| Player 1 |
|
#player1 Quackle_0.97 Quackle 0.97
#player2 Elise_0.0.8_dev Elise 0.0.8 dev
#description Saved by Elise version 0.0.8 dev
#lexicon CSW12
>Quackle_0.97: DFOPPT? 8G FOP +16 16
#note A game between Quackle's championship player and Elise's 10 minute player, using the CSW12 lexicon.
#note Best. DFOPPT? is better than 70% of all opening racks.
>Elise_0.0.8_dev: AINRSST J2 SANTIRS +70 70
#note Elise thinks Quackle is about 33% likely to hold the blank, and that Quackle's average rack is 11.4 points better than average.
#note Elise has six different 70-point bingos here: STRAINS J8, INSTARS J6, INSTARS J2, SANTIRS J2, STRAINS J2, SANTIRS J8. With the exception of SANTIRS J8 (which is definitely inferior -- it gives up higher scoring opponent bingos) they are all very close to each other.
#note 5 ply simulation likes STRAINS J8 a little better (-3.01/57.8%/58.6% versus -6.24/56.8%/58.4%) but the difference in mean win probability is small and the difference in median winprob is even smaller.
#note The starting rack of AINRSST is of course excellent, better than 87% of all racks.
#note 3-ply winprob simulation (880), -4.39 / 57.2% [57.19s]
#note Time remaining on clock: [09:01.00]
>Quackle_0.97: ADOPTU? I3 POUT +22 38
>Elise_0.0.8_dev: AAEORUX K3 ROUX +46 116
#note Time remaining on clock: [08:09.16]
>Quackle_0.97: ADEJKY? 9C JAKEY +43 81
>Elise_0.0.8_dev: AAEEGOT 10B TOGAE +32 148
#note This move was not simulated by Elise's clock player -- we are not late in the game, and it was considered superior enough to the alternatives by static evaluation to just play immediately. According to Elise's 5-ply simulation, this move is indeed best: +22.92/69.6%/71.9% in 2300 sims. GAE 10B is second best with 68.0% mean winprob/70.8% median winprob. Also good is GOA 10B (67.3% mean), everything else is 65% or below.
#note Time remaining on clock: [08:08.85]
>Quackle_0.97: DGGHIN? D8 HAGGING +34 115
>Elise_0.0.8_dev: AAAENRS 11A SARGE +38 186
#note 3-ply winprob simulation (2480), +27.00 / 72.4% [55.19s]
#note Time remaining on clock: [07:12.82]
>Quackle_0.97: DEILNV? 2C VILDNEsS +71 186
#note Only bingo.
>Elise_0.0.8_dev: AADHINO 1G AHA +42 228
#note 3-ply winprob simulation (3280), +2.07 / 62.3% [46.83s]
#note Time remaining on clock: [06:25.35]
>Quackle_0.97: ACDEMOR A11 SCRAM +36 222
>Elise_0.0.8_dev: DEEINOY B13 EYE +28 256
#note Up to this point, although Quackle had a blank and Elise didn't, Elise's tile luck is probably better than Quackle's. Quackle has had 3 rack draws out of 6 that scored better than the median draw; Elise has had 5 out of 6. Elise's best draw up to this point was better than Quackle's, and Quackle's worst draw up to this point was worse than Elise's.
#note 3-ply winprob simulation (1680), +2.97 / 63.2% [30.97s]
#note Time remaining on clock: [05:53.74]
>Quackle_0.97: DEEIOR? 3I PAREO +14 236
#note At this point in the game, the tile luck reverses a bit -- this is Quackle's first truly excellent draw. DEEIOR? is better than 95% of all racks drawn from DEO.
#note However, this is the first move Quackle makes that Elise strongly disagrees with.
#note Quackle has several possible bingos, the best being (G)OiTERED# at 14D for 63 points. Quackle is afraid of counterplay using the TWS at H15. Obviously DEIR? is an excellent leave, and a lot of draws give Quackle a 75+ point play next turn -- but in this position surely you want the points now? Points are scarce on this board, and while a 63 point bingo is not huge, it's still in fact better than what's possible with 95% of all of Quackle's possible racks from this position...
#note Elise at 5 ply thinks this is a major mistake, with (G)OiTERED being -6.51/59.9%/64.4% and (PAR)EO only -14.51/54.0%/57.8%.
>Elise_0.0.8_dev: BDEINNO 4H IONONE +20 276
#note Elise at 5 ply thinks this move also is a mistake. Elise (correctly) figures Quackle's rack is bingo-prone, +12.8 on average with the blank much more likely than chance. At 5 ply, however, Elise favors B(I)NDI D1 by a significant margin -- 59.8% mean winprob versus only 53.1% for I(ONO)NE.
#note In its favor, I(ONO)NE blocks a number of opponent bingos on the next turn.
#note 3-ply winprob simulation (3360), -2.51 / 60.4% [63.41s]
#note Time remaining on clock: [04:49.76]
>Quackle_0.97: DEILQR? D1 QI +21 257
#note Best move, under the circumstances, by miles.
#note This draw is why you take the points -- the draw of LQ is bad, but hardly unthinkably bad, better than 22% of all possible draws.
>Elise_0.0.8_dev: BDINOTW 2L TWO +21 297
#note The 5-ply player likes (L)OWND# E2 a little better, though the difference is small (55.0% mean winprob versus 54.6% winprob).
#note 3-ply winprob simulation (3440), -8.31 / 57.7% [58.31s]
#note Time remaining on clock: [03:50.95]
>Quackle_0.97: DEIILR? C2 VILD +16 273
#note Quackle has four possible bingos, the best of which is (G)RIDELIn# 14D for 63 points.
#note However, all four bingos give substantial counterplay at the TWS at H15... we saw this just a couple of moves before, but the situation is now different. We're closer to the end of the game and a killer play using the TWS now would be a lot more damaging. Quackle is correct here to skip the bingo and play elsewhere.
#note Elise likes (V)ID C2 slightly better, but only slightly... (V)ID is -15.77/50.6%/52.9% while (V)ILD# is -16.30/50.2%/50.8%.
>Elise_0.0.8_dev: BBCDFIN 4B BLIND +16 313
#note This is the second worst tile draw of the game.
#note At 5-ply, Elise actually likes "exchange all" better than BLIND, but it is very close -- the exchange is -8.13/56.4%/63.9%, while B(L)IND 4B is -3.01/55.5%/64.9%. Median winprob and point spread are actually a little better for B(L)IND, so this is basically a toss-up.
#note 3-ply winprob simulation (4400), -7.32 / 58.7% [43.66s]
#note Time remaining on clock: [03:06.89]
>Quackle_0.97: EIIORU? 6C EUOI +15 288
>Elise_0.0.8_dev: BCFLLMN -BCFLLMN +0 313
#note This time Elise does exchange all... the draw of LLMN to BCF is the worst tile draw in this game, better than only 0.4% of all draws -- 1 in 250!
#note Elise figures that "Exchange all" and "Exchange BCFLLM" are the best plays by a large margin -- the two exchanges give winprobs of 59.1% and 56.8% respectively, while everything else is at 33.3% or lower.
#note 3-ply winprob simulation (560), -9.38 / 57.3% [11.22s]
#note Time remaining on clock: [02:55.33]
>Quackle_0.97: EIIORT? I8 POInTIER +61 349
#note This move is a significant mistake according to Elise. The problem is that the Z is by far the single tile most likely to win the game for Elise, and after the exchange 7, you have to assume Elise is quite likely to have it.
#note Much better than POInTIER is simply IO# 1N -- it uses the TWS at O1 (where the Z could otherwise go), has the EIRT? leave, and does not leave row 15 wide open. Elise 5 ply gives IO a 64.9% win probability.
#note Quackle at this point thinks it has about a 72% win probability, while Elise's 5 ply gives the more moderate figure of 58.8%. The difference here is that Elise is playing through to the end of the game at this point, while Quackle is still only looking ahead a couple of ply.
#note Quackle could also have played a bingo, (N)OTIfIER at 13D, that scores only 2 points less but is considerably safer (64.4% vs. 58.8%).
#note In longer simulations Quackle likes (N)OTIfIER better, but it does not "get" the importance of IO except at much longer and deeper settings.
>Elise_0.0.8_dev: ADEESVZ 15G VERD +24 337
#note 3-ply winprob simulation (816), +28.00 / 82.3% [35.41s]
#note Time remaining on clock: [02:19.58]
>Quackle_0.97: ACFMNRT 13I INFARCT +24 373
#note At this point the game is basically lost if Elise has the Z (and vowels to go with it)... Quackle is doubling down creating a hotspot at L12, but it only matters a little. This is why IO 1N on the previous turn was so much better.
#note A human player would likely assume, after VERD 15G, that Elise wasn't using its last vowel -- VERD was not a huge scoring play after all -- but instead assume it was likely Elise saved vowels for the endgame. For that reason INFARCT 13I seems a little more reckless than INFRACT 13I.
>Elise_0.0.8_dev: ABELSUZ 12L ZA +46 383
#note Winning. Also winning are BUS(T)LE O10, ZA 1N, and (C)UZ# N13.
#note Time remaining on clock: [02:18.64]
>Quackle_0.97: LMW N13 CWM +20 393
#note Quackle gave up a total of 19.38 percentage points of win probability, compared to Elise 5-ply. Elise's 10-minute player gave up 8.99 percentage points of win probability, compared to Elise 5-ply. Up until the later part of the game, Elise's tile luck was a little better than Quackle's, but later in the game Elise's tile luck was quite bad.
#note Both players had an equal number of rack draws that Elise thinks are better than the median, and an equal number below the median. Of course, Quackle had both blanks and made two bingos to Elise's one. Despite all that, Elise won. Who was luckiest?
>Elise_0.0.8_dev: BELSU O8 SUBLET +27 410
#note And two minutes still left on the clock.
#note Time remaining on clock: [02:18.64]
>Elise_0.0.8_dev: (L) +2 412
|
