# Dwan–Schwimer: public methodology

Research/extraction date: September 5, 2026. Publisher: GTO Solutions AS.

This is a reported played-hand reconstruction and scoped comparison with existing database solutions. It is not a custom solve or a complete solver grade.

## Hand sources

- PokerNews, February 4, 2021: https://www.pokernews.com/news/2021/02/high-stakes-poker-season-8-episode-8-38623.htm — positions, stack, board, betting sequence and final pot.
- Poker.org, February 7, 2021: https://www.poker.org/latest-news/michael-schwimer-leaves-tom-dwan-stunned-after-200k-high-stakes-poker-pot-a7Cd65C72XYK/ — physical hole cards, river pot and call amount.
- Bart Hanson reconstruction: https://www.youtube.com/watch?v=euMG7R2k4Go — independent reconstruction, including a conflicting CO position label.
- Primary broadcast page: https://www.pokergo.com/videos/high-stakes-poker-season-8-episode-8 — the subscription-gated original footage was not independently reviewed.

The blind/ante description at episode level does not reconcile with the reported hand pots. We use the reported river starting pot54,800 and final pot200,800, and derive flop pot26,800 by subtracting the28,000 flop bet/call. We do not infer an exact forced-bet schedule. The main comparison follows PokerNews's HJ/SB labels; CO is separately tested.

## Database scope

The preflop figures use actual bundled solver data from a separate preflop library. Postflop figures use actual GTO Gecko stored solutions. They are separate extracts of corresponding configurations; no exact continuous preflop-to-river tree is asserted.

Main configuration: eight-seat straddle cash,200 base big blinds,0.5/1/2 blinds,0.5 base bb ante per player (4bb total),4% rake capped at2 base bb,HJ open6bb,SB raise24bb,heads-up postflop. The word bb always refers to the base blind before the straddle. Model200bb means100 straddles.

Main postflop: SB versus HJ three-bet pot,5–4–2 rainbow,55bb pot,176bb behind. On the flop select bet27.5bb/call, then the ace turn. On the turn select SB check to inspect the IP response. The export ends at river-entry states; we did not retrieve or calculate a river strategy. Standard equilibrium files were used, without player-type adjustments or node locking.

## Physical-card mapping

Original suits map c→s,s→h,h→d,d→c, consistently for every card and combination. Actual flop4s5c2h becomes sorted5s4h2d; turnAh becomesAd; riverJs becomesJh; actual heroAdKh becomesAcKd; actual villain5h4c becomes5d4s. The actual villain combination is absent. Analogy5d4d becomes5c4c; analogy5h4h becomes5d4d. Replacing only a hand's suits is not a global relabel and is explicitly treated as a different-hand analogy.

## Frequencies and graphics

Postflop probabilities are stored rounded to two decimal places, including totals of0.99 or1.01. Normalize each legal positive-frequency vector by its sum for summarized comparisons. Selected exact sum1.00 vectors can be displayed directly. Thus the suited54 flop call is approximately always; the turn medium bet for mapped5c4c is98/101=97.03%, not exactly98%. Rounded near-pure outputs are not claimed as theoretical certainty. No exploitability or convergence certificate was available in the retrieved export; no precise EV losses are asserted.

Preflop aggregate weights are six combinations per pair class, four per suited class, twelve per offsuit class, multiplied by stored arrival weight. Normalize each action vector, sum weighted action mass and divide by total arriving mass. Zero arrival means no conditional verdict. The grid uses one category per hand: mixed when two or more actions each reach1%; otherwise the dominant action. Grey/dash denotes zero arrival. A nonzero arrival can still be tiny: the grid does not encode how often each hand first opened. The native article includes a text-equivalent13×13 table.

All graphics are original SVGs created for this article. The seven graphics explain the reported showdown, street actions, SPR comparison, categorical range grid, flop mix, turn analogy and river price. Source data for the grid and selected numeric results are published alongside this note. Full proprietary solution nodes, credentials and unrelated account data are not published.

## Sensitivity and verification

- HJ200bb main flop branches: 13.8/55=25.09%,27.5/55=50%,41.2/55=74.91%. On the ace turn after each bet/call, mappedAK checks30%,77%,96% respectively. These are different conditional ranges, not interpolation points.
- CO200bb half-pot branch: mappedAK checks90% on the ace turn. CO54o still folds100% first in; SB AKo three-bets96.14% versus its open.
- Other checked root geometries:100bb straddle gives76/55=1.38 SPR;150bb unstraddled live gives138.8/23.5=5.91. Their ranges and rake assumptions differ and they were not used for the main strategy claim.
- Main model SPR176/55=3.20; reconstructed actual SPR87,000/26,800=3.2463.
- River call threshold48,000/(54,800+25,000+73,000+48,000)=23.9044%. This is required pot share, not measured opponent bluff frequency. No ties means required win probability; reported pots are assumed to exclude no further unreported deductions.

A separately written verifier checked suit mapping, storage-to-bb scaling, action branch, reach consistency, preflop weights and river arithmetic. Working extraction scripts and scoped source hashes are retained privately for editorial reproducibility. The article also undergoes source/claim review, visual inspection, responsive checks and the site's build pipeline.