Spin & Go Push/Fold Charts
Three-handed winner-take-all, no ante, equal stacks. Nash equilibrium
shove and call ranges, solved from scratch and checked against published charts.
Every number below is reproducible — see the method note at the bottom.
What was solved
A pure shove-or-fold game: every player either moves all in or folds, no raising and
no limping. Three players, equal stacks, blinds of 0.5 and 1, no ante. Action order is
button → small blind → big blind. The heads-up charts are the same game with
one player busted.
Why plain chip EV is the right math here
In a winner-take-all tournament a player's prize equity equals their share of the chips
in play. That makes equity linear in chips, so maximising chips is exactly maximising money
and no ICM adjustment is needed. This is only true because the high multipliers that pay
second and third were deliberately excluded — those would need different, tighter ranges.
How it was built
- An exact 169×169 hand-class all-in equity matrix, built with a C-backed evaluator
at 50,000 trials per matchup.
- Blocker-aware range math: each villain hand class is weighted by how many of its
combinations survive the cards you hold. The weight table sums to exactly C(50,2)=1225
per row, which is the arithmetic check that it is right.
- Equilibrium by damped best-response iteration, with three-way all-in equities
re-sampled during the solve for the spot where the button shoves and the small blind calls.
How it was verified
- Equity layer — spot-checked against full board enumeration. AA vs AKs: exact
0.8786 vs matrix 0.8792. 72o vs AA: exact 0.1180 vs 0.1173. AKs vs 22: exact 0.4989 vs
0.4966. All inside 0.25%.
- Equilibrium quality — measured exploitability, the most any single hand could
gain by deviating. Lower is better; 0 is a perfect equilibrium.
- External cross-check — the heads-up 10bb solve gives 58.2% shove / 37.3% call,
against published values of 58.7/37.6, 58.9/38.3 and ~58/~38 from three independent sources.
Note for anyone comparing elsewhere: some sources quote roughly 53% for this spot.
Those are raise-or-shove solves where the small blind min-raises most hands and only shoves
the rest. Different game, different number — not a contradiction.
- Three-handed cross-check — read this before comparing elsewhere. The
heads-up numbers match published charts closely. The three-handed button shove does not:
this solve says 33% at 10bb, while published figures range from 39% to 52%. Those sources
disagree with each other, and one of them lists the small blind as shoving tighter than the
button, which cannot be right with fewer players left to act. Where a source gives an actual
hand list rather than a headline percentage, it agrees with this solve — one widely used
guide's big-blind calling list works out to about 22%, against 24.3% here. The most credible
disagreeing table is a six-max chart that may assume a big-blind ante, which would widen every
range and would explain a consistent gap of roughly six points. That is a hypothesis and has
not been tested yet. Treat the three-handed button percentage as the one number here still
open to revision.
- Internal consistency — when the button folds, three-handed play becomes heads-up.
The three-handed solver reproduces the separately written heads-up solver in that subgame
(58.2% vs 58.2%, 37.4% vs 37.3% at 10bb).
Known simplifications
- Equal stacks. Real spins drift unequal; these are the standard memorisable charts.
- At 15bb a pure shove-or-fold strategy leaves value on the table versus min-raising.
It is a deliberate simplification for memorability, not the full 15bb strategy.
- A folding player's cards are not removed from the opponents' ranges. Standard practice;
the effect is small.