crtdrops
NOTE · ENGLISH ·

The rule of 4 lies, and not where you think

From 1 to 20 outs on the flop, the rule of 4 as dashes with squares and the exact probability of hitting on the turn or the river as a solid line with circles: they run together up to 8 outs and then the rule pulls away, 80 % against 67.5 % with 20. On the right, the 9-out example with a pot-sized bet: you need 33.3 %, the rule says 36 %, the exact two-card figure is 35.0 % and with one card you buy 19.1 %.

I have always looked at poker through math: it is about making rational decisions with incomplete information, and a good decision can lose a hand and still be a good one. I have played it mostly at home with friends and in an online room with play chips.

At any table, the same count comes up to decide whether to keep going with a draw: your outs times 4 on the flop, times 2 on the turn. I wanted to check with numbers how accurate it is, and it turned out its big error is not in the rounding.

This is the math of the game, not an invitation to gamble: everything is in chips. The figures in the examples are illustrative: the pot and the bets are those of a typical situation, not of a hand that was played; the probabilities are exact for hitting, assuming that hitting wins.

Where the rule comes from

An out is a card that completes your hand. With a flush draw, the 9 left of your suit; with an open-ended straight draw, 8. In Texas Hold’em, after the flop you see 5 cards: your 2 and the 3 on the board. You do not know 47. The opponents’ cards and the burned ones are just as unknown to you, so they do not change the count.

Each out is one of 47 possibilities on the turn: 2.1 %. On the river, one of 46: 2.2 %. Rounded, 2 % per out for each card to come. With one card to come, your outs times 2; with two, times 4.

The exact count with two cards is one minus the probability of missing both: 1 − (47 − n)(46 − n) / (47 × 46). Against it, the rule of 4 is off in two directions. Each out is worth a little more than 2 % per card, which leaves the rule short with few outs. But adding the two cards double-counts the cases where both cards are outs, when one is enough, and that grows with the square of the outs: from 7 on, the rule overstates, and with many outs it overstates a lot.

The table

The error is the rule minus the exact figure, in points: positive, the rule overstates; negative, it falls short. The rule of 4 is measured against the turn and the river from the flop; the rule of 2, against the river from the turn.

Outs Rule of 4 Exact, 2 cards Error Rule of 2 Exact, 1 card Error Typical draw
2 8 % 8.4 % −0.4 4 % 4.3 % −0.3 pocket pair looking for a set
4 16 % 16.5 % −0.5 8 % 8.7 % −0.7 gutshot straight
6 24 % 24.1 % −0.1 12 % 13.0 % −1.0 two overcards looking for a pair
8 32 % 31.5 % +0.5 16 % 17.4 % −1.4 open-ended straight
9 36 % 35.0 % +1.0 18 % 19.6 % −1.6 flush
12 48 % 45.0 % +3.0 24 % 26.1 % −2.1 flush and gutshot
15 60 % 54.1 % +5.9 30 % 32.6 % −2.6 flush and open-ended straight
20 80 % 67.5 % +12.5 40 % 43.5 % −3.5 —

Up to 8 outs, the rule of 4 stays within a point; with 9, it is one off. With 15 it already overstates by almost 6 points, and with 20, by more than 12. The rule of 2 always falls short, and its error grows slowly: −1.6 with 9 outs, −3.5 with 20.

For the draw you see most, the flush draw, a one-point difference almost never changes a decision. The rule of 4’s big lie is somewhere else.

What the rule takes for granted

To know whether keeping a draw pays, the probability is compared with the pot odds: what you pay divided by the pot you can win.

A typical situation: on the flop you have a flush draw, 9 outs. The pot was 100 chips and the opponent bets 100, the size of the pot. Now the pot is 200 and you have to put in 100 to stay. If you call, the final pot is 300, and 100 of those 300 are yours: to not lose chips in the long run you need to win at least one time in three. 100 / 300 = 33.3 %.

The rule of 4 says 9 × 4 = 36 %. More than 33.3 %: calling pays. The exact count says 35.0 %, a point less, and it pays too. So far, the rounding changes nothing.

But that 35.0 % is the probability of completing the flush on the turn or the river, and the call of 100 does not buy you both cards. It buys you the turn. If your card does not come and the opponent bets again, seeing the river costs you again. What you really bought with those 100 chips is one card: 9 in 47, 19.1 %. Against the 33.3 % you need, it does not pay, and by a lot.

On the turn the count repeats: if they bet the size of the pot again, you need 33.3 % again, and with one card to come you have 19.6 %.

The rule of 4 assumes you will see two cards paying once. That is only guaranteed when there is nothing left to bet, because someone is all-in (or when the opponent checks the turn, but that is not your call). There it holds: 35.0 % against 33.3 %.

In chips, in that same situation, the expected value of calling is the probability times the final pot minus what you put in. All-in, +4.9 chips per call on average. Calling for the turn only, and folding if you miss and get bet into again, −42.6. What you could win in later betting if you complete your hand changes that count.

The correction for many outs

For what the rule overstates with many outs there is a well-known correction: with more than 8 outs, subtract from 4n whatever is above 8, 4n − (n − 8). With 15 outs, 60 − 7 = 53 %, against 54.1 % exact. With 20, 68 % against 67.5 %. Between 9 and 20 outs, the corrected rule stays within a little over a point at most: the worst case is 14 outs, where it falls short by 1.2.

The rule of 2 needs no correction to keep from overstating: it always falls a little short, because on the river each out is worth 2.2 % and not 2 %. As errors go, it errs on the cautious side: it can make you fold a call that did pay, but never make you call one that does not.

The correction does not fix the earlier problem. With the corrected rule, the flush draw on the flop comes to 35 %, almost exact, and it is still the probability of two cards when you only bought one.

To try it with your outs, your pot and what you have to call, there is the rule of 4 calculator: it runs in the browser and gives you the probability you really buy, the pot odds and the expected value in chips.

The rule of 4 does not lie in the rounding: with a flush draw it is off by a point. It lies in what it takes for granted, that you will see two cards for the price of one. And not even the exact count says what will happen in a hand: at 35.0 %, the flush misses almost two times in three, and that does not make the decision a bad one. The full expected value, with what happens on the later streets, and why there are good plays that lose, are left for the next notes on poker: first, computing a probability by Monte Carlo simulation and knowing how far off it is; then, whether poker is a game of skill or of luck.

Israel Negrete Lepe is an electronics engineer. Twenty-five years making different systems work together under real conditions: from the sensor, the vehicle or the transaction to the backend, the data, security and operations. Today, increasingly, edge AI. I take on projects by the hour or by deliverable: israel0negrete@gmail.com.