Texas Hold'em stats library
5 deterministic datasets, each on its own page with the full table, CSV download and BibTeX/APA citations. Every figure is reproducible (exhaustive matrix / combinatorics / self-solved Nash).
- All 169 starting hands vs random — exact equity ranking — 169 rows · CSV
- Flop texture frequencies (exact combinatorics) — 6 rows · CSV
- Outs odds master table (Rule of 2&4 vs exact, 1-20 outs) — 20 rows · CSV
- Opening and jam/call range shares by position and stack — 25 rows · CSV
- All 169 starting hands: equity vs a random field at 2, 3 and 6 players — 169 rows · CSV
Related: Odds dataset (hands & outs)
How these statistics are produced
The distributions on this page come from enumeration, not sampling: C(52,2)=1,326 starting combinations (78 pairs, 312 suited, 936 offsuit), and matchup equities are produced by enumerating opponent combos card by card — so "AA has 85.20% versus a random hand" is a recomputable exact value, not a simulation estimate. With 205 terms and 469 matchup pages, this page only aggregates; it introduces no new definition.
What to keep in mind when reading them
Enumeration gives a static showdown equity: it contains no betting, folding or position, so it is not an expectation and cannot be converted directly into a call-or-fold decision. Turning it into a decision needs pot odds (compare price with equity), fold equity (what a fold is worth) and position. This page supplies the verifiable base numbers; the strategy section explains how to combine them.
What each of these 5 datasets settles
· All 169 starting hands vs random — exact equity ranking — 169 rows · exhaustive class-average, card-removal weighted (weights sum 1225 per hand) · Flop texture frequencies (exact combinatorics) — 6 rows · exact counting · Outs odds master table (Rule of 2&4 vs exact, 1-20 outs) — 20 rows · deterministic formula · Opening and jam/call range shares by position and stack — 25 rows · combinatorial share of 1,326 combos · All 169 starting hands: equity vs a random field at 2, 3 and 6 players — 169 rows · seeded Monte Carlo (300,000 样本/组, seed 20260716); N=2 与穷举 vs-random 对账最大偏差 0.202pp
All 5 datasets share one property: a third party can recompute the conclusion. The hand ranking comes from pairwise enumeration over 1,326 combos, flop textures from categorising C(50,3)=19,600 flops, and draw probabilities from ratios of combination counts rather than simulation. Anyone holding the CSV can verify it with a table instead of taking it on trust. That also dictates how to cite them: cite the dataset and its scope, not a single sentence.