Pot 0
Pre-flop
Texas Hold'em against six computer players, each with a different and consistent style. You are dealt two cards, five community cards come out in three rounds, and the best five-card hand from your seven wins. The corner panel shows what your hand is actually worth while you decide.
Two players post forced bets before the cards come out — a small blind and a big blind, starting at 10 and 20 — and the obligation moves one seat to the left every hand. Without them nobody would ever have a reason to play a hand, so the blinds are what makes the game move.
They double every 20 hands, up to a ceiling of 160 / 320, which is reached at hand 81 and holds from then on. In a real tournament the levels run on a clock, but a clock means nothing here — "fast opponents" changes how long a hand takes without changing how much poker has been played — so this table counts hands instead. The effect is the same one a tournament is built around: as the blinds grow against your stack, folding and waiting stops being free, and hands you would happily pass early on become hands you have to play.
The ceiling matters as much as the climb. Blinds that kept doubling would pass the 2,000 starting stack within a few more levels, and every hand would turn into a shove-or-fold decision before the flop — a real part of tournament poker, but a much narrower game than the one this table is for.
Your equity is the share of the pot your hand is worth right now. The page measures it by dealing the rest of the board and your opponents' cards at random a few thousand times and counting how often you finish best. It is recomputed every time a card appears, and it counts only the players still in the hand — folding opponents out is worth real equity to you.
Pot odds are the price you are being offered. Calling 50 into a pot of 150 risks 50 to win 200, so you need to be right 50 ÷ 200 = 25% of the time just to break even. Compare that with your equity: if equity is higher, calling makes money over the long run even when this particular hand loses.
The verdict line does exactly that comparison, and nothing more. It weighs equity against the immediate price, so it ignores implied odds — the chips you might win on later streets when your draw comes in — and it ignores the chance of winning by betting and making everyone fold. Follow it mechanically and you will play a little too tight with drawing hands.
Outs are the cards still to come that would genuinely improve you. A card that merely pairs the board is not an out, because it pairs it for everyone. Nine outs — a flush draw — comes in about 35% of the time between the flop and the river. The shortcut at the table is to multiply your outs by four on the flop and by two on the turn.
Each unseen card is roughly a 1 in 47 shot, or about 2.1%, so one out is worth about two points per card still to come. That gives the shortcut every player learns: multiply your outs by 4 on the flop (two cards to come) and by 2 on the turn (one card to come). It is close enough to bet on:
| Outs | Flop → river | ×4 rule | Turn → river | ×2 rule |
|---|---|---|---|---|
| 4 gutshot | 16.5% | 16% | 8.7% | 8% |
| 8 open-ender | 31.5% | 32% | 17.4% | 16% |
| 9 flush draw | 35.0% | 36% | 19.6% | 18% |
| 12 | 45.0% | 48% | 26.1% | 24% |
| 15 flush + straight | 54.1% | 60% | 32.6% | 30% |
The ×4 rule drifts high past about ten outs, because it counts the run-outs where you hit on both cards twice over. Above that, multiply by 4 and subtract the outs above 8.
Your final hand is the best five of seven cards. These frequencies are exact — the page's own evaluator was run over all 133,784,560 seven-card combinations to produce them:
| Hand | Chance | Odds |
|---|---|---|
| One pair | 43.82% | 1 in 2.3 |
| Two pair | 23.50% | 1 in 4.3 |
| High card | 17.41% | 1 in 5.7 |
| Three of a kind | 4.83% | 1 in 20.7 |
| Straight | 4.62% | 1 in 21.6 |
| Flush | 3.03% | 1 in 33.1 |
| Full house | 2.60% | 1 in 38.5 |
| Four of a kind | 0.168% | 1 in 595 |
| Straight flush | 0.031% | 1 in 3,217 |
Two pair arrives once every four hands or so, which is why a single pair wins far fewer showdowns than beginners expect — and why the strength of your kicker decides so many pots.
There are 1,326 possible two-card holdings, which reduce to 169 distinct types once suits that do not matter are folded together.
| Holding | Combinations | Chance |
|---|---|---|
| Any pocket pair | 78 | 5.88% — 1 in 17 |
| One specific pair (aces, say) | 6 | 0.45% — 1 in 221 |
| Suited | 312 | 23.53% |
| Suited connector | 48 | 3.62% |
| Ace-king, any form | 16 | 1.21% |
With the odds panel showing, every hand is turned face up once the hand is over — including the players who folded, and including the hands nobody stayed to see. That is not how a real table works, and it would be information you could not have in a real game, but it is most of the lesson here: it is the only way to find out whether the player who pushed you off a pot actually had it.
Turn the panel off and the table behaves properly again — only the players who paid to reach a showdown reveal, and a pot won by everyone else folding is never shown.
Counting every possibility is not an option. Heads-up on the flop, an exhaustive answer means checking every opponent holding against every run-out — 893,970 deals for one opponent, and astronomically more against six.
So the page samples instead. It deals the rest of the board and everyone's hole cards at random a few thousand times and counts how often you finish best. That is a Monte Carlo estimate, and its error falls with the square root of the sample count:
| Samples | Typical error |
|---|---|
| 500 | ± 2.2 points |
| 2,000 | ± 1.1 points |
| 6,000 what the panel runs | ± 0.65 points |
| 20,000 | ± 0.35 points |
The samples are spread across animation frames rather than run in one burst, so the table stays responsive — which is why you can watch the figure settle as the sample count climbs. The computer players use the same sampler on a smaller budget, and how many samples each one is allowed is a fair description of how well it reads the board.
Before the flop there is no board to measure against, so the panel shows Bill Chen's starting-hand score instead: −1 for the worst hand (7-2 offsuit) up to 20 for a pair of aces. Roughly the top 18% of hands score 7 or better, and the top 26% score 6 or better. The computer players use the same number to decide what to play, so their advertised styles are literally true.