yeah, it's hard. It sucks to be reminded, but I don't want to avoid bringing him up either when making videos about the scene or talking about the history. He was and still is at the heart of everything
Yeah, I know. I’m not trying to imply that we shouldn’t mention him anymore. It’s important to remember and celebrate him and his accomplishments. It’s important to grieve, and confronting the pain of loss is a part of that
Yeah, this really caught me by surprise too... Definitely still missing him, and as bitter-sweet as it is, it's nice to be reminded of just how great he was, both in terms of skill and kindness.
This might be the biggest magnum opus tool in the history of the community. I mean, my god, this video went in every direction I could've possibly wanted. This is a major game changer, not just for broadcasts, but even potentially to the understanding and approach to the game. We're entering the Moneyball sabermetrics era of NES Tetris and you love to see it. Breathtaking stuff here, Scout. Shoutouts to every genius hero who contributed and will contribute to this project.
As a Data Analysis student and NES Tetris fan I found this incredibly interesting and entertaining. The problem, progress and solution that you recounted in this video is what I hope to do with what I'm learning in school! I will definitely take a look at the code and documents you linked when I have time! GGs
Absolutely AMAZING. I have no words...its crazy how a driven musician can dive into the world of coding, programing, data entry, and rendering to create something so sophisticated and technical for the entire community to benefit from. Players, viewers, commentators alike...everyone can now thrive in a new way with this creation. Your dedication and passion to this game and community is unparalleled. Thank you, Scout! Now lets get that maxout for the cherry on top!
Man it's crazy how you went from an unknown creator dipping your toes into competitive classic Tetris to becoming one of the biggest content creators in the scene, very glad to have you here man
I have also thought of a Statistic that has never been seen, but would be very useful to know; the True Drought. Part 1: The problem with the regular drought statistic. The problem with the regular drought statistic is that it will count towards a drought if you don't get an I piece, even if you don't have a 4-tall open well, where getting an I piece won't result in a Tetris. Part 2: The True Drought. The True Drought aims to resolve this issue. The True Drought only counts up when you have an open 4-high well with all the lines fully completed, and don't get an I piece. AKA it only counts up if getting an I piece would result in a Tetris, and you don't get one. This also means that doing a delayed burn would temporarily stop the counter until the well is cleared. Part 3: The problem with the True Drought. This Statistic isn't perfect. Due to the nature of randomness you would expect to see an average of a 7 piece true drought. However, you can expect a skew in the average due to minor adjustments that players can make after processing the next piece preview (examples: adjusting your piece placement to create a high enough well because you see the I piece is coming next, adjusting to a delayed burn that you know will get resolved with the next piece meaning that the counter won't count up for one piece) I hope GameScout sees this!
That’s a cool statistic! Yobi does have something somewhat similar to that in NEStrischamps, where he tracks streaks of Tetris-readiness...so if you don’t pass up on a Tetris, it is basically a “true drought”. You should suggest it on his server!
@@aGameScout I was thinking that it might be interesting to see a "tetris conversion percentage", which count how many I pieces that are preceded by enough non-I pieces to be likely able to score a Tetris, and report what fraction of those did in fact result in Tetrises. What I'd most like to see, though, would be for Nintendo and the Tetris Company to cooperate with the release of an official licensed cartridge with RNG seeding control and perhaps a facility for showing encoded information about recent plays on the side of the screen (so that if a stream has a momentary drop-out, play could be accurately recorded in cheat-resistant fashion), so as to minimize the effect of randomness.
Another useful statistic idea: the Adjusted Tetris Rate. Maybe someone has their lines that aren't in a Tetris be in triples, which could adjust the score. Answer: give them a multiplier.
@@btf_flotsam478 A statistic like that already exists. It's called efficiency and it's defined as the points per line ignoring level increases, or in other words assuming level 0 scoring. You can see it in nestrischamps at 15:30. The only difference between EFF and your proposed ATRT is that EFF goes between 40 and 300 = 1200/4 instead of maxing out at 100% but that shouldn't make it any harder to understand. EDIT: It is relatively easy to show that EFF = 40SRT + 50DRT + 100TrRT + 300TRT where S(ingle), D(ouble), Tr(iple), T(etris) are the line clears. (You can test this formula with the numbers at 15:30 since nestrischamps provides rates for all line clears) So, EFF is already an adjusted tetris rate in a sense and that's why I would argue that ATRT should be defined as EFF/300. -The only problem I have now is the name. I would prefer something like Efficiency Rate instead because Adjusted Tetris Rate implies the existence of for example an Adjusted Triple Rate.- EDIT 2: Hopefully I won't have to edit this again. I think defining an Adjusted Triple Rate as the triple rate adjusted only by the less valuable line clears (ignoring tetrises) is a satisfactory definition so I'll retract my criticism. (ATrRT = 0,4SRT + 0,5DRT +TrRT) This also works for doubles (and singles technically but that's just the trivial ASRT=SRT). I personally find the name Efficiency Rate better than Adjusted Tetris Rate though.
HECK YECK this is so cool? Ugh I love hearing about Tetris stats. Thanks for walking through your thought process and everything. This is utterly fascinating
BOOM! New stat from GameScout ! The chronological reveal of the process was so well explained along with the usual high clarity production values- you make it a joy to watch. And with the revisited gameplay examples used (plus inclusion of so many beloved players. The 'Late' was a big lump in the throat too...) By half way I was waiting to hear "so i made this and then I was getting Trey to help/add this... " But I never considered the scope and knowledge of the community to handle all the programming and more. Its awesome, and I DO hope it goes through iterations enough to end up in the default comp gameplay view. well done !!
Knowing pace when playing a competitive game is something most players estimate on the fly but this tool is so useful to know exactly how perfect you need to play and how much you need to pray for rng lol
Man, you're a legend. That's another amazing piece of analysis and editing. You're making History my friend, your work will be celebrated by future generations of retro-gaming enthusiasts.
Great video! I know almost nothing about professional Tetris, but as a programmer, I just found all the work that went into making this to be really impressive. It seems like it can really add a lot of depth to professional Tetris matches.
Awesome that you quantified and measured pace. I knew it was important from my start watching competitive Tetris and the only solution to that point seemed to be experience. Fantastic work! I’m sure the late great Jonas would’ve loved this.
I discovered Yobi about 2 years ago and already back then he impressed me with his layouts. It's awesome to see his work now slowly getting more publicly known and appreciated.
So this is the reason why CTWC 2019 Recap has been delayed for several months, right? Anyway, thank you everyone for making these tools for better and more beautiful Tetris match analyses.
Partially. In December I was burnt out and needed a bit of a break, and was on track to finish part 2 in January but then the Jonas news hit. That took a while to process and in the meantime I've gotten back into doing the monthly recaps because any video talking about NES Tetris history inevitably involves Jonas and it's hard. I still fully intend to finish it, but the tone of some of the parts will inevitably be completely different.
@@aGameScout Oh, this is ONE of the reasons and might have least(?) influence on your break. I agree, Jonas' sudden news, the main reason, made us stop what we did previously for a while and make videos in memory of him. Sorry if my comment made you upset, and I hope you do your best on your tasks without hastiness and stress. Thank you always.
Overall, I think this is will be a fantastic addition to tournaments. I have a slight concern on how cluttered the screen may become during a broadcast. I think that any person invested in Tetris would love to know the stats, but you also don't want to overshadow the gameplay happening with a bombardment of ever-changing numbers. I do like the set-up around 19:00 in the GreenTea/Joseph match, but I think the most important stat there is the overall high-score pace. The other numbers are further representations of that overall high-score pace, and should be treated as such. Regardless of that, I think that any additional stats or information that the commentators can utilize is going to be great. And I think you and everyone who has helped has done some amazing work. Awesome!
thanks! Yeah that's true. Yobi created 4 different layouts with different levels of stats for whatever people may want during restreaming matches. Having ALL the stats is more fun when doing post game analysis
I’m no mathmagician so parts felt overwhelming, but I’m so glad you saw it all the way through. I can’t wait to see some of this implemented going forward, especially in real-time.
This is really illustrative of why diversity of viewpoints can be so effective in progression of understanding something. The math (and some of the programming) are quite simple, if a little buggy, for someone who has that background, but it takes someone also recognizing that the gap in commentary exists to motivate someone to try to fix the issue. When we have no collaboration, you just need to hope whoever comes up with this issue has the capabilities of building the program. With collaboration, we can work together and flesh out some amazing things. We see this all the time in science too! Having diversity of backgrounds and skill sets can be monumental in tackling an issue.
Love it! One giant piece of hole in this real time analysis holy grail is board state. The next leap will have compare the condition of each player's matrix, and give heavy weight to boards in a tetris-ready state vs a board with lots of swiss cheese. Previous games will have to be analyzed for historic predictions. In baseball, there is a predictive chart based on each game situation on what the outcome will be based on the hundreds of thousands of box scores collected over the past 150 years of games played. How far down the rabbit hole can we go?
Yes, I’d love to see something like fangraphs for Tetris! Yobi is tinkering with a way to save the entire board state of every game played through NEStrischamps, and once a library of data is there, then the possibilities are wide open for creative ways to use it.
@@CheezTetris yeah, there’s a ton more possibilities if players eventually have their historical game stats tracked. I agree that killscreen scores would be a good thing to add if such data were available
Holy cow! That's amazing, but what's even more amazing is that someone would do those changes in numbers themself. My hat's off to you, friend, and every other Tetris commentator that edits the numbers in post. I honestly had no idea that was the case, and always just thought it was a program running that did all that for you and/or the CTWC.
Damn! Nice work on the number crunching and Yobi on the programming. I have programmed with OCR libraries and instantly recognized some of the hurdles you mentioned with detecting the characters and so on.
This was an absolutely amazing video! I never knew how much work went into this whole project. I'm glad I could help in a small way, and I can't wait to see more people start to use it!
Respect. It's great to see the lengths that this community will go to to make tools to improve the game for not only the players, but everyone watching. It's much more than just a statistic, it really highlights how collaborative and inclusive the community is. Well done to all involved, and looking forward to seeing it used in future events!
I love the variety of different skill sets people have applied to this game! Learning new piece stats, learning new ways to control the pieces, and now adding all these improved ways to track and watch games. I've loved watching this sport grow. Huge props on your contributions to the game - this is a huge tool for fans and players!
wow wow wow, I saw you getting help on the pace stat over the last few months so I knew what this was about but it came out SO GOOD. Amazing community support, obviously big shout out to Yobi in particular!! Question -- have you mentioned this to Trey at all?? I feel like adding this to MOC/Treyvision would just be... the best, and would really make CTWC even better.
Yes, I will mention it to Trey! I wasn't expecting to have any discussions about adding it to maxoutclub/ctwc until the stat had been in use for a while and proven itself, since there could still be more ways to improve it and MOC has a sleek minimalist layout that I think would only add something if it noticeably had a huge benefit. Yobi working on his own restreaming program on the side when I reached out to him ended up being a happy coincidence, since his goal was to include as many stats as possible...so hopefully if some people start using it for restreaming games, the pace stat/any other stats he's included that end up being useful will be considered for MOC!
Ah makes sense!! There's definitely plenty of time and no rush, so it makes a lot of sense to wait and let it get some use and see if it can be refined. Great plan!!
Great video! I think the biggest part of the Classic Tetris community doesn't know yet how amazing NESTrisChamps is and what it can do to improve this e-sport. Good job, as always!
Man, the research you did for all the examples is incredible. And it makes the video that much more interesting! Honestly, you did an amazing job, scout!
Great video as usual! It's the one time I really regret not being more involved with the community. I worked on similar projects and came up with some interesting ideas but never put in too much effort in collaborating.
incredible dude, had me saying "wow" a bunch of times - at first glance the stat I would personally expect to see and wouldn't find numerically overwhelming would be the "max possible score before 29" that you suggest happy to find your channel recently - really high production value for community commentary for the tetris world - bet you weren't expecting to have to learn so much new computer stuff to get this working, eh? xD
My favourite part was your description of the problems and bugs near the end. You struggling with python and learning these steps sounds banal but it is so relatable. Then when you got into the bugs I was thinking 'ok yah, that's fixable, hmhm yup gotta do that' and then you proceeded to point out the exact solutions I had in mind xD Really, when you started out with spreadsheets in the video I was screaming to automate that stuff. To see you reach that point with the help of others... chefs kiss. Great watch.
Awesome video, as always. Your positive energy actually makes my day every time. Also I appear briefly in the video a few times, so it's perfect in every way
I can only thank you for the amount of time you've put into this new stats. Hopefully they'll be using it in the next CTWC. It would be a huge improvement for the spectators, the commentators and the players as well.
the strategy is going to be more important now in competitive tetris. Great effort and dedication for this video and the program to exist so thank you Scout and everyone who contributed
First of all, this is a gread video and I really enjoyed watchin it! I hope that you come up with something similar to this pace stat that actually gets implemented into the official layout. But now to the problem that I have with the pace stat in its current version: The pace stat seems really useful if a player already topped out, but in most other cases it seems either useless or missleading. If the players are at the same linecounts, then the pace stat is more or less useless since then the score differential just reveals all necessary information. If one player is far ahead in lines though, the problem that you aleady mentioned comes into play: The slower player will almost always lead in pace, since the stat assumes quite a lot of perfect play, which will almost certainly not happen. This means that one could easily see that the slower paced player has the chance to catch up, but it actually lets you believe that the slower player is ahead in actual pace, which might be not true in many cases. You already mentioned this problem in the video already, but you displayed it as a tiny problem, while it actually makes this stat misleading in most situations. If a player is far behind in lines, the other one is close to level 29 and the pace differential displays positive 30k for the slower player, it is more or less guaranteed that the slower player will actually lose the game, which is one example of the way this stat gives misleading results. I think that projecting the players performance to the future based on their current tetris rate (or maybe trble and double rates as well) is a way more useful stat. This might not give you the actual chance of a player winning the game, since there are these problems in calculating the stat if you just burned a lot, as you mentioned. But: The pace stat in its current form does not display the chance that the current player wins either, so there is no advantage for the pace stat in this regard. The fun of the game is, that you can not predict how a player is going to play, which means that most of these stats will not be useful. From a machine learning perspective (i actually study this), the best chance we have to predict the outcome of the game is to extrapolate from the previous performance of the player, which is the current tetris rate up to this point. This might give misleading results in some cases, as discussed before, but it will give good predictions in way more cases than the pace stat in its current version. One could advance that stat in many ways: One way would be to include the players usual tetris rate over many games. This would reduce the influence of too bad or too good play in the current game. But this would be a lot of work to track in a tournament and it would be tedious to add all of the data. Another way to improve the stat would be to include the other players performance in the calculations. If the other player scored a lot in the lines that differentiate him compared to the slower player, then this probably means that a lot of favourable pieces will come and thus the slower player will most likely score better. But if you were to take this into account, you would need to account for the fact that the other player could just be the better player, which would not be easy as well. But overall, the easiest way would be the method you described at first: Just save information about the scores that each player had at different line counts, and compare the score of the slower player to the score that the faster player had at this point. One of the reasons you dismissed this idea was that it is extemely tedious to enter these information into a spreadsheet. But after you began automating this process by programming, this is not a problem any more. You do not even have to restrict yourself to the 10-line threshold, but instead you can track the score continuously throughout the game and calculate the differential at any given point. Dealing with fractions of tetrisses should not be a problem either when using programming. The other problem you had with this stat is that it is misleading if the other player burned a lot recently. This is true, but all stats discussed here had certian problems in some cases and this seems to be the least misleading one, or at least less misleading compared to the current pace stat, since this will always give misleading results if one player is far behind, and this stat would only give misleading results sometimes. Some interesting stuff that might be useful to consider: In machine learning we have a concept called "unbiased estimator". This basically means, that if we calculate a value based on emprirical data (so the value would be out pace difference and our empirical data would be the current performance of both players), then the estimator of the value is unbiased if in its expectation (so if we average it over infinitely many games) it would on average give the correct results. The stat that extrapolates the players current tetris rate would be fairly close to being unbiased, sicne it sometimes overestimates the slower players performance and sometimes underestimates it, such that we have a good result on average (there are some other factors that falsify this statement, like the fact that it is harder to score tetrisses at level 19 compared to 18, but that would be too complicated to account for). This means, that this stat has a chance to provide useful information. The current pace stat on the other hand consistently overestimates the slower players performance, such that is is extremely far from being undbiased. such that the results are not that helpful in most cases. Okay this is the result of me being bored at night ^^ I dont know if anyone is actually going to read this, but it nevertheless feels good to have commented my thoughts I guess :D
This is incredible. There have been so many times that I've yelled at my phone because announcers are reporting on just score or even just tetris rate, when there was clearly more going on under the surface. It just gives the wrong impression to not only the audience, but the players as well -- there's nothing worse than hearing "a win is virtually impossible at this point" and for that to not be true. It's so much better to be able to hear "a win is here is gonna be tough, but if he plays extraordinarily well, it's still possible." No more blind guessing on the part of the announcers.
I love the part where he complains about how terrible Python distribution is. It's not his fault-it's definitely a meme at this point that Python is this way.
Absolutely loved the explanation of everything. A true bliss to watch. Also, I'm so glad to hear that Jonas had the best Tetris Rate in 2018, he totally deserve it. 😢🙂
I am absolutely a much bigger faan of modern guideline vs tetris than NEStris, but the lengths the nestris community has gone to in order to advance the game they love so dearly is just breathtaking. Thank you so much for providing this information for those who don't follow nestris
With these new programming weapons now in your arsenal, I can't wait for the next CTWCs!! 😀 Excellent work from you and the Classic Tetris community! 👏👏👏
The way your Pace stat assumes perfect play parallel how in Ten-Pin Bowling coverage, Strikes and Spares are assumed to be worth their max value (so a Strike in the first frame is shown as being an effective score of 30, even though it isn't officially 30 until the next two bowls happen and both are also Strikes). A thing to maybe consider (although explaining this is the hard part) is to have a comparison to an assumed Tetris rate that's based more on a general average. My rule thumb used to be "80% Tetris rate at level 18 speeds, 60% at level 19 speeds", but so many players are doing better than that these days that 90% and 70% might be more reasonable? Something to look at, perhaps. Speaking of getting the Tetris lead wrong, some of it comes from players simplifying to "30k for a Tetris in the late game", and some of it comes from how a Tetris that increases the level is scored at the new level (this scoring quirk was unnoticed for quite some time). Thank you for respecting the fact that Tetris is a complicated game, and for trying to make it more approachable.
Have you come across the DLS metric from Cricket? The Duckworth-Lewis-Stern method is used to compare performance of cricket teams during a match, as they play sequentially, and interruptions would usually mean the game is abandoned. This allows a winner to be chosen, and is somewhat similar in idea to your pace stat. It works by correcting score according to the difference in 'resources' available to the teams, specifically because in cricket there are two types of resources (balls and players not out) you can't easily scale a target score. I wonder if you could do something similar, as with tetris rate you're compensating for the value of later levels only, and with the new stat you're compensating for value and play improvements, but perhaps there's a scaling factor for lines remaining (like the DLS tables) that more accurately represents real play. You could find this by analysing past games to see if there is a relationship between lines remaining and average burn at that level across many players, using real behaviour rather than programmatically finding the optimal behaviour.
The pace stat is great in terms of analysis, but I do share the sentiment with how it predicts taking some of the fun out of it. Because players can only ever perform worse than the pace, not better. I'm more interested in what the difference would be in their scores if they both were at the same number of lines (with some deviation for burns). If player A is at line 29, and player B is at 34, I want to know if based on recent performance (i.e. last N lines tetris percentage), what does the scores look like if both were at 34. That allows the player to play better than the pace prediction, and it's not about predicting what the match is gonna end up being, it's about showing you accurately what the current state is. How could you use your current algorithm to do that? Instead of estimating the score at 230 lines you could estimate it at the leaders current lines, or even leaders current lines plus some prediction amount like next 20 lines. And instead of assuming perfect playing from both players maybe assume a reasonable upper limit. I don't know, emotionally it definitely is a little daunting that the pace always assumes robotically perfect playing and we'd only see players basically not living up to that standard in chase downs. Secondly I like the aspect of trying to think about "we'll if they do this and this, maybe the score will just tip over". If you were doing this for like odds analysis for betting, I'd call it a smart tool in addition to the usual suspects, but for viewing experience it doesn't make it more entertaining for me, because as stated, I already know who should be winning, af if they don't it's because compared to the prediction they "dropped the ball", even though we don't expect that level of play from them, that's what the algorithm suggests.
@@CobaltTetris funnily enough, a year later with roll chase downs now being a reality, I think the pace statistic is fortunately still far from being obsolete. Most players in the scene aren’t elite rollers and even the elite rollers die quite often without scoring much on 29. Pace is not a guarantee of who wins, but it is definitely a massive advantage to be ahead 100-200k into 29 no matter who you’re facing!
The next major leap forward will be to alter the max score possible based on board position. For example if there are many holes or a misdrop instead of a right side well the max score should shift to reflect that. That means a new metric and analysis on board position based on factors like number of holes, stack shape, height, level, and well-size, and accessibility.
I think this is great, but I had a question on one of your comments. You said a problem was that it the one method wouldn't account for the fact that dog has been burning since 210 lines. But doesn't the final method you came up with also not account for that? What I mean is, let's say dog was at 210 lines 1 tetris lead, but a messy play field, vs andy 210 lines clean play field. Wouldnt it still say that dog has the pace lead because it assumes he will tetris all the lines remaining regardless of playfield?
yeah, currently detecting a dirty playing field and accommodating for that is too complicated and abstract to work in at the moment. In the situation described, andy would be in the better spot, however in terms of raw scoring potential, depending on how bad dog's board was, he could potentially go for dirty tetrises if he really had to, the raw score does still have some application
@@aGameScout ahh ok thanks for the reply. I was just wondering if you were actually able to account for it in the final method. You're right that seems way too hard to calculate. I can't even think of what the logic would be for that. Mayne stack height vs deepest alley available. But yea sooo complicated
Let's put it this way... Jonas was as good as he was before community involvement. He, and a handful of players, invented the wheel. Watching his gameplay is still the most soothing, logical, approach to the game. He does something that few others match. I present: Master Zentris Neubauer!
16:25 The command line isn't hard when you get used to it, although Windows' command line can be a bit painful at times since it has been left to rot in favor of graphical apps.
Hearing “the late Jonas Neubauer” broke my heart but I know he’d love this video and everything you have planned for the future ❤️
I don't know why, but I didn't entirely realize when he said that..
Hearing “the late Jonas Neubauer” is aways difficult
yeah, it's hard. It sucks to be reminded, but I don't want to avoid bringing him up either when making videos about the scene or talking about the history. He was and still is at the heart of everything
Yeah, I know. I’m not trying to imply that we shouldn’t mention him anymore. It’s important to remember and celebrate him and his accomplishments. It’s important to grieve, and confronting the pain of loss is a part of that
@@GuilhermeBF Jonas' memory: "I'm yours forever! We're all together in this life, don't you forget it. We're all connected!"
Yeah, this really caught me by surprise too... Definitely still missing him, and as bitter-sweet as it is, it's nice to be reminded of just how great he was, both in terms of skill and kindness.
@@GuilhermeBF are you portuguese or brazilian?
This might be the biggest magnum opus tool in the history of the community. I mean, my god, this video went in every direction I could've possibly wanted. This is a major game changer, not just for broadcasts, but even potentially to the understanding and approach to the game. We're entering the Moneyball sabermetrics era of NES Tetris and you love to see it. Breathtaking stuff here, Scout. Shoutouts to every genius hero who contributed and will contribute to this project.
NES Tetris analytics FTW!
"Sabertetrics"??? 😁
@@adrianhendricks772 LOL
Let’s go!! This is super cool!
oh hello joseph
oh hello joseph
oh hello joseph
Hello Joseph
oh hello joseph
this was originally meant to be a 5 minute video...welp. TL;DR maffs
I'm glad it ended up being 4 times longer than it was originally going to be lol
slow maffs
Yeah it got deep.
You did such a good job on this.
But the quality of this video is superb! You did a great job!
As a Data Analysis student and NES Tetris fan I found this incredibly interesting and entertaining. The problem, progress and solution that you recounted in this video is what I hope to do with what I'm learning in school! I will definitely take a look at the code and documents you linked when I have time! GGs
Wow thanks for the shoutout, didn't realize my dumb idea would cause all this hard work for Yobi and a bunch of others lmao
lol
Clearly wasn't too dumb lol
Absolutely AMAZING. I have no words...its crazy how a driven musician can dive into the world of coding, programing, data entry, and rendering to create something so sophisticated and technical for the entire community to benefit from. Players, viewers, commentators alike...everyone can now thrive in a new way with this creation. Your dedication and passion to this game and community is unparalleled. Thank you, Scout! Now lets get that maxout for the cherry on top!
i swear, no youtube channel gets me as hyped to see a new video in my sub box as this one
Man it's crazy how you went from an unknown creator dipping your toes into competitive classic Tetris to becoming one of the biggest content creators in the scene, very glad to have you here man
Sheesh yobi's tools is finally being put to use 🥶🥶🥶
heck yeah
I have also thought of a Statistic that has never been seen, but would be very useful to know; the True Drought.
Part 1: The problem with the regular drought statistic. The problem with the regular drought statistic is that it will count towards a drought if you don't get an I piece, even if you don't have a 4-tall open well, where getting an I piece won't result in a Tetris.
Part 2: The True Drought. The True Drought aims to resolve this issue. The True Drought only counts up when you have an open 4-high well with all the lines fully completed, and don't get an I piece. AKA it only counts up if getting an I piece would result in a Tetris, and you don't get one. This also means that doing a delayed burn would temporarily stop the counter until the well is cleared.
Part 3: The problem with the True Drought. This Statistic isn't perfect. Due to the nature of randomness you would expect to see an average of a 7 piece true drought. However, you can expect a skew in the average due to minor adjustments that players can make after processing the next piece preview (examples: adjusting your piece placement to create a high enough well because you see the I piece is coming next, adjusting to a delayed burn that you know will get resolved with the next piece meaning that the counter won't count up for one piece)
I hope GameScout sees this!
That’s a cool statistic! Yobi does have something somewhat similar to that in NEStrischamps, where he tracks streaks of Tetris-readiness...so if you don’t pass up on a Tetris, it is basically a “true drought”. You should suggest it on his server!
@@aGameScout I was thinking that it might be interesting to see a "tetris conversion percentage", which count how many I pieces that are preceded by enough non-I pieces to be likely able to score a Tetris, and report what fraction of those did in fact result in Tetrises. What I'd most like to see, though, would be for Nintendo and the Tetris Company to cooperate with the release of an official licensed cartridge with RNG seeding control and perhaps a facility for showing encoded information about recent plays on the side of the screen (so that if a stream has a momentary drop-out, play could be accurately recorded in cheat-resistant fashion), so as to minimize the effect of randomness.
Another useful statistic idea: the Adjusted Tetris Rate. Maybe someone has their lines that aren't in a Tetris be in triples, which could adjust the score. Answer: give them a multiplier.
@@btf_flotsam478 A statistic like that already exists. It's called efficiency and it's defined as the points per line ignoring level increases, or in other words assuming level 0 scoring. You can see it in nestrischamps at 15:30. The only difference between EFF and your proposed ATRT is that EFF goes between 40 and 300 = 1200/4 instead of maxing out at 100% but that shouldn't make it any harder to understand.
EDIT: It is relatively easy to show that EFF = 40SRT + 50DRT + 100TrRT + 300TRT where S(ingle), D(ouble), Tr(iple), T(etris) are the line clears. (You can test this formula with the numbers at 15:30 since nestrischamps provides rates for all line clears) So, EFF is already an adjusted tetris rate in a sense and that's why I would argue that ATRT should be defined as EFF/300.
-The only problem I have now is the name. I would prefer something like Efficiency Rate instead because Adjusted Tetris Rate implies the existence of for example an Adjusted Triple Rate.-
EDIT 2: Hopefully I won't have to edit this again. I think defining an Adjusted Triple Rate as the triple rate adjusted only by the less valuable line clears (ignoring tetrises) is a satisfactory definition so I'll retract my criticism. (ATrRT = 0,4SRT + 0,5DRT +TrRT) This also works for doubles (and singles technically but that's just the trivial ASRT=SRT). I personally find the name Efficiency Rate better than Adjusted Tetris Rate though.
when GameScout uploads, its a good day
now i realise just how much effort those past videos took
HECK YECK this is so cool? Ugh I love hearing about Tetris stats. Thanks for walking through your thought process and everything. This is utterly fascinating
BOOM! New stat from GameScout ! The chronological reveal of the process was so well explained along with the usual high clarity production values- you make it a joy to watch. And with the revisited gameplay examples used (plus inclusion of so many beloved players. The 'Late' was a big lump in the throat too...) By half way I was waiting to hear "so i made this and then I was getting Trey to help/add this... " But I never considered the scope and knowledge of the community to handle all the programming and more. Its awesome, and I DO hope it goes through iterations enough to end up in the default comp gameplay view. well done !!
This video just kept getting cooler and cooler.
"Still, ideally (something)"
"Ahh, of course"
"...but what if I told you..."
"NO WAY'
Wow mad respect for all the work you and others put into this. Super cool.
Knowing pace when playing a competitive game is something most players estimate on the fly but this tool is so useful to know exactly how perfect you need to play and how much you need to pray for rng lol
Man, you're a legend. That's another amazing piece of analysis and editing. You're making History my friend, your work will be celebrated by future generations of retro-gaming enthusiasts.
I never thought i would watch until the end, but i did and do not regret at all. What a cool video.This Channel is so underrated!!!!
It's cool to see the process of how an e-sport evolves on how it tracks statistics.
oh god im a nerd.
Great video! I know almost nothing about professional Tetris, but as a programmer, I just found all the work that went into making this to be really impressive. It seems like it can really add a lot of depth to professional Tetris matches.
Awesome that you quantified and measured pace. I knew it was important from my start watching competitive Tetris and the only solution to that point seemed to be experience. Fantastic work! I’m sure the late great Jonas would’ve loved this.
I discovered Yobi about 2 years ago and already back then he impressed me with his layouts. It's awesome to see his work now slowly getting more publicly known and appreciated.
So this is the reason why CTWC 2019 Recap has been delayed for several months, right? Anyway, thank you everyone for making these tools for better and more beautiful Tetris match analyses.
Partially. In December I was burnt out and needed a bit of a break, and was on track to finish part 2 in January but then the Jonas news hit. That took a while to process and in the meantime I've gotten back into doing the monthly recaps because any video talking about NES Tetris history inevitably involves Jonas and it's hard. I still fully intend to finish it, but the tone of some of the parts will inevitably be completely different.
@@aGameScout Oh, this is ONE of the reasons and might have least(?) influence on your break. I agree, Jonas' sudden news, the main reason, made us stop what we did previously for a while and make videos in memory of him. Sorry if my comment made you upset, and I hope you do your best on your tasks without hastiness and stress. Thank you always.
Overall, I think this is will be a fantastic addition to tournaments.
I have a slight concern on how cluttered the screen may become during a broadcast. I think that any person invested in Tetris would love to know the stats, but you also don't want to overshadow the gameplay happening with a bombardment of ever-changing numbers. I do like the set-up around 19:00 in the GreenTea/Joseph match, but I think the most important stat there is the overall high-score pace. The other numbers are further representations of that overall high-score pace, and should be treated as such.
Regardless of that, I think that any additional stats or information that the commentators can utilize is going to be great. And I think you and everyone who has helped has done some amazing work. Awesome!
thanks! Yeah that's true. Yobi created 4 different layouts with different levels of stats for whatever people may want during restreaming matches. Having ALL the stats is more fun when doing post game analysis
I’m no mathmagician so parts felt overwhelming, but I’m so glad you saw it all the way through.
I can’t wait to see some of this implemented going forward, especially in real-time.
This is really illustrative of why diversity of viewpoints can be so effective in progression of understanding something. The math (and some of the programming) are quite simple, if a little buggy, for someone who has that background, but it takes someone also recognizing that the gap in commentary exists to motivate someone to try to fix the issue. When we have no collaboration, you just need to hope whoever comes up with this issue has the capabilities of building the program. With collaboration, we can work together and flesh out some amazing things. We see this all the time in science too! Having diversity of backgrounds and skill sets can be monumental in tackling an issue.
Love it! One giant piece of hole in this real time analysis holy grail is board state. The next leap will have compare the condition of each player's matrix, and give heavy weight to boards in a tetris-ready state vs a board with lots of swiss cheese. Previous games will have to be analyzed for historic predictions. In baseball, there is a predictive chart based on each game situation on what the outcome will be based on the hundreds of thousands of box scores collected over the past 150 years of games played. How far down the rabbit hole can we go?
Pace stat + [layers' mean killscreen score from last 10 games lol
Yes, I’d love to see something like fangraphs for Tetris! Yobi is tinkering with a way to save the entire board state of every game played through NEStrischamps, and once a library of data is there, then the possibilities are wide open for creative ways to use it.
@@CheezTetris yeah, there’s a ton more possibilities if players eventually have their historical game stats tracked. I agree that killscreen scores would be a good thing to add if such data were available
I can see real time sportsbetting happening with this.
Holy cow! That's amazing, but what's even more amazing is that someone would do those changes in numbers themself. My hat's off to you, friend, and every other Tetris commentator that edits the numbers in post. I honestly had no idea that was the case, and always just thought it was a program running that did all that for you and/or the CTWC.
Damn! Nice work on the number crunching and Yobi on the programming. I have programmed with OCR libraries and instantly recognized some of the hurdles you mentioned with detecting the characters and so on.
This was an absolutely amazing video! I never knew how much work went into this whole project. I'm glad I could help in a small way, and I can't wait to see more people start to use it!
Respect. It's great to see the lengths that this community will go to to make tools to improve the game for not only the players, but everyone watching. It's much more than just a statistic, it really highlights how collaborative and inclusive the community is. Well done to all involved, and looking forward to seeing it used in future events!
I love the variety of different skill sets people have applied to this game! Learning new piece stats, learning new ways to control the pieces, and now adding all these improved ways to track and watch games. I've loved watching this sport grow. Huge props on your contributions to the game - this is a huge tool for fans and players!
wow wow wow, I saw you getting help on the pace stat over the last few months so I knew what this was about but it came out SO GOOD. Amazing community support, obviously big shout out to Yobi in particular!!
Question -- have you mentioned this to Trey at all?? I feel like adding this to MOC/Treyvision would just be... the best, and would really make CTWC even better.
Yes, I will mention it to Trey! I wasn't expecting to have any discussions about adding it to maxoutclub/ctwc until the stat had been in use for a while and proven itself, since there could still be more ways to improve it and MOC has a sleek minimalist layout that I think would only add something if it noticeably had a huge benefit. Yobi working on his own restreaming program on the side when I reached out to him ended up being a happy coincidence, since his goal was to include as many stats as possible...so hopefully if some people start using it for restreaming games, the pace stat/any other stats he's included that end up being useful will be considered for MOC!
Ah makes sense!! There's definitely plenty of time and no rush, so it makes a lot of sense to wait and let it get some use and see if it can be refined. Great plan!!
hey david, great video! and what I find even cooler is that it encourages people to contribute to open source projects like nestrischamps. thank you!
Great video! I think the biggest part of the Classic Tetris community doesn't know yet how amazing NESTrisChamps is and what it can do to improve this e-sport. Good job, as always!
Man, the research you did for all the examples is incredible. And it makes the video that much more interesting! Honestly, you did an amazing job, scout!
20:43 Haha, I'm famous!! Seeing this totally made my day! 😀 Thanks for the shoutout, @aGameScout !! 🙏🙏🙏
Great video as usual! It's the one time I really regret not being more involved with the community. I worked on similar projects and came up with some interesting ideas but never put in too much effort in collaborating.
incredible dude, had me saying "wow" a bunch of times - at first glance the stat I would personally expect to see and wouldn't find numerically overwhelming would be the "max possible score before 29" that you suggest
happy to find your channel recently - really high production value for community commentary for the tetris world - bet you weren't expecting to have to learn so much new computer stuff to get this working, eh? xD
My favourite part was your description of the problems and bugs near the end.
You struggling with python and learning these steps sounds banal but it is so relatable. Then when you got into the bugs I was thinking 'ok yah, that's fixable, hmhm yup gotta do that' and then you proceeded to point out the exact solutions I had in mind xD
Really, when you started out with spreadsheets in the video I was screaming to automate that stuff. To see you reach that point with the help of others... chefs kiss. Great watch.
The classic Tetris community really is amazing. Awesome video
Awesome video, as always. Your positive energy actually makes my day every time. Also I appear briefly in the video a few times, so it's perfect in every way
I love stuff like this so much! Very, very good video. Keep it up, David.
I can only thank you for the amount of time you've put into this new stats. Hopefully they'll be using it in the next CTWC. It would be a huge improvement for the spectators, the commentators and the players as well.
"Isn't this all a bit nerdy?"
- AGamesScout -
My response to that as a gamer is a proud yes.
This is what passion looks like. Great job!
I wish I had something I was this interested in again. Due to an accident I can't do what I used to anymore.
What an amazing work! I love the love the Tetris community puts in its game, I don't even play Tetris but this is truely a work of art!
the strategy is going to be more important now in competitive tetris. Great effort and dedication for this video and the program to exist so thank you Scout and everyone who contributed
First of all, this is a gread video and I really enjoyed watchin it! I hope that you come up with something similar to this pace stat that actually gets implemented into the official layout. But now to the problem that I have with the pace stat in its current version: The pace stat seems really useful if a player already topped out, but in most other cases it seems either useless or missleading. If the players are at the same linecounts, then the pace stat is more or less useless since then the score differential just reveals all necessary information. If one player is far ahead in lines though, the problem that you aleady mentioned comes into play: The slower player will almost always lead in pace, since the stat assumes quite a lot of perfect play, which will almost certainly not happen. This means that one could easily see that the slower paced player has the chance to catch up, but it actually lets you believe that the slower player is ahead in actual pace, which might be not true in many cases. You already mentioned this problem in the video already, but you displayed it as a tiny problem, while it actually makes this stat misleading in most situations. If a player is far behind in lines, the other one is close to level 29 and the pace differential displays positive 30k for the slower player, it is more or less guaranteed that the slower player will actually lose the game, which is one example of the way this stat gives misleading results.
I think that projecting the players performance to the future based on their current tetris rate (or maybe trble and double rates as well) is a way more useful stat. This might not give you the actual chance of a player winning the game, since there are these problems in calculating the stat if you just burned a lot, as you mentioned. But: The pace stat in its current form does not display the chance that the current player wins either, so there is no advantage for the pace stat in this regard. The fun of the game is, that you can not predict how a player is going to play, which means that most of these stats will not be useful. From a machine learning perspective (i actually study this), the best chance we have to predict the outcome of the game is to extrapolate from the previous performance of the player, which is the current tetris rate up to this point. This might give misleading results in some cases, as discussed before, but it will give good predictions in way more cases than the pace stat in its current version.
One could advance that stat in many ways:
One way would be to include the players usual tetris rate over many games. This would reduce the influence of too bad or too good play in the current game. But this would be a lot of work to track in a tournament and it would be tedious to add all of the data. Another way to improve the stat would be to include the other players performance in the calculations. If the other player scored a lot in the lines that differentiate him compared to the slower player, then this probably means that a lot of favourable pieces will come and thus the slower player will most likely score better. But if you were to take this into account, you would need to account for the fact that the other player could just be the better player, which would not be easy as well.
But overall, the easiest way would be the method you described at first: Just save information about the scores that each player had at different line counts, and compare the score of the slower player to the score that the faster player had at this point. One of the reasons you dismissed this idea was that it is extemely tedious to enter these information into a spreadsheet. But after you began automating this process by programming, this is not a problem any more. You do not even have to restrict yourself to the 10-line threshold, but instead you can track the score continuously throughout the game and calculate the differential at any given point. Dealing with fractions of tetrisses should not be a problem either when using programming.
The other problem you had with this stat is that it is misleading if the other player burned a lot recently. This is true, but all stats discussed here had certian problems in some cases and this seems to be the least misleading one, or at least less misleading compared to the current pace stat, since this will always give misleading results if one player is far behind, and this stat would only give misleading results sometimes.
Some interesting stuff that might be useful to consider: In machine learning we have a concept called "unbiased estimator". This basically means, that if we calculate a value based on emprirical data (so the value would be out pace difference and our empirical data would be the current performance of both players), then the estimator of the value is unbiased if in its expectation (so if we average it over infinitely many games) it would on average give the correct results. The stat that extrapolates the players current tetris rate would be fairly close to being unbiased, sicne it sometimes overestimates the slower players performance and sometimes underestimates it, such that we have a good result on average (there are some other factors that falsify this statement, like the fact that it is harder to score tetrisses at level 19 compared to 18, but that would be too complicated to account for). This means, that this stat has a chance to provide useful information. The current pace stat on the other hand consistently overestimates the slower players performance, such that is is extremely far from being undbiased. such that the results are not that helpful in most cases.
Okay this is the result of me being bored at night ^^ I dont know if anyone is actually going to read this, but it nevertheless feels good to have commented my thoughts I guess :D
I'll like this comment just for the effort put in
This is incredible. There have been so many times that I've yelled at my phone because announcers are reporting on just score or even just tetris rate, when there was clearly more going on under the surface. It just gives the wrong impression to not only the audience, but the players as well -- there's nothing worse than hearing "a win is virtually impossible at this point" and for that to not be true. It's so much better to be able to hear "a win is here is gonna be tough, but if he plays extraordinarily well, it's still possible." No more blind guessing on the part of the announcers.
Wooow game scout this amazing. Thank you for everything you do for the Tetris community!
It's amazing how limited are people that don't know the basics of programing.
I love the part where he complains about how terrible Python distribution is. It's not his fault-it's definitely a meme at this point that Python is this way.
I love how aGameScout does the captions. A small detail but well appreciated :)
Thanks. Been making the effort to do them on newer videos
Great job, you guys! I hope this will start getting used in official tournaments.
Absolutely loved the explanation of everything. A true bliss to watch. Also, I'm so glad to hear that Jonas had the best Tetris Rate in 2018, he totally deserve it. 😢🙂
This is great, looking forward to seeing these stats displayed on games in the future
I am absolutely a much bigger faan of modern guideline vs tetris than NEStris, but the lengths the nestris community has gone to in order to advance the game they love so dearly is just breathtaking. Thank you so much for providing this information for those who don't follow nestris
With these new programming weapons now in your arsenal, I can't wait for the next CTWCs!! 😀 Excellent work from you and the Classic Tetris community! 👏👏👏
This is actually mind-blowing... Good job to everyone involved to make this a possibility...
You changed competitive Tetris commentary forever
amazing, incredible work everyone!!
This is so dope. More people are perfecting NES Tetris.
The way your Pace stat assumes perfect play parallel how in Ten-Pin Bowling coverage, Strikes and Spares are assumed to be worth their max value (so a Strike in the first frame is shown as being an effective score of 30, even though it isn't officially 30 until the next two bowls happen and both are also Strikes).
A thing to maybe consider (although explaining this is the hard part) is to have a comparison to an assumed Tetris rate that's based more on a general average. My rule thumb used to be "80% Tetris rate at level 18 speeds, 60% at level 19 speeds", but so many players are doing better than that these days that 90% and 70% might be more reasonable? Something to look at, perhaps.
Speaking of getting the Tetris lead wrong, some of it comes from players simplifying to "30k for a Tetris in the late game", and some of it comes from how a Tetris that increases the level is scored at the new level (this scoring quirk was unnoticed for quite some time).
Thank you for respecting the fact that Tetris is a complicated game, and for trying to make it more approachable.
This is supreme nerdiness. In-depth stats for Tetris. Excellent.
One of the best vids / projects you've ever done
What's the name of the song that starts playing in the background at about 06:30?
Have you come across the DLS metric from Cricket? The Duckworth-Lewis-Stern method is used to compare performance of cricket teams during a match, as they play sequentially, and interruptions would usually mean the game is abandoned. This allows a winner to be chosen, and is somewhat similar in idea to your pace stat.
It works by correcting score according to the difference in 'resources' available to the teams, specifically because in cricket there are two types of resources (balls and players not out) you can't easily scale a target score.
I wonder if you could do something similar, as with tetris rate you're compensating for the value of later levels only, and with the new stat you're compensating for value and play improvements, but perhaps there's a scaling factor for lines remaining (like the DLS tables) that more accurately represents real play.
You could find this by analysing past games to see if there is a relationship between lines remaining and average burn at that level across many players, using real behaviour rather than programmatically finding the optimal behaviour.
"Isn't this all a bit nerdy?"
- AGameScout -
My response to that is a proud yes.
i have never played tetris let alone NES tetris but these vids are so interesting
Try out Puyo Puyo Tetris or any of the newer Tetrises.
Good job, and interesting. Watching this also I still can’t believe Jonas is gone :(
Breaks my heart to hear "late" but Jonas showing his true skill years later is a nice tribute.
The pace stat is great in terms of analysis, but I do share the sentiment with how it predicts taking some of the fun out of it. Because players can only ever perform worse than the pace, not better.
I'm more interested in what the difference would be in their scores if they both were at the same number of lines (with some deviation for burns).
If player A is at line 29, and player B is at 34, I want to know if based on recent performance (i.e. last N lines tetris percentage), what does the scores look like if both were at 34.
That allows the player to play better than the pace prediction, and it's not about predicting what the match is gonna end up being, it's about showing you accurately what the current state is.
How could you use your current algorithm to do that?
Instead of estimating the score at 230 lines you could estimate it at the leaders current lines, or even leaders current lines plus some prediction amount like next 20 lines.
And instead of assuming perfect playing from both players maybe assume a reasonable upper limit.
I don't know, emotionally it definitely is a little daunting that the pace always assumes robotically perfect playing and we'd only see players basically not living up to that standard in chase downs. Secondly I like the aspect of trying to think about "we'll if they do this and this, maybe the score will just tip over".
If you were doing this for like odds analysis for betting, I'd call it a smart tool in addition to the usual suspects, but for viewing experience it doesn't make it more entertaining for me, because as stated, I already know who should be winning, af if they don't it's because compared to the prediction they "dropped the ball", even though we don't expect that level of play from them, that's what the algorithm suggests.
As a person who is extremely into basketball stats (NBA and WNBA mainly) this video was an absolute treat to watch.
"More more more!"
"The late Jonas Neubauer...."
he understood the pain of the "pip install"
Calculating pace is all fun and games until Cheez comes in with consistent 200k+ killscreens via rolling. No lead is safe!
😳
@@aGameScout oh no.
@@CobaltTetris funnily enough, a year later with roll chase downs now being a reality, I think the pace statistic is fortunately still far from being obsolete. Most players in the scene aren’t elite rollers and even the elite rollers die quite often without scoring much on 29. Pace is not a guarantee of who wins, but it is definitely a massive advantage to be ahead 100-200k into 29 no matter who you’re facing!
Wow. I didn’t understand most of this but it was so impressive!
So glad Yobi is recognized for his effort
The next major leap forward will be to alter the max score possible based on board position. For example if there are many holes or a misdrop instead of a right side well the max score should shift to reflect that. That means a new metric and analysis on board position based on factors like number of holes, stack shape, height, level, and well-size, and accessibility.
Once this blows up, the official NES Tetris might add it
You folks have revolutionized Tetris media 😮
Master of Tetris plus an Excel master. Outstanding video with brilliant explanations!!
glad to see nes Tetris footage is getting easier to use
Imagine how viral and practical a romhack that does the CTWC interface would become.
So... when are they gonna implement this? I haven't seen it in any CTC matches and it's kinda aggravating
That's what I've been wondering
aGameScout 2018: I'd like to use my video editing skills to cover competitive Tetris
aGameScout 2022: I am a PhD candidate in computer science.
I think this is great, but I had a question on one of your comments. You said a problem was that it the one method wouldn't account for the fact that dog has been burning since 210 lines. But doesn't the final method you came up with also not account for that? What I mean is, let's say dog was at 210 lines 1 tetris lead, but a messy play field, vs andy 210 lines clean play field. Wouldnt it still say that dog has the pace lead because it assumes he will tetris all the lines remaining regardless of playfield?
yeah, currently detecting a dirty playing field and accommodating for that is too complicated and abstract to work in at the moment. In the situation described, andy would be in the better spot, however in terms of raw scoring potential, depending on how bad dog's board was, he could potentially go for dirty tetrises if he really had to, the raw score does still have some application
@@aGameScout ahh ok thanks for the reply. I was just wondering if you were actually able to account for it in the final method. You're right that seems way too hard to calculate. I can't even think of what the logic would be for that. Mayne stack height vs deepest alley available. But yea sooo complicated
Bro has never heard of rolling bruh 💀
I made this video RIGHT before the rolling playstyle became popular 😭
@@aGameScout Oh, I'm so sorry.
Hey, I wounder, what about the stats for "Burns Available"
with that i mean, how many burns can the player behind effort to still win
Let's put it this way...
Jonas was as good as he was before community involvement.
He, and a handful of players, invented the wheel. Watching his gameplay is still the most soothing, logical, approach to the game. He does something that few others match.
I present: Master Zentris Neubauer!
Gotta love Free Open Source Software projects ❤️ huge shoutout to the maintainers and contributers!
Thank you so very much for the great work you contribute to the scene!
16:25 The command line isn't hard when you get used to it, although Windows' command line can be a bit painful at times since it has been left to rot in favor of graphical apps.
How am I just discovering this masterpiece!
This truly is a video with extreme levels of chonk
Each game of tetris has a time limit.
Rolling - yes but how about no.