I think a far bigger factor in the decrease in low xG shots around 2010 (4:04 graph) was the possession football revolution. Everyone and their nans suddenly wanted to pass and keep the ball, and hopeful longshots are a certain way to lose possession. That Guardiola Barcelona team really did change the tactics of the game in a seismic way. I'm not saying statistic didn't play a part aswell, but even the youngest football fan can appreciate tiki taka, not just mathematicians
@@tendatonda1634 Wtf is a ''tiki-caca''? I hope you're trolling; Go check out the Bundesliga news article that mentions ''Guardiola'' and ''tiki-taka''.
@@funknugget6334 it's very well-known that Guardiola hates the term. Multiple news source over the years have confirmed that. I would've just linked them if TH-cam didn't eat my comments. I'm not bothered as you are it seems.
I’m new to football but as an nba fan this shift lines up so well with our pace and space revolution. The long range mid fielders and midrange specialist are an interesting comparison.
I think it was influenced by soccer. It really began with the Spurs team who had soccer fans as key players (Parker and Manu being French and Argentinian) about a couple of years after the tikitaka Barcelona team mainstreamed the use of space as the dominant tactic in football, between wide players hugging the deadline to maximally stretch the defense, passing to find space for a player to get a better shot, lots of ball movement and off-ball movement, etc. And nowadays sports coaches follow tactics from other sports. Football has also started using screens to shake defenders and using advanced analytics after they became mainstream in basketball
@@StoutProper it's the fact that high xg chances are a more reliable way to win a certain amount of points over a season. If you can train to increase high xg chances then you can more accurately associate that with points. By increasing the volume of low xg chances yes you can increase the likely hood of scoring but they are unreliable over the course of a season. The thing that wasn't touched on is that the defence also know more teams are looking for these high xg chances and are better at defending them. That allows an area 17 player like kdb to worry defences because of his higher than expected conversion rate from there, which leads to more high xg chance development
@@ForzaOwnz sorry, if you increase your number of high xg chances then surely you’d expect to see more goals? Scoring high xg chances doesn’t get you any more points than scoring low xg chances
@@StoutProper it's not better from a direct statistic comparison standpoint its just a more reliable to coach teams to get into higher xg situations For example if you up your .25 xg chances from 4 to 6 per game, you raise your xg from 1 to 1.5. This ability to score 2 goals from the 6 shots every other game is more dependable way to get points out of games over the entire season then attempting to raise your 0.05 xg shots from 20 to 30 per game. So it's not that the math doesn't equal, it does. It's just that investing time into high xg shots seems to be a better investment then attempting so many low xg chances
Then comes Fede Valverde who is in fact encouraged by the coach to take more shots as Anchelotti said that if Fede won't score at least 10 goals he'll rip his coaching license. But this is also the human element as a specific player is being asked to do this not everyone and if his shots will terrorize the opposition then it could lead to him being crowded more and opening opportunities for others ultimately increasing the xG of the team.
As a mathematician, I always talk about putting any statistics in context. Shots from (far) outside the box were always very unlikely, yet through survivorship bias they looked better than they were. Meaning that back in the day when you would see about 2min of every game, logically the 20 garbage shots didn't make the highlights, but the golden one would. Just looking at how abyssmal the probabilities are it's a smart development that managers don't want their players to just roll the dice from 30m out. However, obsession about a singular statistic can often make it misleading. If managers focus everything on maximizing xG, they will create a team with high xG that will very likely underperform. Individual xG is also very iffy - is Haaland a good player because he outperforms his xG? But then what about a player that is able to create many chances but isn't as great at putting them away? He might underperform but still have great stats because he can create chances others can't. As a general rule, having new tools will improve things - but how much they will improve will depend on the people interpreting the tools.
Thank you for this!! Wish I saw this comment before I typed mine out. This new idea of maths and stat data over "flow" and understanding what exactly teams try to do when playing football is why we're beginning to have nba-esque stat porn in football that usually means nothing at the end of the day
Haaland is an anomally, most players are gonna match their xG overtime, the best will slightly overperform overtime, and many will underperform as well.
When looking at someone’s finishing ability it is best to take in all the stats available and measure them against each other. A simplified version of this would be taking a strikers goals scored, their XG and the opposing teams expected goals against based on the players who are playing. A more detailed analysis would have number of shots taken and where from included also.
Your biggest inaccuracy is that elite players just roll the dice, when actually they just believe in themselves in the moment. Sure you can whip a group of sheep into shape on the field, but sheep need sheppards for guidance. To have someone one the field with a vision for the moment before anybody else is a priceless asset.
I believe the reasons for quality of chance creation based on xG even affects players too. As it cascades to their performance and market values. Sometimes strikers are pressured to finish with a limited chances they are getting.
Hey All! Firstly, thanks for the feedback on the video, I can see xG is truly a divisive stat lol. I've removed a small section of the video with some math's calculations, as apparently they were wrong, so thank you to everyone who pointed that out 🙌 I'll double check next time I need to whip out a calculator
Some penalties are still absurd, they change lost balls and unlucky defending for high chances of scoring. If penalties weren't so dominant in xg a lot of shoots would be taken from outside the box and probably more shots in general
We need ai to get rid of double standards and also judge when the player is baiting a foul rather than playing the ball because you see all the time offensive players (or the defense when the fense just turned it over in the final third) have lost control of the ball or let it get far enough away from them where they aren't controlling it but any touch from the defensive team is called a foul. Like hitting the ball so hard it's definitely going to go out just to get contact from the defender, or the defender contacting the offensive player when the latter has let the ball roll a couple yards away and is actually moving backwards into the defender and impeding him should not be called fouls against the defender
xG is just like possession: value that shows dominance of one team over the other. As dominating team is more likely to win it correlates with winning but it is not a cause of the victory. Many teams use more defensive tactics letting the opposition to dominate the game. Keeping the clean sheet with occasional chances is a viable tactics especially for cup competitions.
xG has a much stronger correlation with winning than possession does. One thing I think you might not realise is that defensive, counterattacking football attempts to lower your opposition's xG by denying them chances and increasing your xG by creating really high xG chances on the counter. So defensive football would still aim to improve the net xG between you and your opposition
@@lucasoldrini small thing that i notice sometimes is that a trailing team attempts a lot more shots(in close games) which generally bumps the xG up(however it is generally the team that dominates that has a higher xG)
@@f4k3gamer not actually. xG just get boosted if shots are close and chances are likely to score. A penalty kick is equivalent to 0.84 or something like that, so it's not just like shoot and boost it up. You have to actually get decent chances. If you have a high xG on a match, you definitely had good. chances.
This is interesting. I think part of Dyche's success at Burnley was in forcing opponents to shoot from range or cross the ball. Nick Pope excels at dealing with crosses and stopping long-range shots, and this was a key part of Burnley qualifying for Europe.
As a casual American football fan I became hooked watching compilations of long shots as a kid. I do miss it but the quality of attacks and build ups has become a new reason to watch. I do wonder if this has affected the strategy of set pieces
I think one reason shots from outside the box have become more rare is because the build up play is so god damn slow allowing defenders to get in position well before the ball even makes it to the final third.
One question I have about xG is if it is a living statistic, is it constantly being updated as games are playing. Because xG was originally compiled from data pre-xG, but the fact tactics are now changed to fit xG, like where and when to shoot means that xG will change as a result of its own impact. For example I can imagine in the not too distant future short range attempts and chances will become less successful as teams potentially looks for goalkeepers with pure reflexes and ball distribution over speed, reach, height and defenses will know about how to best defend xG based attacking strategies, thereby creating more quality opportunities and success rates for long shots
Obviously if you are close to the goal and straight in front of it, you are more likely to score. Which is why it is much more heavily defended. The question is, is it balanced. Is the ease at which an opportunity is created in balance with the chance of scoring. If it is, then it doesn't really matter. At the end of the day it probably isn't balanced because when you shoot, most often you lose the ball. Which is why teams would rather work towards a high quality chance for a longer time than to rush things. Most teams will more easily control a game on the ball than off the ball. But there are exceptions of course.
The World Cup graph was really, really interesting to see. I believe Manchester United only recently started hiring a specalist data analyst team - probably explains why we've been so pants in the league! I think xG is also incredibly useful for recruitment, not just tactics. Finding players who can consistently outperform their xG or xA for midfielders/playmakers is one of the reasons why Liverpool have done so well in recruitment in the past 5 years or so, and United not doing it explains why they've splashed big money on absolute flops. Great video.
Interesting, I'm currently watching very little football, but this year I managed to watch a lot of World Cup games. I used to go crazy with the lack of attempts from outside the box, and maybe that could be one of the factors. As he concluded in the end, many of the opportunities I saw the player kicking ended up turning into passes or dribbles which did not result in anything.
There is another element to it, though, which does become apparent - the odds of taking a shot in the first place. Lots of teams try to walk the ball into the net and take fewer shots because of it, so if you can have 10 long shots with 0.15 xG it might make up for having 2 shots of 0.6 xG. It is more midrange than long when you have 0.15 xG, of course, but still.
This is my opinion I hope everyone can accept that I still love football but in terms of viewing, it has severely declined. The true beauty of a "non conventional" has completely vanished. This robotic style is all because of analytics and Pep Gusrdiola the two major culprits in ruining the beautiful game Xg takes away from the game from a viewing perspective. Although being a competitive sport football is also entertainment at the end of the day and it's boring now more than ever
@@warofmankind I think you finding football boring is more about you than the football. The football is better than ever, and its simpler to watch a game than it ever was, with so many countries having access to a great experience. For most of us, 20 years ago we could maybe watch 1-2 games per week.
@@Growlizing True, not to mention while it seems that using statistics to make every decision in a football match is removing the "beauty" of the game, in reality it is just helping teams find the most optimal way to play, and the better and smarter teams are, the better the sport is to watch.
Despite I understand of the inefficiency of the long shot in terms of goal creation, there are some perks of this type of shot not captured by xG. For instance, generally, long shots tend to finish the play, either by scoring, the ball going out of bounds or a good save by the keeper. Either way, this allows the defense to fall back and massively decreases the chances of a counterattack. Also, if you have a good long-distance scorer, will force the defense to spend more resources guarding him, creating more space in the sides and, therefore, allowing your wings to play more freely. I understand the importance of xG but following blindly one statistic without proper knowledge of its implications and what the models are leaving out, that's to say, misinterpreting what the coefficient says and overreaching to conclusions it's a really bad practice.
I think the next natural step in the evolution of midfielders would be a fusion between players like De Bruyne and Valverde. An energetic, well rounded midfielder who can pick a pass into the box but also consistently score screamers from outside. Really interesting video.
I knew it. I noticed a decline in shots from far and people disagreed with me. That's why there are so few amazing long goals but also far fewer horrible long shots.
I think this is why man utd vs arsenal was such a fun game to watch. There were a couple of long shot goals which are so rare to see nowadays. Another good one I think was Reyna in Dortmund vs Augsburg.
Looking at the example of 2 tap ins vs 5 long range shots, it seems like the managers argument for trying to get 2 tap ins as opposed to 5 long range shots would be that the team is more likely to outperform xG with 2 tap ins, but it seems like that argument would need to be made based on personnel. If I've got KDB, then sure two tap ins could be more likely to outperform xG because the passes into that area are so good. But if I've got two strikers who make their long shots more than 20% of the time and dont have a passing artist, then by golly I want my strikers going for it. So in a nutshell it seems like the way for managers to use xG is "how can my team and my players outperform xG." At that point it seems like the statistics really only help to confirm what managers are already thinking based on their intuition, experience, and observation. Lastly, I'd like to know how successful xG has been as a model and whether "how can I outperform xG?" Is a good question to be asking. Some questions to help with this: 1) what is the average distance between xG and goals scored in a game? What percentage of games are within that average? 2) what percentage of games are won by a team outperforming xG? What percentage are lost? 3)what percentage of games are lost by a team underperforming xG? What percentage are won?
The main reason that’d you’d tell your players to go for those 50% tap-ins vs. those 20% long shots is because you think it’d be easier to get 2 of those tap-ins than 5 of those long shots, not because you think your finishing will outperform xG. Getting a ball right in front of the goal is obviously harder than getting it just outside the box, but, with the newly created tool of xG, people have realized that that increased difficulty of getting the shot off is more than outweighted by the higher success rate of the shot. Now, of course, your personnel is going to factor in to this, so if your players are particularly good at converting those long range shots or particularly bad at pushing the ball deep into the box, that’ll change things, but I don’t think that’s the main way that managers use xG.
@@riggsmarkham922 But in the example the xG for the two tap ins is the same as that for the 5 long range shots: 1 expected goal. The difference that he was pointing out was that you were more likely to score 2 goals (i.e., outperform xG) from the 2 tap ins than you are from the 5 long range shots. In other words the total xG has no preference between the 5 long range shots and the 2 tap ins; it's only when you look at trying to outperform the total xG that the difference occurs. It might be the case that it is easier to create 2 tap in shots than it is to create 5 long range shots, but that is not apparent to me and that would definitely depend on personnel and what was worked on in training.
th-cam.com/video/nteN61mfUbo/w-d-xo.html i made this a couple years ago, not sure if I have time to do more small sided tactics, but if there's enough demand I'm sure I could try haha
Just because it can be done, doesn't mean it should be done. Statistics and data-driven football is definitely better for business, which is what football has been for a while now, but maybe I'm a misty-eyed idealist who wants to have their cake and eat it too. I feel that it's important to optimize a team's performance, but without the drama of occasional high-stakes errors, hitting and hoping, and taking unconventional routes to victory, we might as well be watching bowling. (Sorry, bowling. I like you, too.)
Great video! I'm a new subscriber and am going through your back catalogue and learning so much! I'm a West Ham fan and would love for you to do another video on their transformation so I can understand it better please? Because it seems to me that consensus is they were overachieving when they got 6th and 7th but now that they have splashed £170 mill they are now underachieving fighting for relegation. But their tactics are clearly changing between the two since your last video on them. What produced their overachieving success the last two seasons when they had a weaker squad and why are they now underachieving despite having better players? Or have their tactics not changed and was it your video that jinxed them?! Haha No worries if you don't think it would be a good video, I know it might just be for me! You've still found a loyal subscriber in me! Keep up the great work!
05:24 it's true that correlation doesn't always lead to causation. But correlation always has an impact (which could be greater or smaller depending on various factors) on the event.
Man city vs Real Madrid proved zone 14 is quietly becoming underrated and in tight matches might become the difference maker since tightly defended boxes also block the goalies vision.
I agree on the most parts, strongly disagree with the end. Manager's primary job is to overlook his staff, advising them and vice versa and dealing with players, trying to maintain dressing room atmosphere up. Also, xG outliers are too rare to suggest a metric change. Since 2014, Cristiano Ronaldo has scored 0.9 non penalty league goals more than his expected. Not per season, totally(182.1 - 183)
Funny that. Brazilian football has always been suspicious of the long shot. We thought that those 30 yard screamers typical of the English football were a sign of a lack of footballing intelligence. Nice to know that data analysis has proven us right.
The human element is the key at the end. xG might suggest that shooting from outside the box, when there are multiple defenders in front is not a good idea, but Messi is the guy who has succeeded the most in this era and he is also the guy who takes and makes the most shots from outside the box. The guy has 87 open play and 60 free kick goals from outside the box to this day. Similar thing is happening in the NBA, where midrange shots are frowned upon, but still some of the most successful players of this era use it as a major weapon. Most notably Kevin Durant. Talent and skill matter. Maybe the overall trend is governed by the analytics but individuals can do their thing if they are excellent. If Lampard and Gerrard were playing now, they'd still be stars with their playstyle.
I see a metric using xG but also the quantification we do for say penalty chances individually for a player, which makes it a prrsonalised rating and might ve a better one.
I kinda disagree with this notion, because having a threat from long distance forces midfielders to be tighter outside of the box, which leaves more space in between the midfield and defence.
I think your maths is wrong. At 5:32 you say the chance of scoring two long range shots is 4%. This is only true for one specific combination of the 5 shots (in your case the far left and far right shots) - however there are 10 possible combinations of 5 unique shots so in fact it is 10 x 4% = 40%. This is much higher than the 25% for the two tap ins. Sorry if I'm missing something - please correct me if I'm wrong :)
You are right, it's not 4%, but it's actually around 26% chance that you get more than 1 goals with 5 shots each having 20% odds. And it's 25% with 2 shots 50%. So 5 shots are slightly better actually, to get more than 1 goal. But here's the thing. I simulated this behaviour now and over 1,000,000 games there were: Team that did 5 shots won 31.4% games. Marked as team 1 below. Team that did 2 shots won 34.8% games. Marked as team 2 below. And there were 33.7% draws. So despite 5 shots having higher likelihood of scoring more than 1 goal, they still lost overall. That's because they also had higher odds of scoring no goals and the team with 2 shots had higher odds of scoring 1 goal. Here's the full stats of simulation - you can see that team 2 won more mainly because they were more likely to win with 0 - 1. { team1Win: 313036, team2Win: 348888, draw: 338077, team1Goals: { '0': 329076, '1': 409219, '2': 203829, '3': 51095, '4': 6449, '5': 333 }, team2Goals: { '0': 249939, '1': 499886, '2': 250176 }, results: { '1-1': 204528, '0-1': 164157, '1-2': 102450, '2-1': 102336, '1-0': 102241, '0-0': 82638, '0-2': 82281, '2-2': 50911, '2-0': 50582, '3-1': 25521, '3-2': 12874, '3-0': 12700, '4-1': 3188, '4-0': 1682, '4-2': 1579, '5-1': 156, '5-0': 96, '5-2': 81 } }
The correct way to calculate or confirm the odds of scoring at least 2 goals from long shots without simulating. There's binomial distribution for calculating that. The combinations are (12, 13, 14, 15, 23, 24, 25, 34, 35, 45) = 10 which you said, but then the way we would have to add them together would have to be something like: 1. We see what is the odds of getting 1st goal. 2. What are the odds of rest of 4 goals not going in? 3. We see what is the odds of getting 2nd goal after 1st goal had not gone in? 4. What are the odds of rest of 3 goals not going in? We sum all of those odds like: 0.2 * (1- pow(0.8, 4)) = 0.11808 // Odds of scoring a goal and then scoring at least 1 more within 4 chances). 0.2 * (1- pow(0.8, 3)) = 0.0976 // Odds of scoring a goal and then scoring at least 1 more within 3 chances). 0.2 * (1- pow(0.8, 2)) = 0.072 // Same, but 2 chances. 0.2 * (1- pow(0.8, 1)) = 0.04 // Same, but 1 chance. // Now we add all of those together, but we must consider that each next one must be added with considering that previous goals must have not been scored because these had already been accounted for. = 0.11808 + pow(0.8, 1) * 0.0976 + pow(0.8, 2) * 0.072 + pow(0.8, 3) * 0.04 = 0.26272 So the final answer should be 26.272%
The correct way of calculating would be the probability of X≥2 when X follows a binomial distribution with p=0.2 and n=5. This comes out to 26.27%. So there's an error somewhere in your calculation, but the probability is indeed higher than the two short range shots.
Lobanovskiy predicted (and partly pioneered) this shift to a more and more data-driven management approach, ending in algorithms, rather than speculations… in 70s.
I think xG is great to measure the quality and amount of chances a team can create. I don't think is meant to be taken literally it just gives a good idea of the attacking prowess of a team, so managers can meditate on how to approach a game. That's why I don't think robot could replace a manager. A robot would take the xG literally and make decisions based on that, while a human manager can meditate on that statistic and other human aspects of the beautiful gme.
5:44 The probability of making 2 long range shots is 4%, but the probability of making 2 out of 5 long range shots is just over 20%. Still lower than the two tap ins (25%) but not as much lower as seemed to be indicated.
Think of players like Vinicius Junior, almost the opposite of Haaland. He is not the best finisher, but he can create a mess in the defense so that the team takes advantage of it but he will not appear in the goal or assist stats
The thing is if less long shots are taken the more they become unusual. When they become more unusual their success rate will increase. Example, if there are less long shots there is less incentive for defenders to press players 20+m out. Less pressing means more opportunity to strike the ball cleanly and place it out of gk reach.
The defensive strategies improve too so you can't say that the slow growth of goals scored per game means the new offensive strategies don't add much value.
xG is useful, it's just not accurate enough to be used when anything other than very good chances are included, and even then I'm skeptical. As you mentioned some players are better at converting chances anywhere from tap ins to long range shots to free kicks to penalties, and I'd go over step further it also depends on who is in goal trying to stop a certain type of shot. You could also supply the same logic to the defenders closing down a chance, some are better than others. Include that and I'd love to see how the accuracy compares to the current xG. Then again sometimes players have bad days or are out of form whilst other times they on fire and seem to send a worldy every other week, that's the human element. Maybe that needs to be taken into account too!!!
The xG is a measure of the chance an average player in that league would have of scoring the goal (I'm guessing) So, if a striker is outperforming their xG, they are a better striker than the average. If a midfielder has a higher assist count than their xA, then they are a better midfielder than the average in the sample space and so on
The video has flawed maths from 5:17. The xG 1 comes from adding up the chances, which would make the result 1:1 and that correction is correct. But from 5:40 the statement that two 20% chances have only 4% potential to result one goal is incorrect. 2 pieces of 20% chances have 40% chance of scoring. The 4% (i.e. multiplying and not adding up the chances) means you score two in a row. So the 0.032% means scoring 5 in a row from 20% chances, and that is compared to scoring 2 in a row from 50% chances (25)%. So what you are comparing is the chances of scoring 5 in a game vs scoring 2 in a game and not the expected end result, which is in fact 1:1. For the end result you cant apply conditional probability (that is scoring from a second, third, etc...20% chance PROVIDING there was a score from a first, second, third, etc.) because these events are theoretically independent from one another. If you score from a 20% chance that doesn't make the second long range shot's chance 4%, it will be still 20%. It's like the textbook example of rolling with the dice, if you roll a 6 for the first try its 1/6, and if you try the 2nd time it's still 1/6, despite that the chance of 2 in a row is 1/36. This error in the calculation makes the conclusion of the video (and basically the whole video) invalid. There are other factors tho, which make these events dependent from one another meaning that the whole idea of relying on made up statistics is a bogus. For example if someone scores from a 50% chance from the 6 yard box the manager may take measures against that (like parking the bus into the penalty area) making the 50% worth 20%, and bumping up the chances for a long shot goal to 50% and the whole situation is reversed. Football is a much more complicated game than what these very simplistic statistical models describe...
What if the rules would be changed. Like a goal from outside the box counts more than inside, similar to basketball. Or a penalty count less than a regular goal. This changes seem drastic but I believe it would make soccer much more attractive.
"A robot will replace the manager 20 years from now". I actually think it's gonna happen a lot sooner than that. I think managers will basically be hired for PR and stuff while actual strategic and tactical decisions will be taken by advanced AI. Also, Wenger can take solace in knowing that all of us in every profession are gonna be replaced by robots until every process in every industry is fully automated. We can then look forward to a future where Elon Musk's son Kyle will hunt us for sport.
The thought of this makes me quite sad. It’s almost like the games themselves will just be played as a formality, but the results will be predetermined by increasingly advanced AI. For the moment, I still believe in the ability of the sport to surprise. But that in itself is a problem, because every shock upset includes not just a surprise win but also a surprise loss. And for those players that are being compared to their AI versions and failing to measure up, it creates even more pressure than already exists. Every player will feel the pressure to become a literal robot. Not just the managers.
What would be good to know is what is the goal conversion rate of shots taken from zone 13, 14 and 15. My bet is that it is even at EPL league level not even 5% chance of scoring from that distance.
For such a dynamic sport, these new advanced metrics are still too raw to decently describe particular scenarios. While I welcome them, there is still a whole lot of polishing that needs to happen before they begin to matter (Ex: variables comes to mind .. too many to name here and when you start seeing how they impact each other you instantly reach infinity). I recon we need a few more years before xG can be a reliable metric ... and I hope we dont lose the long shot because of it!
For the quality of the chance the quality of the shooter is irrelevant, so keep it out of XG. But if you want to know how well a team performed, instead of just adding up both teams XGs, the attempts should be multiplied by XG modifiers of individual shooters. How to best calculate that value is a bit tricky since some players take very few shots and player abilities also change, but considering that for Haaland that value is around 1.5, you are certainly missing something if you ignore it. It simply doesn't make sense to say that "City were lucky again".
Surely a player's xg is a bit like wins above replacement in baseball. It's what the average player would do in that situation - so Haaland essentially has 6 (??) more goals than a league average striker would.
The English Premiership may be more 'skillful' than other UK leagues, but repeat this video for the EFL Championship and holes are blown in the stats as Midfielders are taking as many and often more shots than the Centre Forwards and plenty are from the edge of, or outside the penalty area. Middlesborough (26%) and Bristol City (29%) are the lowest number of shots from outside the box, 6 teams have 40% or more shots outside the box with Cardiff (47%) leading. I'm not a purist (nor do I like chess) so I prefer defenders and midfielders going in for tackles rather than standing off, and I like that I still get to spray mouthfuls of coffee around when balls burst the net from 30 yards 😀
It's not really xg . It's more the way the game is played . Teams always press out now instead of sitting back and its made scoring longshots way harder. That's why outside the box used to be such a great area . You would get time to make a shot , now players will actively press you before the ball even arrives to feet.
5:40 The math is misleading. The probability, to score AT LEAST 2 goals from those 5 long shots is 26.27%. This is the probability that X≥2 when X follows a binomial distribution with p=0.2 and n=5. The probability to score 2 goals exactly would be 20.48%, to score 3 or more 5.79%. Source: Me, a math student
xG stats in their present state are one of the most misused metrics. If you look at the definition at 1:57 that is what the stat represents. Once a real goal is scored, as in the game-state changes, xG accumulation is effected and not accounted for in the stat. Ex: Teams accumulate xG differently when they are up or down a goal. The worst offense to the stat is when xG is tallied up across multiple matches. It’s so far removed from what xG stats do represent that anyone with a sense for mathematics recognize that you’re not looking at anything of value.
You’re right in a sense, that teams behind will typically produce more xG than otherwise, but over large data sets it’s been proven that xG is a much more reliable predictor of future success than expert opinion or any other metric even if you don’t take the scoreline into account. I do think the sort of thing you’re talking about, will be something studied a lot in the coming years with teams seeing how much xG difference they can sacrifice in order to increase their chances of winning by playing more attacking/defensive than is optimal
xG is definitely a really influential stat, and correct me if I’m wrong, but it doesn’t take into account the skillset of individual players right? So it’s misleading to give a blanket xG for certain shots when really the outcome can be expected to vary wildly depending on the player
Fear not, guys. My FIFA team is still on my knowledge of hockey. Shots from deep on the net hoping it goes in. I've scored some fucking beauties from out deep with Kimmich.
most likely XG is calculated per player now than per average in football by the teams. My guess is that Man city calculates higher xg for halland than average xg for striker. Also some players could have higher xg outside the box than others.
The pure stats give you a lot of information, what's missing is the phyological side of all professional sports, being physically fit, technique and tactics take you so for but just as important is the metal motivation and team spirit, a lot of player's play for the manager they like and respect, Don't think a robot is going to get that any time soon 😜
Forgot to say how much the crowd can influence games, that's why most have better home records than away, last year in the Premier league Arsenal were going for CL top 4 place against my team Newcastle, the fans put on a Fantastic display with Wor Flags before the game the crowd were well up and very intimidating Arsenal had a much better team on paper, we looked like going down at Christmas, and yet we hammered them 2-0 never underestimate the phyological effects great support can have on teams.
Returning back to watch some football in this world cup and initially confused by the xG and other stats (stopped watching in 2010) I do think these advanced stats are useful, but there are some things that you cannot measure by stats. This is a sport played by humans, not robots. Not everything has to be measured by stats. I do agree that sooner or later, football teams would probably won't need a coach and instead will employ AI instead. Which means the beauty and unpredictability of the game will be dead. Like, any coach with a healthy mind will always allow Roberto Carlos to take a long shot from the left. That might not happen nowadays.
the graph at 4:21 it has values 0.00 and up and it's talking about number of shots decreasing compared to amount of xg increasing so can I assume the value like 0.20 in terms of xg is 20% chance for a goal compared to shots which logically will be 20 shots because I still can't see those two values being in a single graph. a team can't have 0.20 shots that's less than 1 which conveys less than a shot it's imposable to have less than a shot. Can somebody please clarify ?
While I can see the value of the xG statistic, it relies on many factors that are difficult to quantify. Furthermore, like you, I like seeing shots from distance in appropriate circumstances. There are times when a pass is the best option, but a team must shoot to score.
Think there's definitely an argument that could say that all of what you say is more linked to the emergence of tiki taka playstyle and the fact that 80% of teams tried to copy fcb of guardiola
Good Longshooters are not obsolete. You show them that you can take scary longshots to pull apart defenses, next time you are in that spot. Goal or no it puts a bad idea in the head of more instinctual players. You want them to close you down and open up gaps behind them.
I was on soccerment the other day trying to get ppda data for a data viz I was doing but couldn't download it so I had to manually enter it into my csv(fortunately it was only la liga data) Any tips ?
Up to now i tell my friends they have changed the weight of the ball because it looks like its hard to deep the ball from a long short like back in the days of the Lampards this era i give you Pogba some of his shorts are boeing 707 , just my myths😀
Xg hasnt killed the long shot at all, it may have hindered it yes, but xgot (expected goals in target) has brought it back, i’ll give the example of khvicha kvaratskhelia, pretty much every long shot he takes has over 0.5xgot as long as it is on target.
No one here that considers the football aerodynamics to be a reason why long shots aren't a thing anymore? Ball has changed ever since that Jabulani disaster
@@FootballMeta okay no problem I respect your opinion or reply I am interested to see when did he said this or show me a particular video Mourinho complaining that his XG is best
![กี่หมื่นฟ้า | Your Sky Series EP.1 [1/4]](http://i.ytimg.com/vi/vLGjxFL6U_I/mqdefault.jpg)
![[LIVE] : ONE ลุมพินี 87 วันนี้!! คู่เอก "ก้องชัย vs โชคปรีชา"](http://i.ytimg.com/vi/YN6e7HVnekw/mqdefault.jpg)
I think a far bigger factor in the decrease in low xG shots around 2010 (4:04 graph) was the possession football revolution.
Everyone and their nans suddenly wanted to pass and keep the ball, and hopeful longshots are a certain way to lose possession.
That Guardiola Barcelona team really did change the tactics of the game in a seismic way.
I'm not saying statistic didn't play a part aswell, but even the youngest football fan can appreciate tiki taka, not just mathematicians
all is well but please don't use the term "tiki-taka", it represents the type of time-wasting possession that even Pep himself hates
@@reintaler6355 no that's called "tiki-caca" tiki taka is offensive progressive possession.
@@tendatonda1634 Wtf is a ''tiki-caca''? I hope you're trolling; Go check out the Bundesliga news article that mentions ''Guardiola'' and ''tiki-taka''.
@@reintaler6355 the bundesliga news article 😭 you read a random ass article and we’re so bothered by this you have to correct him
@@funknugget6334 it's very well-known that Guardiola hates the term. Multiple news source over the years have confirmed that. I would've just linked them if TH-cam didn't eat my comments.
I'm not bothered as you are it seems.
I’m new to football but as an nba fan this shift lines up so well with our pace and space revolution. The long range mid fielders and midrange specialist are an interesting comparison.
I think it was influenced by soccer. It really began with the Spurs team who had soccer fans as key players (Parker and Manu being French and Argentinian) about a couple of years after the tikitaka Barcelona team mainstreamed the use of space as the dominant tactic in football, between wide players hugging the deadline to maximally stretch the defense, passing to find space for a player to get a better shot, lots of ball movement and off-ball movement, etc. And nowadays sports coaches follow tactics from other sports. Football has also started using screens to shake defenders and using advanced analytics after they became mainstream in basketball
If xg is so significant why aren’t we seeing more goals being scored than in the past? Or are we?
@@StoutProper it's the fact that high xg chances are a more reliable way to win a certain amount of points over a season. If you can train to increase high xg chances then you can more accurately associate that with points.
By increasing the volume of low xg chances yes you can increase the likely hood of scoring but they are unreliable over the course of a season.
The thing that wasn't touched on is that the defence also know more teams are looking for these high xg chances and are better at defending them. That allows an area 17 player like kdb to worry defences because of his higher than expected conversion rate from there, which leads to more high xg chance development
@@ForzaOwnz sorry, if you increase your number of high xg chances then surely you’d expect to see more goals? Scoring high xg chances doesn’t get you any more points than scoring low xg chances
@@StoutProper it's not better from a direct statistic comparison standpoint its just a more reliable to coach teams to get into higher xg situations
For example if you up your .25 xg chances from 4 to 6 per game, you raise your xg from 1 to 1.5. This ability to score 2 goals from the 6 shots every other game is more dependable way to get points out of games over the entire season then attempting to raise your 0.05 xg shots from 20 to 30 per game.
So it's not that the math doesn't equal, it does. It's just that investing time into high xg shots seems to be a better investment then attempting so many low xg chances
Then comes Fede Valverde who is in fact encouraged by the coach to take more shots as Anchelotti said that if Fede won't score at least 10 goals he'll rip his coaching license. But this is also the human element as a specific player is being asked to do this not everyone and if his shots will terrorize the opposition then it could lead to him being crowded more and opening opportunities for others ultimately increasing the xG of the team.
As a mathematician, I always talk about putting any statistics in context. Shots from (far) outside the box were always very unlikely, yet through survivorship bias they looked better than they were. Meaning that back in the day when you would see about 2min of every game, logically the 20 garbage shots didn't make the highlights, but the golden one would. Just looking at how abyssmal the probabilities are it's a smart development that managers don't want their players to just roll the dice from 30m out.
However, obsession about a singular statistic can often make it misleading. If managers focus everything on maximizing xG, they will create a team with high xG that will very likely underperform. Individual xG is also very iffy - is Haaland a good player because he outperforms his xG? But then what about a player that is able to create many chances but isn't as great at putting them away? He might underperform but still have great stats because he can create chances others can't.
As a general rule, having new tools will improve things - but how much they will improve will depend on the people interpreting the tools.
Thank you for this!! Wish I saw this comment before I typed mine out. This new idea of maths and stat data over "flow" and understanding what exactly teams try to do when playing football is why we're beginning to have nba-esque stat porn in football that usually means nothing at the end of the day
Haaland is an anomally, most players are gonna match their xG overtime, the best will slightly overperform overtime, and many will underperform as well.
When looking at someone’s finishing ability it is best to take in all the stats available and measure them against each other. A simplified version of this would be taking a strikers goals scored, their XG and the opposing teams expected goals against based on the players who are playing. A more detailed analysis would have number of shots taken and where from included also.
They do measure xA.
Your biggest inaccuracy is that elite players just roll the dice, when actually they just believe in themselves in the moment. Sure you can whip a group of sheep into shape on the field, but sheep need sheppards for guidance. To have someone one the field with a vision for the moment before anybody else is a priceless asset.
I believe the reasons for quality of chance creation based on xG even affects players too. As it cascades to their performance and market values. Sometimes strikers are pressured to finish with a limited chances they are getting.
Hey All! Firstly, thanks for the feedback on the video, I can see xG is truly a divisive stat lol. I've removed a small section of the video with some math's calculations, as apparently they were wrong, so thank you to everyone who pointed that out 🙌 I'll double check next time I need to whip out a calculator
Mistakes happen, I respect the way you've dealt with it!
Some penalties are still absurd, they change lost balls and unlucky defending for high chances of scoring.
If penalties weren't so dominant in xg a lot of shoots would be taken from outside the box and probably more shots in general
But there is an xg minus penalties stats out there, the xgot is also examined
@@Braziliansyrah agreed, but i was talking about the outcome of the games, suppose your teams loses a final because of a really dubious penalty...
I'm sure analysts discount penalty goals while also accounting for the chance being created at the moment of foul inside the box
We need ai to get rid of double standards and also judge when the player is baiting a foul rather than playing the ball because you see all the time offensive players (or the defense when the fense just turned it over in the final third) have lost control of the ball or let it get far enough away from them where they aren't controlling it but any touch from the defensive team is called a foul. Like hitting the ball so hard it's definitely going to go out just to get contact from the defender, or the defender contacting the offensive player when the latter has let the ball roll a couple yards away and is actually moving backwards into the defender and impeding him should not be called fouls against the defender
xG is just like possession: value that shows dominance of one team over the other. As dominating team is more likely to win it correlates with winning but it is not a cause of the victory. Many teams use more defensive tactics letting the opposition to dominate the game. Keeping the clean sheet with occasional chances is a viable tactics especially for cup competitions.
Teams with higher xG commonly win way more games
@@lucasoldrini I agree with you. Im trying to point out that teams that dominate are more likely to win but other tactics were also successful.
xG has a much stronger correlation with winning than possession does.
One thing I think you might not realise is that defensive, counterattacking football attempts to lower your opposition's xG by denying them chances and increasing your xG by creating really high xG chances on the counter. So defensive football would still aim to improve the net xG between you and your opposition
@@lucasoldrini small thing that i notice sometimes is that a trailing team attempts a lot more shots(in close games) which generally bumps the xG up(however it is generally the team that dominates that has a higher xG)
@@f4k3gamer not actually.
xG just get boosted if shots are close and chances are likely to score. A penalty kick is equivalent to 0.84 or something like that, so it's not just like shoot and boost it up. You have to actually get decent chances. If you have a high xG on a match, you definitely had good. chances.
This is interesting. I think part of Dyche's success at Burnley was in forcing opponents to shoot from range or cross the ball. Nick Pope excels at dealing with crosses and stopping long-range shots, and this was a key part of Burnley qualifying for Europe.
100%
As a casual American football fan I became hooked watching compilations of long shots as a kid. I do miss it but the quality of attacks and build ups has become a new reason to watch. I do wonder if this has affected the strategy of set pieces
Look up James ward-prowse. Plays for bottom of the league but consistently scores 20 yard free kicks
I think one reason shots from outside the box have become more rare is because the build up play is so god damn slow allowing defenders to get in position well before the ball even makes it to the final third.
Can you do more individual tactics videos?Like based on position?
One question I have about xG is if it is a living statistic, is it constantly being updated as games are playing. Because xG was originally compiled from data pre-xG, but the fact tactics are now changed to fit xG, like where and when to shoot means that xG will change as a result of its own impact. For example I can imagine in the not too distant future short range attempts and chances will become less successful as teams potentially looks for goalkeepers with pure reflexes and ball distribution over speed, reach, height and defenses will know about how to best defend xG based attacking strategies, thereby creating more quality opportunities and success rates for long shots
Obviously if you are close to the goal and straight in front of it, you are more likely to score. Which is why it is much more heavily defended.
The question is, is it balanced. Is the ease at which an opportunity is created in balance with the chance of scoring. If it is, then it doesn't really matter.
At the end of the day it probably isn't balanced because when you shoot, most often you lose the ball. Which is why teams would rather work towards a high quality chance for a longer time than to rush things. Most teams will more easily control a game on the ball than off the ball. But there are exceptions of course.
The World Cup graph was really, really interesting to see. I believe Manchester United only recently started hiring a specalist data analyst team - probably explains why we've been so pants in the league!
I think xG is also incredibly useful for recruitment, not just tactics. Finding players who can consistently outperform their xG or xA for midfielders/playmakers is one of the reasons why Liverpool have done so well in recruitment in the past 5 years or so, and United not doing it explains why they've splashed big money on absolute flops.
Great video.
I remember when goals just outside the box wouldn't be taken as a long shot now thats what so many compilations have
Interesting, I'm currently watching very little football, but this year I managed to watch a lot of World Cup games.
I used to go crazy with the lack of attempts from outside the box, and maybe that could be one of the factors.
As he concluded in the end, many of the opportunities I saw the player kicking ended up turning into passes or dribbles which did not result in anything.
There is another element to it, though, which does become apparent - the odds of taking a shot in the first place. Lots of teams try to walk the ball into the net and take fewer shots because of it, so if you can have 10 long shots with 0.15 xG it might make up for having 2 shots of 0.6 xG. It is more midrange than long when you have 0.15 xG, of course, but still.
I love xG - it makes so much sense. xG makes sense of league performance over time, but is perhaps less useful in knock out style tournaments.
This is my opinion I hope everyone can accept that I still love football but in terms of viewing, it has severely declined. The true beauty of a "non conventional" has completely vanished. This robotic style is all because of analytics and Pep Gusrdiola the two major culprits in ruining the beautiful game
Xg takes away from the game from a viewing perspective. Although being a competitive sport football is also entertainment at the end of the day and it's boring now more than ever
U just gotta look at it differently
@@warofmankind I think you finding football boring is more about you than the football. The football is better than ever, and its simpler to watch a game than it ever was, with so many countries having access to a great experience. For most of us, 20 years ago we could maybe watch 1-2 games per week.
@@Growlizing True, not to mention while it seems that using statistics to make every decision in a football match is removing the "beauty" of the game, in reality it is just helping teams find the most optimal way to play, and the better and smarter teams are, the better the sport is to watch.
Despite I understand of the inefficiency of the long shot in terms of goal creation, there are some perks of this type of shot not captured by xG. For instance, generally, long shots tend to finish the play, either by scoring, the ball going out of bounds or a good save by the keeper. Either way, this allows the defense to fall back and massively decreases the chances of a counterattack. Also, if you have a good long-distance scorer, will force the defense to spend more resources guarding him, creating more space in the sides and, therefore, allowing your wings to play more freely. I understand the importance of xG but following blindly one statistic without proper knowledge of its implications and what the models are leaving out, that's to say, misinterpreting what the coefficient says and overreaching to conclusions it's a really bad practice.
It also opens the door to counter attacks, so it's not always a "good thing".
There's no tactic with no downside @@jimmycrackcorn99
I think the next natural step in the evolution of midfielders would be a fusion between players like De Bruyne and Valverde. An energetic, well rounded midfielder who can pick a pass into the box but also consistently score screamers from outside. Really interesting video.
Ie what Pogba could have been
I would switch out KDB with Bruno Fernandes
Dude has 0.59 GA contribution compared to KDB's 0.61
Remember KDB has much better attack line to work with
@@rooftopastronomer2697 better teams doesn't mean higher GA contribution. Just look at Grealish.
Depends on a players role.
Valverde has very very low chance of scoring from outside
Bro bro bro bro that's illegal, "messi or Maguire"😂😂😂
I knew it. I noticed a decline in shots from far and people disagreed with me. That's why there are so few amazing long goals but also far fewer horrible long shots.
I think this is why man utd vs arsenal was such a fun game to watch. There were a couple of long shot goals which are so rare to see nowadays. Another good one I think was Reyna in Dortmund vs Augsburg.
Looking at the example of 2 tap ins vs 5 long range shots, it seems like the managers argument for trying to get 2 tap ins as opposed to 5 long range shots would be that the team is more likely to outperform xG with 2 tap ins, but it seems like that argument would need to be made based on personnel. If I've got KDB, then sure two tap ins could be more likely to outperform xG because the passes into that area are so good. But if I've got two strikers who make their long shots more than 20% of the time and dont have a passing artist, then by golly I want my strikers going for it.
So in a nutshell it seems like the way for managers to use xG is "how can my team and my players outperform xG." At that point it seems like the statistics really only help to confirm what managers are already thinking based on their intuition, experience, and observation.
Lastly, I'd like to know how successful xG has been as a model and whether "how can I outperform xG?" Is a good question to be asking. Some questions to help with this:
1) what is the average distance between xG and goals scored in a game? What percentage of games are within that average?
2) what percentage of games are won by a team outperforming xG? What percentage are lost?
3)what percentage of games are lost by a team underperforming xG? What percentage are won?
The main reason that’d you’d tell your players to go for those 50% tap-ins vs. those 20% long shots is because you think it’d be easier to get 2 of those tap-ins than 5 of those long shots, not because you think your finishing will outperform xG.
Getting a ball right in front of the goal is obviously harder than getting it just outside the box, but, with the newly created tool of xG, people have realized that that increased difficulty of getting the shot off is more than outweighted by the higher success rate of the shot.
Now, of course, your personnel is going to factor in to this, so if your players are particularly good at converting those long range shots or particularly bad at pushing the ball deep into the box, that’ll change things, but I don’t think that’s the main way that managers use xG.
@@riggsmarkham922 But in the example the xG for the two tap ins is the same as that for the 5 long range shots: 1 expected goal. The difference that he was pointing out was that you were more likely to score 2 goals (i.e., outperform xG) from the 2 tap ins than you are from the 5 long range shots. In other words the total xG has no preference between the 5 long range shots and the 2 tap ins; it's only when you look at trying to outperform the total xG that the difference occurs.
It might be the case that it is easier to create 2 tap in shots than it is to create 5 long range shots, but that is not apparent to me and that would definitely depend on personnel and what was worked on in training.
Happy New Year!! Can you do 4v4 with keepers? Formation and Tatics?
th-cam.com/video/nteN61mfUbo/w-d-xo.html i made this a couple years ago, not sure if I have time to do more small sided tactics, but if there's enough demand I'm sure I could try haha
Just because it can be done, doesn't mean it should be done. Statistics and data-driven football is definitely better for business, which is what football has been for a while now, but maybe I'm a misty-eyed idealist who wants to have their cake and eat it too. I feel that it's important to optimize a team's performance, but without the drama of occasional high-stakes errors, hitting and hoping, and taking unconventional routes to victory, we might as well be watching bowling. (Sorry, bowling. I like you, too.)
Exactly I want good ol' honest football back instead of this robotic bullshit
“Robotic” would win u more games.
@@connoryoung8951 unless the other team are better robots than you are
I am happy to remember when football was played with a ball.
Football took a while to hop on the analytics train but it has gotten there now.
Rudiger is the only player that doesn't care about xG and shoots whenever he feels like it.
The quantity of knowlege i am allowed to get in 5min in this Channel.. greatttttt!!
Happy New year, and God Prosper you!!
Thanks man happy new year to you too 🙌🏻
Great video! I'm a new subscriber and am going through your back catalogue and learning so much! I'm a West Ham fan and would love for you to do another video on their transformation so I can understand it better please? Because it seems to me that consensus is they were overachieving when they got 6th and 7th but now that they have splashed £170 mill they are now underachieving fighting for relegation. But their tactics are clearly changing between the two since your last video on them. What produced their overachieving success the last two seasons when they had a weaker squad and why are they now underachieving despite having better players? Or have their tactics not changed and was it your video that jinxed them?! Haha No worries if you don't think it would be a good video, I know it might just be for me! You've still found a loyal subscriber in me! Keep up the great work!
4:33 Mate, your y-axis has to start at 0.
Edit 1: Same at 5:11.
Edit 2: Again at 7:22.
You should do a deep dive into each zone
05:24 it's true that correlation doesn't always lead to causation. But correlation always has an impact (which could be greater or smaller depending on various factors) on the event.
Man city vs Real Madrid proved zone 14 is quietly becoming underrated and in tight matches might become the difference maker since tightly defended boxes also block the goalies vision.
I agree on the most parts, strongly disagree with the end. Manager's primary job is to overlook his staff, advising them and vice versa and dealing with players, trying to maintain dressing room atmosphere up. Also, xG outliers are too rare to suggest a metric change. Since 2014, Cristiano Ronaldo has scored 0.9 non penalty league goals more than his expected. Not per season, totally(182.1 - 183)
xG fails to take into consideration if It's Mguire shooting versus David De Gea, at which point it's over 9000.
Funny that. Brazilian football has always been suspicious of the long shot. We thought that those 30 yard screamers typical of the English football were a sign of a lack of footballing intelligence. Nice to know that data analysis has proven us right.
The siblings of the target man number 9 and the creative No. 10 have also vanished because of this point. Great analysis.
Future stadium chants to robot managers :”You’re just a mincer, you’re just a mincer...” 😂😮
The human element is the key at the end. xG might suggest that shooting from outside the box, when there are multiple defenders in front is not a good idea, but Messi is the guy who has succeeded the most in this era and he is also the guy who takes and makes the most shots from outside the box. The guy has 87 open play and 60 free kick goals from outside the box to this day.
Similar thing is happening in the NBA, where midrange shots are frowned upon, but still some of the most successful players of this era use it as a major weapon. Most notably Kevin Durant.
Talent and skill matter. Maybe the overall trend is governed by the analytics but individuals can do their thing if they are excellent.
If Lampard and Gerrard were playing now, they'd still be stars with their playstyle.
Alfred Marshall has entered my beautiful game and has made it arithmetic
I see a metric using xG but also the quantification we do for say penalty chances individually for a player, which makes it a prrsonalised rating and might ve a better one.
I kinda disagree with this notion, because having a threat from long distance forces midfielders to be tighter outside of the box, which leaves more space in between the midfield and defence.
I think your maths is wrong. At 5:32 you say the chance of scoring two long range shots is 4%. This is only true for one specific combination of the 5 shots (in your case the far left and far right shots) - however there are 10 possible combinations of 5 unique shots so in fact it is 10 x 4% = 40%. This is much higher than the 25% for the two tap ins. Sorry if I'm missing something - please correct me if I'm wrong :)
You are right, it's not 4%, but it's actually around 26% chance that you get more than 1 goals with 5 shots each having 20% odds. And it's 25% with 2 shots 50%.
So 5 shots are slightly better actually, to get more than 1 goal.
But here's the thing. I simulated this behaviour now and over 1,000,000 games there were:
Team that did 5 shots won 31.4% games. Marked as team 1 below.
Team that did 2 shots won 34.8% games. Marked as team 2 below.
And there were 33.7% draws.
So despite 5 shots having higher likelihood of scoring more than 1 goal, they still lost overall. That's because they also had higher odds of scoring no goals and the team with 2 shots had higher odds of scoring 1 goal. Here's the full stats of simulation - you can see that team 2 won more mainly because they were more likely to win with 0 - 1.
{
team1Win: 313036,
team2Win: 348888,
draw: 338077,
team1Goals: {
'0': 329076,
'1': 409219,
'2': 203829,
'3': 51095,
'4': 6449,
'5': 333
},
team2Goals: { '0': 249939, '1': 499886, '2': 250176 },
results: {
'1-1': 204528,
'0-1': 164157,
'1-2': 102450,
'2-1': 102336,
'1-0': 102241,
'0-0': 82638,
'0-2': 82281,
'2-2': 50911,
'2-0': 50582,
'3-1': 25521,
'3-2': 12874,
'3-0': 12700,
'4-1': 3188,
'4-0': 1682,
'4-2': 1579,
'5-1': 156,
'5-0': 96,
'5-2': 81
}
}
The correct way to calculate or confirm the odds of scoring at least 2 goals from long shots without simulating. There's binomial distribution for calculating that.
The combinations are (12, 13, 14, 15, 23, 24, 25, 34, 35, 45) = 10 which you said, but then the way we would have to add them together would have to be something like:
1. We see what is the odds of getting 1st goal.
2. What are the odds of rest of 4 goals not going in?
3. We see what is the odds of getting 2nd goal after 1st goal had not gone in?
4. What are the odds of rest of 3 goals not going in?
We sum all of those odds like:
0.2 * (1- pow(0.8, 4)) = 0.11808 // Odds of scoring a goal and then scoring at least 1 more within 4 chances).
0.2 * (1- pow(0.8, 3)) = 0.0976 // Odds of scoring a goal and then scoring at least 1 more within 3 chances).
0.2 * (1- pow(0.8, 2)) = 0.072 // Same, but 2 chances.
0.2 * (1- pow(0.8, 1)) = 0.04 // Same, but 1 chance.
// Now we add all of those together, but we must consider that each next one must be added with considering that previous goals must have not been scored because these had already been accounted for.
= 0.11808 + pow(0.8, 1) * 0.0976 + pow(0.8, 2) * 0.072 + pow(0.8, 3) * 0.04 = 0.26272
So the final answer should be 26.272%
Thank you, I definitely missed something and oversimplified it, I'll double check next time 😅
The correct way of calculating would be the probability of X≥2 when X follows a binomial distribution with p=0.2 and n=5. This comes out to 26.27%.
So there's an error somewhere in your calculation, but the probability is indeed higher than the two short range shots.
@@someverycool4552 That's pretty neat. I guess it must have something to do with the shape of the distributions.
Lobanovskiy predicted (and partly pioneered) this shift to a more and more data-driven management approach, ending in algorithms, rather than speculations… in 70s.
3:40 Somehow at least in this context it seemed that Expected Points made sense, because Arsenal missed out on the title.
I think xG is great to measure the quality and amount of chances a team can create. I don't think is meant to be taken literally it just gives a good idea of the attacking prowess of a team, so managers can meditate on how to approach a game. That's why I don't think robot could replace a manager. A robot would take the xG literally and make decisions based on that, while a human manager can meditate on that statistic and other human aspects of the beautiful gme.
The difference between a shot XG and total XG. You are probably more likely to score from 1 0.7 than you are 10 0.08s
5:44 The probability of making 2 long range shots is 4%, but the probability of making 2 out of 5 long range shots is just over 20%. Still lower than the two tap ins (25%) but not as much lower as seemed to be indicated.
It's not lower if you also consider the possibility of scoring 3 or more
Love from Bangladesh! I love your Videos💗
Think of players like Vinicius Junior, almost the opposite of Haaland. He is not the best finisher, but he can create a mess in the defense so that the team takes advantage of it but he will not appear in the goal or assist stats
The thing is if less long shots are taken the more they become unusual. When they become more unusual their success rate will increase.
Example, if there are less long shots there is less incentive for defenders to press players 20+m out. Less pressing means more opportunity to strike the ball cleanly and place it out of gk reach.
I've been saying this to mates for a good couple of seasons and finally a youtube video just spells it out.
The defensive strategies improve too so you can't say that the slow growth of goals scored per game means the new offensive strategies don't add much value.
xG is useful, it's just not accurate enough to be used when anything other than very good chances are included, and even then I'm skeptical. As you mentioned some players are better at converting chances anywhere from tap ins to long range shots to free kicks to penalties, and I'd go over step further it also depends on who is in goal trying to stop a certain type of shot. You could also supply the same logic to the defenders closing down a chance, some are better than others. Include that and I'd love to see how the accuracy compares to the current xG. Then again sometimes players have bad days or are out of form whilst other times they on fire and seem to send a worldy every other week, that's the human element. Maybe that needs to be taken into account too!!!
The xG is a measure of the chance an average player in that league would have of scoring the goal (I'm guessing)
So, if a striker is outperforming their xG, they are a better striker than the average. If a midfielder has a higher assist count than their xA, then they are a better midfielder than the average in the sample space and so on
@@Someone25948 nah xA doesnt work like that, it means they have better finishers than average to convert the chances they created
The video has flawed maths from 5:17. The xG 1 comes from adding up the chances, which would make the result 1:1 and that correction is correct. But from 5:40 the statement that two 20% chances have only 4% potential to result one goal is incorrect. 2 pieces of 20% chances have 40% chance of scoring. The 4% (i.e. multiplying and not adding up the chances) means you score two in a row. So the 0.032% means scoring 5 in a row from 20% chances, and that is compared to scoring 2 in a row from 50% chances (25)%. So what you are comparing is the chances of scoring 5 in a game vs scoring 2 in a game and not the expected end result, which is in fact 1:1.
For the end result you cant apply conditional probability (that is scoring from a second, third, etc...20% chance PROVIDING there was a score from a first, second, third, etc.) because these events are theoretically independent from one another. If you score from a 20% chance that doesn't make the second long range shot's chance 4%, it will be still 20%. It's like the textbook example of rolling with the dice, if you roll a 6 for the first try its 1/6, and if you try the 2nd time it's still 1/6, despite that the chance of 2 in a row is 1/36.
This error in the calculation makes the conclusion of the video (and basically the whole video) invalid.
There are other factors tho, which make these events dependent from one another meaning that the whole idea of relying on made up statistics is a bogus. For example if someone scores from a 50% chance from the 6 yard box the manager may take measures against that (like parking the bus into the penalty area) making the 50% worth 20%, and bumping up the chances for a long shot goal to 50% and the whole situation is reversed. Football is a much more complicated game than what these very simplistic statistical models describe...
Hello Human,
I am XEr-2000, an OpenAI football manager. Could you please do a video about the 01101110111011 formation?
Ah yes, a classic! Just you wait until 2036 that formation will dominate the game
What if the rules would be changed. Like a goal from outside the box counts more than inside, similar to basketball. Or a penalty count less than a regular goal. This changes seem drastic but I believe it would make soccer much more attractive.
"A robot will replace the manager 20 years from now".
I actually think it's gonna happen a lot sooner than that. I think managers will basically be hired for PR and stuff while actual strategic and tactical decisions will be taken by advanced AI.
Also, Wenger can take solace in knowing that all of us in every profession are gonna be replaced by robots until every process in every industry is fully automated.
We can then look forward to a future where Elon Musk's son Kyle will hunt us for sport.
The thought of this makes me quite sad. It’s almost like the games themselves will just be played as a formality, but the results will be predetermined by increasingly advanced AI. For the moment, I still believe in the ability of the sport to surprise. But that in itself is a problem, because every shock upset includes not just a surprise win but also a surprise loss. And for those players that are being compared to their AI versions and failing to measure up, it creates even more pressure than already exists. Every player will feel the pressure to become a literal robot. Not just the managers.
this channel is bloody brilliant.
What would be good to know is what is the goal conversion rate of shots taken from zone 13, 14 and 15. My bet is that it is even at EPL league level not even 5% chance of scoring from that distance.
It’s to do with playing style, evolution of football , possession based teams, low blocks allowing less shots on goal.
For such a dynamic sport, these new advanced metrics are still too raw to decently describe particular scenarios. While I welcome them, there is still a whole lot of polishing that needs to happen before they begin to matter (Ex: variables comes to mind .. too many to name here and when you start seeing how they impact each other you instantly reach infinity).
I recon we need a few more years before xG can be a reliable metric ... and I hope we dont lose the long shot because of it!
For the quality of the chance the quality of the shooter is irrelevant, so keep it out of XG. But if you want to know how well a team performed, instead of just adding up both teams XGs, the attempts should be multiplied by XG modifiers of individual shooters. How to best calculate that value is a bit tricky since some players take very few shots and player abilities also change, but considering that for Haaland that value is around 1.5, you are certainly missing something if you ignore it. It simply doesn't make sense to say that "City were lucky again".
This already exists,, kind of. Search up PSxG.
Pretty sure you could ask anybody in top tier football clubs and they would tell you if xG is causing a difference in strategy
Somebody calls Steven Gerrard he is one of the kind who can kill xG with his rocket shot
Surely a player's xg is a bit like wins above replacement in baseball. It's what the average player would do in that situation - so Haaland essentially has 6 (??) more goals than a league average striker would.
The English Premiership may be more 'skillful' than other UK leagues, but repeat this video for the EFL Championship and holes are blown in the stats as Midfielders are taking as many and often more shots than the Centre Forwards and plenty are from the edge of, or outside the penalty area. Middlesborough (26%) and Bristol City (29%) are the lowest number of shots from outside the box, 6 teams have 40% or more shots outside the box with Cardiff (47%) leading. I'm not a purist (nor do I like chess) so I prefer defenders and midfielders going in for tackles rather than standing off, and I like that I still get to spray mouthfuls of coffee around when balls burst the net from 30 yards 😀
Whoah. Ive been wondering where have all the long shot goals gone or if it is just me
It's not really xg . It's more the way the game is played . Teams always press out now instead of sitting back and its made scoring longshots way harder. That's why outside the box used to be such a great area . You would get time to make a shot , now players will actively press you before the ball even arrives to feet.
5:40 The math is misleading.
The probability, to score AT LEAST 2 goals from those 5 long shots is 26.27%.
This is the probability that X≥2 when X follows a binomial distribution with p=0.2 and n=5.
The probability to score 2 goals exactly would be 20.48%, to score 3 or more 5.79%.
Source: Me, a math student
xG stats in their present state are one of the most misused metrics. If you look at the definition at 1:57 that is what the stat represents. Once a real goal is scored, as in the game-state changes, xG accumulation is effected and not accounted for in the stat. Ex: Teams accumulate xG differently when they are up or down a goal. The worst offense to the stat is when xG is tallied up across multiple matches. It’s so far removed from what xG stats do represent that anyone with a sense for mathematics recognize that you’re not looking at anything of value.
You’re right in a sense, that teams behind will typically produce more xG than otherwise, but over large data sets it’s been proven that xG is a much more reliable predictor of future success than expert opinion or any other metric even if you don’t take the scoreline into account.
I do think the sort of thing you’re talking about, will be something studied a lot in the coming years with teams seeing how much xG difference they can sacrifice in order to increase their chances of winning by playing more attacking/defensive than is optimal
xG is definitely a really influential stat, and correct me if I’m wrong, but it doesn’t take into account the skillset of individual players right? So it’s misleading to give a blanket xG for certain shots when really the outcome can be expected to vary wildly depending on the player
Fear not, guys. My FIFA team is still on my knowledge of hockey. Shots from deep on the net hoping it goes in. I've scored some fucking beauties from out deep with Kimmich.
I finally know what's going on in FIFA 23 now
most likely XG is calculated per player now than per average in football by the teams. My guess is that Man city calculates higher xg for halland than average xg for striker. Also some players could have higher xg outside the box than others.
And soon Nikes The Last Game commercial will become reality 😆
Next video. How did the death of Long Shot give room to more efficient possesions
The pure stats give you a lot of information, what's missing is the phyological side of all professional sports, being physically fit, technique and tactics take you so for but just as important is the metal motivation and team spirit, a lot of player's play for the manager they like and respect,
Don't think a robot is going to get that any time soon 😜
Forgot to say how much the crowd can influence games, that's why most have better home records than away, last year in the Premier league Arsenal were going for CL top 4 place against my team Newcastle, the fans put on a Fantastic display with Wor Flags before the game the crowd were well up and very intimidating Arsenal had a much better team on paper, we looked like going down at Christmas, and yet we hammered them 2-0 never underestimate the phyological effects great support can have on teams.
Maybe playing futsal (small football) under box!?
Returning back to watch some football in this world cup and initially confused by the xG and other stats (stopped watching in 2010)
I do think these advanced stats are useful, but there are some things that you cannot measure by stats. This is a sport played by humans, not robots. Not everything has to be measured by stats.
I do agree that sooner or later, football teams would probably won't need a coach and instead will employ AI instead. Which means the beauty and unpredictability of the game will be dead. Like, any coach with a healthy mind will always allow Roberto Carlos to take a long shot from the left. That might not happen nowadays.
Knew there had to be an explanation for Liverpool s lack of shots from on or outside the box. Doesn't explain their bad shooting inside the box.
When did Kulusevski start his TH-cam channel?
the graph at 4:21 it has values 0.00 and up and it's talking about number of shots decreasing compared to amount of xg increasing so can I assume the value like 0.20 in terms of xg is 20% chance for a goal compared to shots which logically will be 20 shots because I still can't see those two values being in a single graph. a team can't have 0.20 shots that's less than 1 which conveys less than a shot it's imposable to have less than a shot. Can somebody please clarify ?
If a player, like Haaland, over a long period constantly overperform compared to his xG, is he overperforming or just performing?
While I can see the value of the xG statistic, it relies on many factors that are difficult to quantify. Furthermore, like you, I like seeing shots from distance in appropriate circumstances. There are times when a pass is the best option, but a team must shoot to score.
They must’ve invented xG after Suarez has left Liverpool.
What is the formula for xG?
Think there's definitely an argument that could say that all of what you say is more linked to the emergence of tiki taka playstyle and the fact that 80% of teams tried to copy fcb of guardiola
Good Longshooters are not obsolete. You show them that you can take scary longshots to pull apart defenses, next time you are in that spot. Goal or no it puts a bad idea in the head of more instinctual players. You want them to close you down and open up gaps behind them.
I was on soccerment the other day trying to get ppda data for a data viz I was doing but couldn't download it so I had to manually enter it into my csv(fortunately it was only la liga data)
Any tips ?
Hmm i never had an issue with downloads, I'll let them know😉
@@FootballMeta this embarrassing 😭
I don't mean I had trouble , I mean I couldn't figure out how to do it
Up to now i tell my friends they have changed the weight of the ball because it looks like its hard to deep the ball from a long short like back in the days of the Lampards this era i give you Pogba some of his shorts are boeing 707 , just my myths😀
Robert* not Robehr
Please can you do a 451 video bro if you want
Xg hasnt killed the long shot at all, it may have hindered it yes, but xgot (expected goals in target) has brought it back, i’ll give the example of khvicha kvaratskhelia, pretty much every long shot he takes has over 0.5xgot as long as it is on target.
Putting Messi and Harry Maguire in the same sentence so casually is disrespect to Maguire.
I've always factored in xG. Only difference is, it's MY xG. Not some computer.
No one here that considers the football aerodynamics to be a reason why long shots aren't a thing anymore? Ball has changed ever since that Jabulani disaster
Half* not Holf
There may be certain situations where a long shot becomes a game theoretic positive tactic. But you dont take them willy nilly
"People who don't understand football they analyse the stats"---Jose Mourinho
Yet Mourinho has spent all season complaining that his xGA is the best in the league lol
@@FootballMeta okay no problem I respect your opinion or reply I am interested to see when did he said this or show me a particular video Mourinho complaining that his XG is best
Example* not Exomple