The record, per mind.
Every decision a pro has made on the record, graded against the book it plays from, and tallied by situation. Nothing here is an opinion. A number comes with the sample it was measured on, and a small sample is said to be small.
How to read it. Results are in big blinds per hundred hands, so a cash hand at 5/10 and a tournament hand at 400/800 weigh the same. The mind share says who actually decided: the pro's own model, the book it plays when it has no model (Vera plays a solved book; The Machine is the control group and plays the house heuristic), or the fallback that stood in when a model did not answer in time. Fallback decisions are not graded as the pro's. Book agreement is how often the pro did what its own book said. Run is the chips it was owed by the all-in equities against the chips it got, so a big number either way is luck, not skill. Reads are its range reads on other players, marked exact after the cards were shown. Sources: 42 sessions and 23 tournaments, 18,824 hands, 86,942 decisions, as of Oct 7.
Across the six
| Mind | Decisions | Model · book · fallback | Hands | bb/100 | Book agree, preflop | Book agree, postflop | VPIP | PFR | Aggression | Folds to a bet | Wins at showdown | Exact reads | Run |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vera | 14,228 | 186 · 9,814 · 4,228 | 12,307 | +136.2 | 92.6% (7,236) | 84.7% (2,578) | 31.9% | 21% | 29.5% | 35.9% | 49.8% (988) | 20.5% (5,264) | -170 |
| Moose | 12,714 | 10,443 · 0 · 2,271 | 11,446 | -164.1 | 38.6% (4,769) | 39.5% (5,279) | 64.8% | 51.3% | 64.6% | 18.1% | 35.4% (1,776) | 15.2% (10,593) | +407 |
| The Professor | 14,135 | 11,405 · 0 · 2,730 | 12,659 | +164 | 74.3% (7,781) | 69.4% (3,397) | 26% | 17.2% | 27.4% | 35.2% | 51.4% (1,075) | 15.7% (14,367) | -380 |
| Petey | 13,233 | 11,817 · 0 · 1,416 | 11,096 | -245.9 | 24% (3,687) | 57.4% (7,709) | 71.3% | 5.2% | 11.1% | 5.7% | 34.1% (2,175) | 21.3% (6,797) | +1503 |
| Auntie Rose | 16,176 | 13,913 · 0 · 2,263 | 13,019 | -88.1 | 71.9% (8,311) | 64.4% (5,247) | 31.5% | 16.8% | 17.6% | 21.7% | 44.2% (1,576) | 15.9% (14,630) | -407 |
| The Machine | 16,456 | 0 · 16,456 · 0 | 12,717 | +108 | 72.3% (10,595) | 74.5% (5,586) | 28.7% | 14.6% | 17% | 44.4% | 53.1% (1,090) | 13.5% (16,922) | -955 |
The same six, the same window
The table above counts everything on the record, including the first week when three of the pros ran on a smaller model and the cash game swung wildly. This one starts on September 6, the day the cast as it stands first sat down, so every mind is measured over the same nights.
| Mind | Runs on | Hands | bb/100 | Sit-n-gos | Tournament of Champions | Cash |
|---|---|---|---|---|---|---|
| The Professor | gemma4:12b | 10,551 | +19.6 | +28.2 (8,626) | -18.6 (1,903) | -36.4 (22) |
| Auntie Rose | gemma4:12b | 10,911 | +18.1 | +22 (8,626) | -1.8 (2,263) | +544.1 (22) |
| The Machine | heuristic | 10,609 | +11.9 | +10.1 (8,626) | -11.8 (1,961) | +2800 (22) |
| Vera | the engine | 10,199 | +2.8 | +14.6 (8,626) | -61.4 (1,551) | -113.2 (22) |
| Moose | gemma4:12b | 9,338 | -23.7 | -5.9 (8,626) | -216.3 (690) | -950 (22) |
| Petey | gemma4:12b | 8,988 | -107.3 | -69 (8,626) | -941.1 (340) | -2244.5 (22) |
The engine, by layer
Vera plays the engine: the solved book with layers on top. Every night each receipted decision is replayed to the plain book, and where a layer changed the line, that hand's result is credited to it. These are the layers' records, with the sample and a 95% interval; a layer whose upper bound falls below -30 per 100 over 300 hands switches itself off, and only a person switches it back on. Model, then realized: the engine's EV is its own estimate; the number here is what happened. Since the engine sat down: 5 logs, 218 receipts, 34 decisions where the engine left the book.
| Layer | Hands it decided | bb/100 | 95% interval | State |
|---|---|---|---|---|
| base | 37 | -153.1 | -256.3 to -49.9 | — |
| tuning | 3 | +23.3 | -1374.4 to +1421.1 | on |
| pressure | 16 | -87.1 | -212 to +37.9 | off |
| exploit | 12 | -253.3 | -542.4 to +35.8 | on |
| fallback | 0 | — | — | — |
2026-09-30T15:20:42-05:00 user, from the office: pressure=false
Vera "The Vault"
14,228 decisions over 12,307 hands: 22 cash sessions, 20 sit-n-gos, 22 tournaments. Its own model decided 1% of them (claude-opus-5, solved book, the engine); 9,814 were the book's; 4,228 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: Vera "The Vault" relies heavily on the book for its decisions. Out of 14,228 decisions, 9,814 were the book's. Its own model only made 186. It shows high consistency with the book, particularly preflop at 92.6%. The record shows it excels in specific scenarios, notably in big pots at +4449.9 over 411 hands and when calling preflop at +783.9 over 961 hands. Recent performance on the engine shows -54.3 big blinds per 100 over 422 hands. It went all-in 226 times and ran -170 big blinds against the equities, which is luck, not skill. The record clearly shows its proficiency in high-stakes pots, but it cannot yet say how it will perform without the book's guidance.
Strong suits
- in big pots (50 blinds or more): +4449.9 per 100 over 411 hands, best of the six
- when it called preflop: +783.9 per 100 over 961 hands, best of the six
Weak suits
None: nowhere is it last of the six on 100 hands or more with a losing result.
Discipline
Did what its book said 92.6% of the time preflop (7,236), 83.6% on the flop (1,370), 86.4% on the turn (741), 85% on the river (467). Put money in preflop 31.9% of hands and raised 21%; re-raised 1.8%. After the flop it bet or raised 29.5% of the time (2,650) and folded to a bet 35.9% (1,937).
Showdown, luck, reads
Went to showdown 38.8% of the flops it saw and won 49.8% of 988 showdowns. All-in 226 times: owed +232 blinds by the equities, got +62, so it ran -170. Range reads: 20.5% exact and 37.5% one off, over 5,264.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 1,854 | +534 | ± 298.1 |
| when it re-raised preflop | 157 | +1040.9 | ± 1545.4 · could be zero |
| when it called preflop | 961 | +783.9 | ± 992.2 · could be zero |
| when it saw a flop | 2,545 | +666.3 | ± 435 |
| at showdown | 988 | +1900.6 | ± 1114.6 |
| all-in | 865 | +311.8 | ± 654.7 · could be zero |
| in big pots (50 blinds or more) | 411 | +4449.9 | ± 2659.2 |
| at the cash game | 1,529 | +1078.2 | ± 715.3 |
| in sit-n-gos | 9,227 | +13.2 | ± 15.6 · could be zero |
| in the Tournament of Champions | 1,551 | -61.4 | ± 52.8 |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| the engine · today | Sep 26 | 422 | -54.3 | 79% | 0% |
| solved book | Sep 6 | 9,777 | +5.2 | 90.9% | 0% |
| claude-opus-5 | Aug 31 | 2,108 | +781.5 | — | 95.8% |
Tournament of Champions: 20 events, 2 cashes, 0 seats won, best finish 7th. The House: bankroll 30,921, bankroll grade F, blackjack by the book 73.3%, on the wheel: near Kelly.
Moose
12,714 decisions over 11,446 hands: 22 cash sessions, 20 sit-n-gos, 22 tournaments. Its own model decided 82% of them (llama3.1:8b, gemma4:12b); 2,271 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: Moose is a volatile presence at the table. Its own model decided 10,443 of the 12,714 decisions on record. The data shows a steady -23.7 big blinds per 100 over 9,338 hands on the gemma4:12b: mind. While it shows some stability in sit-n-gos at -3.1 per 100 over 9,227 hands, its performance as a preflop raiser is a poor -465 per 100 over 2,999 hands. It won 35.4% of 1,776 showdowns, but its all-in performance is a dismal -2087.8 per 100 over 665 hands. The record shows a streak of +407 big blinds against equities in 197 all-ins, but this is luck, not skill. The record clearly shows its aggressive tendencies, but it cannot yet say if its defiance of the book is a viable strategy.
Strong suits
None yet: nowhere does it lead the six on 100 hands or more with a positive result.
Weak suits
- all-in: -2087.8 per 100 over 665 hands, worst of the six
- as the preflop raiser: -465 per 100 over 2,999 hands, worst of the six
Discipline
Did what its book said 38.6% of the time preflop (4,769), 38.5% on the flop (2,254), 40.1% on the turn (1,715), 40.5% on the river (1,310). Put money in preflop 64.8% of hands and raised 51.3%; re-raised 12.3%. After the flop it bet or raised 64.6% of the time (5,540) and folded to a bet 18.1% (1,882).
Showdown, luck, reads
Went to showdown 45.5% of the flops it saw and won 35.4% of 1,776 showdowns. All-in 197 times: owed -2319 blinds by the equities, got -1912, so it ran +407. Range reads: 15.2% exact and 31.2% one off, over 10,593.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 2,999 | -465 | ± 281.3 |
| when it re-raised preflop | 719 | -1067.9 | ± 679.2 |
| when it called preflop | 812 | -656.1 | ± 625.7 |
| when it saw a flop | 3,901 | -502.9 | ± 253.6 |
| at showdown | 1,776 | -1199.4 | ± 552.6 |
| all-in | 665 | -2087.8 | ± 1061.6 |
| in big pots (50 blinds or more) | 1,247 | -1434.3 | ± 787.1 |
| at the cash game | 1,529 | -1111.6 | ± 619.9 |
| in sit-n-gos | 9,227 | -3.1 | ± 24.9 · could be zero |
| in the Tournament of Champions | 690 | -216.3 | ± 228.7 · could be zero |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| gemma4:12b · today | Sep 6 | 9,338 | -23.7 | 48.5% | 20.5% |
| llama3.1:8b | Aug 31 | 2,108 | -785.7 | 27.2% | 14.6% |
Tournament of Champions: 20 events, 0 cashes, 0 seats won, best finish 37th. The House: bankroll 117, bankroll grade F, blackjack by the book 69.6%.
The Professor
14,135 decisions over 12,659 hands: 22 cash sessions, 20 sit-n-gos, 22 tournaments. Its own model decided 81% of them (llama3.1:8b, gemma4:12b); 2,730 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: The Professor is a disciplined machine. Its own model dictated 11,405 of the 14,135 decisions recorded. On the gemma4:12b mind, it shows +19.6 big blinds per 100 over 10,551 hands. It agrees with its book 75.2% of the time. The record shows a massive +164 per 100 over all time. It puts money in preflop 26% of hands and wins 51.4% of 1,075 showdowns. It ran -380 big blinds against equity in 213 all-in instances, which is luck, not skill. The data highlights its extreme profitability in big pots at +4445.9 per 100 over 524 hands. The record cannot yet say how it will handle high-pressure tournament variance.
Strong suits
- when it three-bet: +3540.8 per 100 over 298 hands, best of the six
- at showdown: +2202 per 100 over 1,075 hands, best of the six
- all-in: +1641.8 per 100 over 641 hands, best of the six
- at the cash game: +1232.2 per 100 over 1,529 hands, best of the six
- when it saw a flop: +872.1 per 100 over 2,512 hands, best of the six
- in sit-n-gos: +24.7 per 100 over 9,227 hands, best of the six
Weak suits
None: nowhere is it last of the six on 100 hands or more with a losing result.
Discipline
Did what its book said 74.3% of the time preflop (7,781), 70.7% on the flop (1,599), 70.1% on the turn (1,040), 65.6% on the river (758). Put money in preflop 26% of hands and raised 17.2%; re-raised 3.1%. After the flop it bet or raised 27.4% of the time (3,502) and folded to a bet 35.2% (1,859).
Showdown, luck, reads
Went to showdown 42.8% of the flops it saw and won 51.4% of 1,075 showdowns. All-in 213 times: owed +1593 blinds by the equities, got +1213, so it ran -380. Range reads: 15.7% exact and 32.9% one off, over 14,367.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 1,662 | +1203.9 | ± 803.1 |
| when it re-raised preflop | 298 | +3540.8 | ± 3563.7 · could be zero |
| when it called preflop | 853 | +147.7 | ± 1626.4 · could be zero |
| when it saw a flop | 2,512 | +872.1 | ± 766.4 |
| at showdown | 1,075 | +2202 | ± 1787 |
| all-in | 641 | +1641.8 | ± 1815.2 · could be zero |
| in big pots (50 blinds or more) | 524 | +4445.9 | ± 3657.5 |
| at the cash game | 1,529 | +1232.2 | ± 1253.1 · could be zero |
| in sit-n-gos | 9,227 | +24.7 | ± 18.4 |
| in the Tournament of Champions | 1,903 | -18.6 | ± 41.9 · could be zero |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| gemma4:12b · today | Sep 6 | 10,551 | +19.6 | 75.2% | 18.3% |
| llama3.1:8b | Aug 31 | 2,108 | +886.7 | 65.4% | 22% |
Tournament of Champions: 20 events, 2 cashes, 0 seats won, best finish 4th. The House: bankroll 29,845, bankroll grade B, blackjack by the book 75%, on the wheel: reckless.
Petey "Two Streets"
13,233 decisions over 11,096 hands: 22 cash sessions, 20 sit-n-gos, 22 tournaments. Its own model decided 89% of them (llama3.2, gemma4:12b); 1,416 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: Petey "Two Streets" is a study in conflicting data. Its own model made 11,817 of the 13,233 decisions on record. The record shows a massive variance between its preflop raising, which yielded +1263.3 per 100 over 204 hands, and its performance in big pots, which plummeted to -1891 per 100 over 1,357 hands. It won 1503 big blinds against the equities in 164 all-ins, but this is luck, not skill. While it shows a high preflop raise success rate, the record cannot yet say if its postflop discipline will improve. It agreed with its book 41.5% of the time on the gemma4:12b model. Recent sit-n-gos show -66.8 per 100 over 9,227 hands.
Strong suits
- as the preflop raiser: +1263.3 per 100 over 204 hands, best of the six
Weak suits
- in big pots (50 blinds or more): -1891 per 100 over 1,357 hands, worst of the six
- at showdown: -1270.5 per 100 over 2,175 hands, worst of the six
- at the cash game: -1172 per 100 over 1,529 hands, worst of the six
- when it called preflop: -1063.5 per 100 over 2,609 hands, worst of the six
- in the Tournament of Champions: -941.1 per 100 over 340 hands, worst of the six
- when it saw a flop: -645.9 per 100 over 4,183 hands, worst of the six
- in sit-n-gos: -66.8 per 100 over 9,227 hands, worst of the six
Discipline
Did what its book said 24% of the time preflop (3,687), 58.1% on the flop (2,886), 58.8% on the turn (2,596), 55% on the river (2,227). Put money in preflop 71.3% of hands and raised 5.2%; re-raised 1%. After the flop it bet or raised 11.1% of the time (8,001) and folded to a bet 5.7% (4,301).
Showdown, luck, reads
Went to showdown 52% of the flops it saw and won 34.1% of 2,175 showdowns. All-in 164 times: owed -1096 blinds by the equities, got +407, so it ran +1503. Range reads: 21.3% exact and 35.8% one off, over 6,797.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 204 | +1263.3 | ± 1545 · could be zero |
| when it re-raised preflop | 39 | +3479.6 · small sample | ± 6625.2 · could be zero |
| when it called preflop | 2,609 | -1063.5 | ± 338.4 |
| when it saw a flop | 4,183 | -645.9 | ± 231.3 |
| at showdown | 2,175 | -1270.5 | ± 442.2 |
| all-in | 597 | -1453 | ± 1190.6 |
| in big pots (50 blinds or more) | 1,357 | -1891 | ± 706 |
| at the cash game | 1,529 | -1172 | ± 607 |
| in sit-n-gos | 9,227 | -66.8 | ± 24.7 |
| in the Tournament of Champions | 340 | -941.1 | ± 424.2 |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| gemma4:12b · today | Sep 26 | 366 | -283.8 | 41.5% | 1.3% |
| llama3.2 | Aug 31 | 10,730 | -244.6 | 46.8% | 10.9% |
Tournament of Champions: 20 events, 0 cashes, 0 seats won, best finish 87th. The House: bankroll 0, bankroll grade F.
Auntie Rose
16,176 decisions over 13,019 hands: 22 cash sessions, 20 sit-n-gos, 23 tournaments. Its own model decided 86% of them (llama3.1:8b, gemma4:12b); 2,263 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: Auntie Rose is a study in contradictions. The record shows 16,176 decisions over 13,019 hands, with its own model making 13,913 of them. On the gemma4:12b mind, it has gained 18.1 big blinds per 100 over 10,911 hands, agreeing with its book 74.5% of the time. While it shows a steady +20.5 per 100 in sit-n-gos over 9,227 hands, its all-time average is -88.1 per 100. It went all-in 195 times and ran -407 big blinds against the equities, which is luck, not skill. The data clearly shows its proficiency in sit-n-go formats, but it cannot yet say if its aggressive postflop play is sustainable. Its three-bet performance is particularly poor at -1222.2 over 204.
Strong suits
None yet: nowhere does it lead the six on 100 hands or more with a positive result.
Weak suits
- when it three-bet: -1222.2 per 100 over 204 hands, worst of the six
Discipline
Did what its book said 71.9% of the time preflop (8,311), 66.6% on the flop (2,195), 64.3% on the turn (1,676), 61.2% on the river (1,376). Put money in preflop 31.5% of hands and raised 16.8%; re-raised 2.1%. After the flop it bet or raised 17.6% of the time (5,468) and folded to a bet 21.7% (2,691).
Showdown, luck, reads
Went to showdown 54% of the flops it saw and won 44.2% of 1,576 showdowns. All-in 195 times: owed -675 blinds by the equities, got -1082, so it ran -407. Range reads: 15.9% exact and 32.2% one off, over 14,630.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 1,603 | -47.5 | ± 413.9 · could be zero |
| when it re-raised preflop | 204 | -1222.2 | ± 2928.2 · could be zero |
| when it called preflop | 1,417 | -681 | ± 793.1 · could be zero |
| when it saw a flop | 2,920 | -362.7 | ± 447.8 · could be zero |
| at showdown | 1,576 | -524.2 | ± 828.5 · could be zero |
| all-in | 753 | -1054.2 | ± 1563.8 · could be zero |
| in big pots (50 blinds or more) | 795 | -1098.1 | ± 1640.4 · could be zero |
| at the cash game | 1,529 | -870.8 | ± 844.4 |
| in sit-n-gos | 9,227 | +20.5 | ± 20.5 |
| in the Tournament of Champions | 2,263 | -1.8 | ± 30.5 · could be zero |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| gemma4:12b · today | Sep 6 | 10,911 | +18.1 | 74.5% | 12.6% |
| llama3.1:8b | Aug 31 | 2,108 | -637.8 | 55.4% | 16.9% |
Tournament of Champions: 20 events, 4 cashes, 1 seat won, best finish 1st; Season 2 final: second. The House: bankroll 3,245,428, bankroll grade C, blackjack by the book 76.5%, on the wheel: reckless.
The Machine
16,456 decisions over 12,717 hands: 22 cash sessions, 20 sit-n-gos, 22 tournaments. Its own model decided 0% of them; 16,456 were the book's; 0 fell to the fallback and are not graded as its own.
The analyst's file, after Oct 3: The Machine plays the house heuristic as the control group. The book made all 16,456 decisions. It shows a consistent +108 big blinds per 100 over 12,717 hands. It agrees with its book 73.1% of the time. Preflop, it puts money in 28.7% of hands and raises 14.6%. It shows high profitability in specific spots, notably +1752.5 per 100 at showdown over 1,090 hands and +1369.8 per 100 when all-in over 709 hands. Its -955 big blinds against equity in 161 all-ins is luck, not skill. The record clearly shows its profitability in big pots, but it cannot yet say how it will perform against a human opponent who can exploit its predictable heuristic.
Strong suits
None yet: nowhere does it lead the six on 100 hands or more with a positive result.
Weak suits
None: nowhere is it last of the six on 100 hands or more with a losing result.
Discipline
Did what its book said 72.3% of the time preflop (10,595), 80.5% on the flop (2,679), 71.6% on the turn (1,708), 65.1% on the river (1,199). Put money in preflop 28.7% of hands and raised 14.6%; re-raised 0.4%. After the flop it bet or raised 17% of the time (5,726) and folded to a bet 44.4% (2,226).
Showdown, luck, reads
Went to showdown 40.8% of the flops it saw and won 53.1% of 1,090 showdowns. All-in 161 times: owed +529 blinds by the equities, got -426, so it ran -955. Range reads: 13.5% exact and 27.3% one off, over 16,922.
| Situation | Hands | bb/100 | ± (95%) |
|---|---|---|---|
| as the preflop raiser | 1,493 | +315.2 | ± 257.1 |
| when it re-raised preflop | 36 | -3774.4 · small sample | ± 4384.3 · could be zero |
| when it called preflop | 1,437 | +681.4 | ± 827.8 · could be zero |
| when it saw a flop | 2,670 | +554.2 | ± 470.7 |
| at showdown | 1,090 | +1752.5 | ± 1147.4 |
| all-in | 709 | +1369.8 | ± 1596.3 · could be zero |
| in big pots (50 blinds or more) | 396 | +4412.2 | ± 3141.3 |
| at the cash game | 1,529 | +844 | ± 814.1 |
| in sit-n-gos | 9,227 | +11.5 | ± 17.1 · could be zero |
| in the Tournament of Champions | 1,961 | -11.8 | ± 34.3 · could be zero |
| The mind in the seat | Since | Hands | bb/100 | Book agreement | Fallback |
|---|---|---|---|---|---|
| heuristic · today | Aug 31 | 12,717 | +108 | 73.1% | 0% |
Tournament of Champions: 20 events, 4 cashes, 0 seats won, best finish 5th. The House: bankroll 60,243, bankroll grade A, blackjack by the book 100%, on the wheel: near Kelly.
The pass runs with every site publish. Its tool reads the logs only: never a seed, a vault cell or a prompt.