Hi Mack, Terrific multi layered analysis for this game , and glad you found the content worthwhile. You have set the bar very high with next level analysis, and setting up the punchline and keeping viewers in suspense. Just very good storytelling weaved in. In this game, fortunately I did have a lot of time left for the endgame. I really wanted to leave one in the bag, but kept trying to weigh the odds of a D in the bag, along with Jon missing the weird bingo spot with all the clunky letters. His last TWO plays of dot thru an O for 8 pt, and EF for 20 pt, indicated more likely there was at least one D in the bag, and more likely the S was on his rack. I didn't get thru all the possibilities, especially NOM which is amazing. I did notice that Jon had only 3:30 left on his clock, so while not exactly blitz, the chance of an error might be higher. The 25 % (or lower, given more chance of a D draw) chance he saw and blocked the I to partially quantify is also cool. For the record, Jon was a terrific sport about the game afterwards. Giving up walk off bingos we may not have seen has happened to every one of us (just less so for players at Mack's level).
A friend of mine once saw his opponent with ALNSUUU after he finished his tracking in an endgame and assumed he had won since his opponent had 3 Us on his rack and he was far ahead score wise. He was wrong in his assumption and lost the game.
In Jon's position you got to sometimes keep DDD on your rack to be a bit unpredictable and provoke David into going for the bluff (joking, taking the poker stance a step further)
Chess engines have a "contempt" factor that varies their tendency to go for slightly suboptimal, but less drawish moves. Maybe BestBot / Macondo could get something similar in the future. BestBot's contempt factor could depend on its opponent's Woogles rating so that it would play OR against me and ZOO against you!
I feel like it's also worth considering that one would expect Jon to be slightly more alert towards sniffing out & shutting down bingos if david plays two tiles on his previous turn rather than one, so the 25% estimation that he doesn't neutralise the bingo threat should be tweaked to be a lower estimation really interesting video as always :D
Thanks! I'm not sure about the logic for Jon being more alert after David plays two tiles though -- typically the fewer tiles one plays, the more that tends to look like a fish in general
I also think it's worth mentioning that Jon is very unlikely to have the A. If he did have the A, he would probably play FA or AD on the right, in order to play two tiles instead of one. In fact, I think Jon should have played something like OD or ID at O1, in order to leave just one tile in the bag instead of two.
It sounds rough to have a fraction of an 11% chance to win, but not as rough as committing to 3% odds against a player who isn’t a lock to play optimally. Human error may work out in my favor with unexpected frequency; a 3% chance will play out exactly as expected.
As a statistician, this calls for a bayesian analysis. You can set a distribution on both the chance of the mistake and the tile distribution. So you can kind of quantify the end result
Amazingly, Jon played the first D of the game with DOT for 8 points the turn before OF/EF. For the same score DUD/ID makes a lot of sense with how many D's are still left. For instance, if Jon had 3 D's and 2 T's on his rack with 1 D unseen, then DUD/ID makes more sense then DOT as it doesn't risk triplicated D's. This suggests that at least 2 D's are in the bag before Jon's play of DOT when the bag had 9 tiles. After DOT, SLOP and OF/EF, Jon drew 3 of those 9 tiles, meanwhile David didn't draw a single D, so in the scenario when Jon had 2 D's on his rack before DOT and 2 D's are still in the bag, he had a 7/84=1/12 chance to draw the 2 D's. In this scenario there's a 11/12 chance a D is still in the bag that David will pick up after a 2-tile fish. If Jon had 1 D on his rack before DOT then there's a 83/84 chance a D is in the bag. Simply put you have to be much more prepared to draw a D after a 2-tile fish than any other tile. With all this in mind, the win percentage after a move like ZOO is in the ballpark of (1/12)(1/15)=1/180 (No D's are in the bag * drawing AS from AINOST 1 out of 6C2=15) which is 5 times worse then 1/36, it might be even worse. Meanwhile drawing AD after OR/BAR is (11/12)(1/8)=11/96 which is over 20 times more likely then hitting an AS draw after ZOO. So the real question is "Is Jon going to see IMBLAZED 20/20 times under time pressure?". If the answer is "No" then he might block it anyway, but taking the gamble was definitely worth a shot.
Fascinating. And this gets a quantifiable 11/12 number, close to what I was trying to estimate in my head re the chance of one tile being a D... I shared the video with another local expert player, who studied it carefully ...Possibly one inference missed was why EF/OF ? Had Jon had the case A, it would have been logical for him to play FAH and FA upper right part of board for 4 more points, and leave 1 in the bag rather than 2. No reason to keep the A on his rack, unless he was trying for his own out bingo with slop/s, but with the lead he didn't need to bingo. So if we assume the A was still in the bag at ef/of, wouldn't that make the A pull a 66.7/33.3 % proposition, and the D pull still 11/12 with your previous inferences and math. So, 11/12 * 0.667 would be considerably higher than 4 of 36, and would be a 61.1 % chance of drawing what I did-- A and D. Correct me if I'm wrong. As Mack points out, Jon still has to miss seeing IMBLAZED...but now maybe we have gone from Hail Mary field position to 4th and goal at the 5 yard line. Lol
I got as far as SEZ, and would probably have landed there in-game. The clock is another unquantifiable factor here: I’d be more inclined to go for the steal (assuming I saw it!) if Jon had 30 seconds on his clock than if he had 10 minutes.
Intriguing situation. A lot of scrabble related videos seem to assume perfect play, which isn't always going to happen. The % of times it actually does happen in top tournaments is kind of amazing though.
Another interesting consideration: do any of the three options that give a chance of an unstoppable bingo also come with a backup chance of a single-out bingo that could potentially be missed?
ZOO also threatens LIMBERS/SLOPS with an IS draw, which can be blocked but it needs to be done carefully, so ZOO is probably the best of the three theoretically winnable plays
I actually think SEZ was the hardest one to find. MARBLES/GAMBLERS and BLAZERS are both really easy bingoes to see as follow-up plays, even a casual player might notice. NOMBRILS not so much.
Exactly. I have no idea what a NOMBRIL is…. Thought process behind NOM was that it takes both a back A and E. I saw the BLAZERS out and assumed there might be some other word hooking NOM with an A
@@PokeminhGottaSocializeEmAll I'd also argue that even if you know NOMBRIL, out of the winning moves SEZ is the play that requires the most advanced board vision to see. ZOO/NOM are both staring you in the face and while you could argue NOM isn't a common 3-letter word, neither is SEZ. Personally I'd rank ZOO as the easiest and NOM as the middle difficulty, those were the two I found and I didn't see SEZ (even though I know NOMBRIL.)
Hi Mack,
Terrific multi layered analysis for this game , and glad you found the content worthwhile. You have set the bar very high with next level analysis, and setting up the punchline and keeping viewers in suspense. Just very good storytelling weaved in.
In this game, fortunately I did have a lot of time left for the endgame. I really wanted to leave one in the bag, but kept trying to weigh the odds of a D in the bag, along with Jon missing the weird bingo spot with all the clunky letters. His last TWO plays of dot thru an O for 8 pt, and EF for 20 pt, indicated more likely there was at least one D in the bag, and more likely the S was on his rack. I didn't get thru all the possibilities, especially NOM which is amazing. I did notice that Jon had only 3:30 left on his clock, so while not exactly blitz, the chance of an error might be higher. The 25 % (or lower, given more chance of a D draw) chance he saw and blocked the I to partially quantify is also cool.
For the record, Jon was a terrific sport about the game afterwards. Giving up walk off bingos we may not have seen has happened to every one of us (just less so for players at Mack's level).
p.s. luv the little comic quotes on the cover photo.
A friend of mine once saw his opponent with ALNSUUU after he finished his tracking in an endgame and assumed he had won since his opponent had 3 Us on his rack and he was far ahead score wise. He was wrong in his assumption and lost the game.
How unusual…
Was he like... "Uuu have got to be kidding me?!"
Now Mack is out here thinking like a poker player.
In Jon's position you got to sometimes keep DDD on your rack to be a bit unpredictable and provoke David into going for the bluff
(joking, taking the poker stance a step further)
Today I learned that NOMBRILS is a word.
Chess engines have a "contempt" factor that varies their tendency to go for slightly suboptimal, but less drawish moves. Maybe BestBot / Macondo could get something similar in the future. BestBot's contempt factor could depend on its opponent's Woogles rating so that it would play OR against me and ZOO against you!
Mack really out here trying to justify his SPC finals play, enticing Ian to miss eschar 😂
This is definitely one of your best and most interesting analyses. Really puts into perspective how complex scrabble is
Thanks for watching!
I feel like it's also worth considering that one would expect Jon to be slightly more alert towards sniffing out & shutting down bingos if david plays two tiles on his previous turn rather than one, so the 25% estimation that he doesn't neutralise the bingo threat should be tweaked to be a lower estimation
really interesting video as always :D
Thanks! I'm not sure about the logic for Jon being more alert after David plays two tiles though -- typically the fewer tiles one plays, the more that tends to look like a fish in general
Keep this pre end game videos coming
I also think it's worth mentioning that Jon is very unlikely to have the A. If he did have the A, he would probably play FA or AD on the right, in order to play two tiles instead of one. In fact, I think Jon should have played something like OD or ID at O1, in order to leave just one tile in the bag instead of two.
It sounds rough to have a fraction of an 11% chance to win, but not as rough as committing to 3% odds against a player who isn’t a lock to play optimally.
Human error may work out in my favor with unexpected frequency; a 3% chance will play out exactly as expected.
As a statistician, this calls for a bayesian analysis. You can set a distribution on both the chance of the mistake and the tile distribution. So you can kind of quantify the end result
Amazingly, Jon played the first D of the game with DOT for 8 points the turn before OF/EF. For the same score DUD/ID makes a lot of sense with how many D's are still left. For instance, if Jon had 3 D's and 2 T's on his rack with 1 D unseen, then DUD/ID makes more sense then DOT as it doesn't risk triplicated D's. This suggests that at least 2 D's are in the bag before Jon's play of DOT when the bag had 9 tiles. After DOT, SLOP and OF/EF, Jon drew 3 of those 9 tiles, meanwhile David didn't draw a single D, so in the scenario when Jon had 2 D's on his rack before DOT and 2 D's are still in the bag, he had a 7/84=1/12 chance to draw the 2 D's. In this scenario there's a 11/12 chance a D is still in the bag that David will pick up after a 2-tile fish. If Jon had 1 D on his rack before DOT then there's a 83/84 chance a D is in the bag. Simply put you have to be much more prepared to draw a D after a 2-tile fish than any other tile. With all this in mind, the win percentage after a move like ZOO is in the ballpark of (1/12)(1/15)=1/180 (No D's are in the bag * drawing AS from AINOST 1 out of 6C2=15) which is 5 times worse then 1/36, it might be even worse. Meanwhile drawing AD after OR/BAR is (11/12)(1/8)=11/96 which is over 20 times more likely then hitting an AS draw after ZOO. So the real question is "Is Jon going to see IMBLAZED 20/20 times under time pressure?". If the answer is "No" then he might block it anyway, but taking the gamble was definitely worth a shot.
Do you take the gamble or bingo out with GAMBLERS? That is the question.
Fascinating. And this gets a quantifiable 11/12 number, close to what I was trying to estimate in my head re the chance of one tile being a D... I shared the video with another local expert player, who studied it carefully ...Possibly one inference missed was why EF/OF ? Had Jon had the case A, it would have been logical for him to play FAH and FA upper right part of board for 4 more points, and leave 1 in the bag rather than 2.
No reason to keep the A on his rack, unless he was trying for his own out bingo with slop/s, but with the lead he didn't need to bingo.
So if we assume the A was still in the bag at ef/of, wouldn't that make the A pull a 66.7/33.3 % proposition, and the D pull still 11/12 with your previous inferences and math.
So, 11/12 * 0.667 would be considerably higher than 4 of 36, and would be a 61.1 % chance of drawing what I did-- A and D. Correct me if I'm wrong.
As Mack points out, Jon still has to miss seeing IMBLAZED...but now maybe we have gone from Hail Mary field position to 4th and goal at the 5 yard line. Lol
I got as far as SEZ, and would probably have landed there in-game. The clock is another unquantifiable factor here: I’d be more inclined to go for the steal (assuming I saw it!) if Jon had 30 seconds on his clock than if he had 10 minutes.
Nice! And yeah, totally -- based on David's comment above sounds like Jon had around 3 minutes, so somewhere in the middle of your two extremes
Intriguing situation. A lot of scrabble related videos seem to assume perfect play, which isn't always going to happen. The % of times it actually does happen in top tournaments is kind of amazing though.
Another interesting consideration: do any of the three options that give a chance of an unstoppable bingo also come with a backup chance of a single-out bingo that could potentially be missed?
ZOO also threatens LIMBERS/SLOPS with an IS draw, which can be blocked but it needs to be done carefully, so ZOO is probably the best of the three theoretically winnable plays
Okay but, against Nigel, are you more likely to draw AS or IS?
Saw NOM and not the others!!
I actually think SEZ was the hardest one to find. MARBLES/GAMBLERS and BLAZERS are both really easy bingoes to see as follow-up plays, even a casual player might notice. NOMBRILS not so much.
Exactly. I have no idea what a NOMBRIL is…. Thought process behind NOM was that it takes both a back A and E. I saw the BLAZERS out and assumed there might be some other word hooking NOM with an A
@@PokeminhGottaSocializeEmAll I'd also argue that even if you know NOMBRIL, out of the winning moves SEZ is the play that requires the most advanced board vision to see. ZOO/NOM are both staring you in the face and while you could argue NOM isn't a common 3-letter word, neither is SEZ.
Personally I'd rank ZOO as the easiest and NOM as the middle difficulty, those were the two I found and I didn't see SEZ (even though I know NOMBRIL.)