HoldemMath Poker Math Data

Precomputed Texas Hold'em datasets powering the free tools at holdemmath.com — no account, no paywall, everything runs in your browser.

Dataset	What it contains	Files
Preflop equity matrix	All 14,365 unique starting-hand matchups (169×169 grid), 20,000 Monte Carlo runouts per pair	matrix/
Nash push-fold charts	Heads-up / 6-max / 9-max equilibrium shove & call ranges at 7 stack depths (4–20bb), no-ante and 10%-ante variants	pushfold/

Interactive versions: odds calculator · matchup matrix · push-fold charts · push-fold trainer

A zero-dependency in-browser viewer for the charts ships in viewer/ — enable GitHub Pages (or serve the folder with any static file server) to browse all ranges without writing code.

---

1. Preflop equity matrix

matrix/preflop_equity_matrix.json

{
  "n": 20000,          // Monte Carlo runouts per pair
  "order": ["AA", "AKs", "AQs", "..."],  // 169 hand classes, same order everywhere
  "p": [               // 14,365 rows, one per unique unordered pair {i, j}
    [0, 0, 21, 956, 22],   // [i, j, i_wins_permille, tie_permille, j_wins_permille]
    ...
  ]
}

• Hand notation: AKs = ace-king suited, AKo = ace-king offsuit, TT = pocket tens. order[i] gives the display name of class index i.
• Probabilities are integer permille (out of 1000). Example row [0, 0, 21, 956, 22]: AA vs AA wins 2.1% / chops 95.6% / loses 2.2% — the famous near-guaranteed split.
• All-in equity (pot equity) of i vs j = (i_wins + tie/2) / 1000.

matrix/preflop_equity_matrix.csv

169×169 wide table, cell = equity of row hand vs column hand in percent (win + ½·tie, 2 decimals).

matrix/preflop_matchups_long.csv

Long format, one row per unique matchup: hand1, hand2, win_pct, tie_pct, lose_pct, equity_pct.

---

2. Nash push-fold charts

pushfold/pushfold_charts.json

{
  "v": 1,
  "depths": [4, 6, 8, 10, 12, 15, 20],        // effective stack in big blinds
  "order": ["AA", "AKs", "..."],              // same 169-hand order as the matrix
  "formats": { "hu": {"seats": [...]}, "max6": {...}, "max9": {...} },
  "anteVariants": { "data": 0, "dataAnte10": 0.1 },  // ante per player, in bb
  "data":      { "<fmt>": { "<depth>": { "<seat>": PfRange } } },
  "dataAnte10":{ "<fmt>": { "<depth>": { "<seat>": PfRange } } }
}

A PfRange entry is:

{
  "s": [0, 1, 5, ...],  // hand-class indices (into "order") in this range
  "p": 12.3,            // coverage: % of all combos this range covers
  "e": [3.02, ...]      // EV of the strategy per hand class, aligned with "order"
}

Seat keys (<fmt> = hu / max6 / max9, <depth> = one of the depths above):

• "UTG", "HJ", "CO", "BTN", "SB", ... — that seat's first-in all-in (shove) range.
• "BB|UTG", "CO|HJ", "BB|SB", ... — the calling range of the seat left of | facing a shove by the seat right of |. E.g. "BB|SB" = big blind call vs small blind, "BB|UTG" = big blind call vs UTG shove.
• Heads-up has exactly two entries per depth: "SB" (SB shove) and "BB|SB" (BB call).

EV convention: chip EV in big blinds, relative to folding (fold = 0), aligned with order. For shove ranges e[h] is the EV of jamming hand h against the field's calling strategy; for call ranges it is the EV of calling.

Model, honestly stated

• Heads-up is a near-exact Nash equilibrium (fictitious play, best-response deviation ε ≤ 0.05bb; ε ≤ 0.13bb in the ante variant).
• Multiway tables solve each first-in seat's sub-game independently with a single-caller approximation, and fix the SB first-in range to the heads-up solution so tables stay mutually consistent.
• Chip EV, no rake, no limps, no ICM. The 10% ante follows the common (HRC-style) convention: every player posts 0.1bb before the cards are dealt.
• Multi-equilibrium degeneracy in push-fold games means other sites' charts may differ slightly and still be equilibria — compare EVs, not exact hand lists.

---

3. Quick start

Python and JavaScript examples live in examples/:

from examples.equity_lookup import equity
equity("AA", "KK")   # 81.95  (% for AA all-in preflop vs KK)

const { equity } = await import('./examples/equity_lookup.js');
await equity('AA', 'KK');   // 81.95

Load the push-fold charts and read the HU 10bb SB range:

from examples.pushfold_ranges import pushfold
pf = pushfold()
sb = pf["data"]["hu"]["10"]["SB"]
hands = [pf["order"][i] for i in sb["s"]]   # e.g. ['AA', 'AKs', ...]

4. Viewer (no code needed)

viewer/index.html renders every chart as a 13×13 grid (dropdowns: format, depth, ante, seat). Serve the repo root with any static server, e.g. python3 -m http.server 8000, then open http://localhost:8000/viewer/. On GitHub it works out of the box once Pages is enabled (Settings → Pages → Deploy from branch → main → /root).

5. Attribution & license

• Data (matrix/, pushfold/): CC BY 4.0 — free to use and redistribute, please credit "HoldemMath (holdemmath.com)" and link back if you build on it.
• Code / examples / viewer: MIT.

Generated by HoldemMath's own tooling (Monte Carlo via the MIT-licensed eval7 engine; charts solved by HoldemMath's fictitious-play solver).