Equity, outs, and conditional probability
5.1 Equity is conditional
If your hand wins 60% of runouts, loses 38%, and ties 2%, your showdown equity is 60% + 0.5*2% = 61%. Against a range, average this quantity over the opponent combinations, weighted by their probabilities. Equity is therefore a conditional expectation over hidden cards and future public cards.
Showdown equity
Equity = P(win) + 0.5 * P(tie) [heads-up, equal split ties]5.2 Outs
An out is an unseen card that improves you to a hand you expect to be good often enough to matter. If you hold four hearts after the flop, nine hearts remain in the 47 unseen cards. A flush draw therefore has 9 nominal outs to make a flush on the turn.
Draw probability
P(hit on turn) = outs / 47 | Nine-out draw by river: 1 - (38/47)*(37/46) = 34.97%The river expression uses the complement of missing on both turn and river. The common "rule of 2 and 4" says roughly 2% per out with one card to come and 4% per out when you will see two cards. It is a mental arithmetic shortcut, not an identity. Nine outs times four suggests 36%, close to the exact 35%.
5.3 Common draw probabilities
| Draw | Nominal outs | One card from flop | By river from flop |
|---|---|---|---|
| Flush draw | 9 | 19.1% | 35.0% |
| Open-ended straight draw | 8 | 17.0% | 31.5% |
| Gutshot straight draw | 4 | 8.5% | 16.5% |
A rough six-out overcard estimate would be 12.8% for one card and 24.1% by the river, but overcards are often dirty: pairing one can still leave you behind. Conversely, some backdoor sequences create probability mass not captured by a simple out count.
5.4 Dirty outs and domination
Suppose you hold Q-J on T-9-2. Any king or eight completes a straight, giving eight nominal outs. But if two board cards are hearts and you hold no heart, a heart that completes your straight can also complete an opponent’s flush. Nominal outs are not necessarily clean outs: the probability of improving is different from the probability of winning.
Out counts must be tied to the opponent’s range and to the probability that the improved hand actually wins.
5.5 Monte carlo versus exact enumeration
For small states, exact enumeration is possible: iterate over opponent combinations and legal runouts, compute the best five-card hand, and average. For large weighted ranges, Monte Carlo simulation is often convenient. The methodological lesson from the MIT analytics lecture is broader: use computation to replace vague intuition where the state can be parameterized, but keep track of modeling assumptions and sampling error.
Equity is not EV
A hand with 45% equity can be an excellent call if the price is cheap, and a 70% equity hand can face a bad call if the price and future branches are extreme. Equity is an input to expected value, not the decision rule by itself.