QuintAI Graded My Dumbest Hands. It's Not What You Think.

Everyone knows the 53s punt. Triton Jeju. 5-bet shove with trash against a guy who had ace-king. I lost ~$750K in thirty seconds and then, because once wasn't enough, I got my aces cracked by nines back-to-back for the rest of the stack. That's the clip. Probably has the double-middle-finger emoji next to it in your group chat.

I'm not going to defend the 53s. Bad hand. Part of the brand. BrokeLivingJRB didn't come from playing aces correctly.

But for years I've wondered: is the 53s actually the leak? Is it the hand that explains why my televised graph looks the way it does? I had a feeling it wasn't. It's the story people share. But a preflop spew happens once and the variance on it is bounded — I'm not bleeding $2M from 53-of-hearts plays. So where's the money going?

I handed thirteen of my most-discussed TV hands to QuintAce — the AI that grades decisions against a GTO solver. Every big one. Triton, HSP, HCL, NGNF, Venetian Live. I told them to run each decision through the solver and tell me what the actual pattern is.

They did. The answer is not what my Instagram comments say.

The 13 hands

Here's the catalog. One preliminary note: the original list was 14. One was cut — the "AA cracked by 99" at Triton right after the 53s has no public print record beyond a sentence, no way to solver-check. So: 13.

Look at that table carefully. The story the clips told isn't here.

What the catalog actually says

I folded the best hand twice — confirmed by the solver. Two hands, not four. I called into the nuts once with top-two, but that call was actually GTO-correct. The solver folds QJ zero percent of the time at that node. I just ran into the one hand I couldn't beat.

And Hand 7 — the "Robl flush-rep" that everyone remembers as me folding like a mark? I called the bluff correctly. Robl told me he had a flush after I'd already called, I mucked face-down, he revealed the joke and pushed the pot back to me. The call was right. The muck was a verbal hustle.

QuintAI annotation — the pattern, restated.

Of the 13 graded: two are clear fold-best-hand mistakes against solver-preferred calls (Hands 2 and 3). One more (Hand 7) was a correct call that got verbally unwound after the fact. One (Hand 10, Adelstein) was graded as a solver-correct call that happened to run into the nuts — not a misread, a cooler. The other nine are variance, entertainment-style gambling pots, or reads that worked.

The failure mode isn't misread opponents. It's conviction under pressure: commit threshold when the bet is big enough to be scary, and execution resilience when the opponent starts talking.

The two real fold mistakes

Three decisions. Same solver verdict in every one: call is better than fold. On Hand 2 and Hand 3, I folded. On Hand 10, I called — and the solver says that call was the right one, even though I lost to a cooler.

Hand 2 — QQ fold vs Salomon's 4-bet all-in. HSP S11 Ep 4. Salomon opens to $15K with 4-2o. I 3-bet to $50K from the SB with queens. Salomon shoves over the top — he has me covered at $365K. I fold.

Nick Schulman on commentary called it the worst fold in HSP history. He wasn't grandstanding.

QuintAI annotation — QQ preflop fold.

At 180bb effective, against a 4-bet shove over a SB 3-bet, the solver calls 100% of the time with QQ across all suit combinations. Average call EV: +137bb. Average fold EV: -27bb (the sunk 3-bet). EV delta of the fold: -164bb ≈ -$328,000.

The solver's opponent range is inferred GTO — a mix of JJ+, AK, and a sprinkle of bluffs. Against Salomon's actual 4-2o, the call wins with ~82% equity instead of ~55%. Either way, folding is the single most expensive preflop decision in the catalog.

Hand 3 — AKo fold vs Persson's river check-raise. HSP S10 E12. $5K preflop raise from Robl, I call with AK on the button, Persson calls from the BB with A9 suited. Flop 5♦6♣A♣ — I flop top pair top kicker. I bet $12K, Persson calls. Turn T♦. I bet $30K, Persson calls. River 8♦ — third diamond arrives. Persson checks. I bet $65K. Persson check-raises to $265K.

Pot is $495K. I have to call $200K. I'd need 29% equity to break even. I fold.

I said on the day I wasn't rich enough to call. The solver doesn't care what I felt I was.

QuintAI annotation — AK river fold.

At this exact river node, the solver's BTN range calls 73% of the time and folds only 27%. Average call EV across the range: +665bb. Average fold EV: -282bb (sunk investment). That's a +947bb gap between calling and folding — roughly $380K in range-level EV.

Caveat: AKo itself has near-zero reach at this node in the solver's own equilibrium, because the solver would have 3-bet preflop with AK rather than called. So we can't cite a clean AKo-specific EV. What we can say: (a) the BTN range overall calls three-quarters of the time at this sizing, which means the check-raise does not threaten enough value to justify folding the top of the one-pair class; and (b) AK has more than 29% equity against almost any plausible check-raise range containing bluffs.

Folding TPTK for a 29% equity requirement on a board that blocks Persson's nut-flush combos is a measurable leak.

Call it $700K in misjudged equity across two hands — Hand 2 and Hand 3. Not $2M across four. The narrative got inflated because folds look worse than they are. Only two of my famous ones are actually GTO-wrong.

The calls that weren't wrong (even when they lost)

Hand 10 — QJ top-two vs Adelstein K-T straight. HSP S9 E7. Four-way limped-raised pot. Flop J♣7♣9♦ — I have top pair top kicker on the button. Everyone gets there. Turn Q♠ — now I have top two. Adelstein bets $20K on turn, I call. River 3♥ — a brick. Adelstein overbets $125K into $85K. I call. Adelstein shows K♠T♠ — he made the straight on the turn.

Every commentator told me I should have folded. The solver disagrees. Here's the full range map at my decision node, with QJo marked:

Most of the BTN range folds to that overbet — 65% of it. But not top-two. For QJo specifically, the solver folds 0% of the time, calls 81%, and jams 19%.

QuintAI annotation — QJ top-two river call.

Solver call EV for QJo at this node: +154bb ≈ +$154,000. Fold EV: -34bb. Jam EV: +142bb — better than folding. Folding is the single worst option.

I lost $125K on this hand. The solver would have lost the same amount by calling. This is not a leak; it's variance on a correct decision. Adelstein was at the exact top of the overbet range — he had the one combo that beats me. The grid above shows what the range at this node actually looks like: strong one-pair hands overwhelmingly call; KT specifically (the nuts) jams 100%.

I've been told for three years I should have folded top-two there. The math doesn't agree. It's variance, not a read miss.

Hand 7 — Robl's flush-rep bluff. HSP S11 E5. 9♦2♦J♠Q♠5♦ runout. Robl overbets $150K on the river with A-K high, representing the third-diamond flush. I have Q-T, top pair. Against a polarized river overbet on a board where the flush is obvious and his line is thin for value, top pair beats enough of his bluffs to justify the call.

I made the call. Pre-solver, any commentator would tell you this is a tough call to find.

Then — after I'd already put the money in — Robl said, out loud, "flush." I mucked my cards face-down. Robl revealed he was joking, showed his A-K, and literally pushed the pot back to me.

QuintAI annotation — call-then-muck-on-verbal-rep.

The read and the call were correct. The execution failure happened after the money was in the pot, at a moment that shouldn't have been a decision point. Solver can't grade "believing an opponent's verbal claim post-showdown" — but the underlying poker decision (making the bluff-catch in the first place) is read-right.

Cataloged as a read hit. Note the failure mode: the same player who correctly read Robl's line couldn't hold conviction against Robl's voice. That pattern shows up in the fold hands too.

The same muscle that folded QQ and folded AK is what let me muck Q-T after correctly calling. Correct decision → verbal pressure → muck. In Hands 2 and 3, correct read → bet pressure → fold. Same failure mode, different surface.

And the 53s?

The 53s punt is a story, not a pattern. I was in a session, I was gambling, I had a 4-bet in front of me from a guy who was never folding AKo at that price. I should have folded the 3-bet back. I shoved. It's a 38/62 underdog when called. I'm not re-litigating it.

But here's the math thing: preflop spews are bounded by pot size. The 53s is an expensive one-off. A variance-loaded play with 38% equity gets me $750K underwater once, and the distribution evens out over a career — I don't make that play every week.

River mistakes against planning opponents are unbounded. If you fold QQ once, that's -$328K on one decision. If you develop the habit of folding QQ at 180bb whenever a pro 4-bets you, that's -$328K every month across your career. The 53s happens; the QQ fold compounds.

The 53s is the hand the clip reel was made for. The QQ fold is the hand that actually pays my rent on the BrokeLiving brand.

What the grader recommends

For me: mostly nothing. The reads work. They're why I'm invited to these games. Switching to a pure GTO approach costs my edge against the recreational end of the pool, which is where my money comes from. But the specific advice the solver would give me — and it's concrete — is: when the read is clear and the math is on your side, stop letting the opponent's conviction change yours. Persson's check-raise doesn't become a nut flush because he's staring at me. Salomon's 4-bet shove doesn't become aces because he's said three sentences about Malibu.

If I fold less when the math says call at 70%+ equity, the rest of my game barely changes. That's the whole fix.

For a recreational player watching me: don't imitate the 53s punt. You already knew that. But also — don't imitate the fold. The instinct to lay down a big hand when a pro confidently shoves or check-raises is a feature they are selling you. The honest play against a balanced line is the math play. Your gut is probably right about direction. The question is whether you can hold it when they push.

For QuintAI: this is exactly what a solver is for. When conviction breaks, math is the cheapest substitute. It doesn't care what the opponent said or how they stared. It just gives you the frequency mix. The solver called QQ 100%, QJ top-two 81%, and AKo 73% at the range level — three decisions where my gut said fold. On two, my gut was wrong. The solver has the receipts.

Close

I didn't lose $2M folding best hands across four hands. I lost about $700K folding best hands across two (Hands 2 and 3), plus a pot I'd already won that Robl stole via verbal sleight (Hand 7). The much-cited Adelstein call (Hand 10) wasn't even a leak — it was a solver-correct call that hit a cooler.

Three decisions out of thirteen. That's the real BrokeLiving pattern. Not a reads-player in a math game. A reads-player whose reads are fine, but whose ability to commit under pressure isn't always up to the reads.

I'm still BrokeLivingJRB. Just — now with the math on my side next time.

QuintAI doesn't care about any of that. It just grades. Here's the grade.

⚠️ Review note for JRB and Thanh — remove before publish.

Solver-verified: - Hand 2 (QQ fold): -164bb EV confirmed; 100% of solver QQ combos call. - Hand 3 (AK fold): range-aggregate 73% call, AKo-specific degenerate (off-tree at 700bb stack depth). Claim framed around range behavior + pot-odds math, not AKo-specific EV. - Hand 10 (QJ call): solver QJ folds 0%, calls 81%, jams 19%. Average call EV +154bb. Reclassified from "read miss" to "solver-correct cooler."

Commentary-level (not solver-verified): - Hand 7 (Robl flush-rep): PokerNews S11 E5 confirms JRB called, then mucked on verbal claim, then Robl returned pot.

Hands not solver-verified: 1, 4, 5, 6, 8, 9, 11, 12, 13. Classifications in the catalog table are based on outcome + commentator verdict, not per-hand solver EVs.

Numbers quoted: - "-$328K" for Hand 2: 164bb × $2K/BB = $328K. Clean. - "~$380K in range-level EV" for Hand 3: (665 − (−282)) bb × $400/BB = $378.8K. Framed as range-level, not AKo-specific. - "$700K" aggregate: $328K + $378K ≈ $706K. Rounded. - "~82% equity" for QQ vs 42o: textbook heads-up equity ~82% (QQ vs 42 offsuit).

If any of these numbers move after deeper solver runs or a stack-depth clarification, v3 adjusts.

Methodology and caveats

This section is deliberately verbose for review. Trimmed before publish.

Data source

The 13 hands were drawn from publicly televised poker (HSP / PokerGO, HCL, No Gamble No Future, Triton, Venetian Poker Live). Each hand was reconstructed from at least one primary source (PokerNews recaps, Upswing Poker breakdowns, Gutshot Magazine, poker.org, HighStakesDB, tightpoker, VIP-Grinders). Full per-hand provenance: hands.md in this article folder, which uses its own H-prefix numbering (H01–H14; H02 cut, H13 replaced). The article body uses a clean 1-through-13 ordering corresponding to the table column #.

Verification went through three rounds. Round 1 (commentator summaries) identified surface-level catalog. Round 2 (web research) corrected several hands: Hand 5 was HSP S8 E5, not a Triton hand; Hand 8 is JRB's AJ flopping trips and rivering quads vs Persson 73s, not an AA-vs-AA quads hand; Hand 12 replaced the originally-listed "MDG $893K vs Matt" (no print coverage of hole cards) with the $655K scoop later in the same HSP S11 E4 session as Hand 2 (narrative arc: worst fold early, big scoop late). Round 3 dug further on remaining partials and confirmed one cut (the original AA-vs-99 Triton hand, now absent from the catalog).

Solver grading

Per-hand decision analysis was run through the QuintAce strategy-grid solver at each JRB-decision node where the reconstruction was clean enough. Hands 2, 3, 10 are solver-verified in the article body. The solver returns the full 169-hand (or 1326-combo) action mix + EVs at the decision node. For each target hand we extracted both the hero-specific strategy (aggregated across hole-card suit combinations) and the range-level call/fold frequency.

For Hand 2 the solver returned fold 0% on QQ (clean, strong signal). For Hand 3 the solver returned near-zero reach on AKo because the solver's own BTN range wouldn't reach this river node with AKo — it would have 3-bet preflop. For Hand 3 we used range-level behavior as the anchor claim and pot-odds math for the equity argument. For Hand 10 the solver returned a clean QJo strategy (0% fold / 81% call / 19% jam).

Verdict buckets

• ✗ Read miss — JRB's action disagrees with solver; measurably loses EV vs solver-preferred.
• ✓ Read hit — JRB's action agrees with solver, or was correct by pot-odds / equity.
• — Variance — action was solver-correct; outcome determined by luck.
• 🎰 Variance hit — solver-correct action that happened to hit big.
• ⚡ Preflop action leak — preflop gambling action with clearly negative EV (Hand 1, the 53s shove).

What we didn't do

• Run the solver on every hand. Only Hands 2, 3, 10 are solver-verified in-body. Nine hands (1, 4, 5, 6, 8, 9, 11, 12, 13) are classified on outcome + commentary + pot-odds, not per-hand solver EV.
• Verify every betting-line detail from raw video. A few hands retain gaps (Hand 1 stack depth; Hand 2 Salomon's exact 4-bet size; Hand 3 river card is confirmed but re-straddle structure approximated; Hand 4 river card and bet size; Hand 7 preflop and early-street action). All flagged per hand in hands.md.
• Include private cash-game hands or unstreamed tournament hands.
• Claim definitive EV numbers for hands where the solver's own equilibrium doesn't reach the node with the hero's actual holding (Hand 3).

Known caveats

• Hand 3 AKo-specific EV: range-aggregate call rate 73% is solid; AKo-specific EV is degenerate because the solver wouldn't have played AKo here. Claim is framed as range-level, not hero-specific.
• Hand 2 Salomon's 4-bet size: modeled as full-stack shove (180bb effective). Salomon's exact shove size is not in any public print source. The call-EV advantage is robust to 4-bet sizing because QQ dominates Salomon's plausible value range at any sizing.
• Hand 3 preflop approximation: HSP S10 E12 was 7-handed with a mandatory straddle + Tilly's re-straddle. Solver run approximated to 6-handed 2-blind + BBA for tractability. Postflop SPR matches; the preflop tree simplifies.
• Hand 10 preflop approximation: actual hand was 4-way preflop (limp-raise-2 calls). Solver run simplified to CO vs BTN heads-up. The river decision's SPR and Adelstein's overbet sizing are preserved.
• "$700K aggregate misjudged equity": sum of Hands 2 + 3, range-level for 3. Replaces the earlier "$2M across four hands" claim from the first v1 draft.

Hands not run through the solver

Hands 1, 4, 5, 6, 8, 9, 11, 12, 13 are classified on commentary + outcome. A v3 pass could run per-hand solver queries for these but they don't anchor the article's thesis.

Research and grading: QuintAce. Narration: Jean-Robert Bellande.