6 Expected Value and Decision-Making
Every decision you make at the poker table (every fold, call, bet, raise, and check) has a price tag and a payoff. Expected value (EV) is the tool that lets you read those price tags. It is the most important concept in this book: once you understand it, poker stops being a list of memorized rules and becomes one question you ask on every street: which of my available actions makes me the most money on average?
By the end of this chapter you will be able to calculate the EV of a call, a bet, and a bluff, compare candidate lines by EV, and judge your decisions by their quality rather than their result.
6.1 What Expected Value Actually Means
Expected value is the average result of a decision if you could repeat it an infinite number of times. It is a weighted average: multiply each possible outcome’s value by its probability, then add them up.
The general formula is:
\[ EV = \sum_{i} P(\text{outcome}_i) \times V(\text{outcome}_i) \]
In plain English: probability of each result times the money won or lost in that result, summed across all results.
A few ground rules before we touch a single hand:
- Work in big blinds (bb), not dollars or chips. A decision that wins 3bb wins 3bb whether you play a $0.50 game or a $50 game, so your reasoning stays portable across stakes.
- Count only the money still in play. Chips you put in on earlier streets are gone. EV is always computed from this decision forward. This is the most common place where intuition betrays beginners, and we will return to it.
- Positive vs. negative EV. A +EV play makes money on average; a –EV play loses money on average. Your job at every decision point is to pick the option with the highest EV available.
EV is not a prediction of what will happen this hand. You will often make a clearly +EV call and lose the pot. EV is a statement about the long run, the average of thousands of identical spots. Your bankroll is the sum of those averages, so playing +EV relentlessly is how you win even though any single hand can go either way.
6.2 The EV of a Call
Calling is the simplest action to analyze because there are usually only two outcomes: you win the pot, or you lose your call.
Suppose the pot is 10bb and your opponent bets 6bb, making the total 16bb. It costs 6bb to call, and you believe you have the best hand, or will improve to it, 40% of the time.
When you win, you collect the 16bb in the pot (the original 10bb plus your opponent’s 6bb bet). When you lose, you forfeit the 6bb you called with. So:
\[ EV_{call} = (0.40 \times 16) - (0.60 \times 6) \] \[ EV_{call} = 6.4 - 3.6 = +2.8\ \text{bb} \]
Calling here is worth +2.8bb on average. The “win” amount is the full 16bb pot after the bet went in, since that is what you drag; we do not subtract the call from the win side. (You can also compute it the gross way: invest your 6bb call every time and drag the full 22bb pot 40% of the time, so 0.40 × 22 − 6 = +2.8bb. Same answer; pick one method and stay consistent.)
This formula is the engine behind pot odds, covered in its own chapter. The call breaks even when your win probability equals the price you are getting: here that breakeven is 6bb / (16bb + 6bb) = 6/22 ≈ 27%, and your 40% clears it comfortably.
6.3 The EV of a Value Bet
When you bet for value, you want worse hands to call. Three outcomes are possible: villain folds, villain calls and you win, or villain calls and you lose (you were behind, or got outdrawn). On the river, with no more cards to come, split the bet by what villain does.
Pot is 20bb. You bet 15bb. Your read: villain folds 50% of the time, and when he calls you are ahead 70%.
Break the tree into branches:
- Villain folds (50%): you win the 20bb pot. Value = +20bb.
- Villain calls and you win (50% × 70% = 35%): you win the 20bb pot plus his 15bb call = +35bb.
- Villain calls and you lose (50% × 30% = 15%): you lose your 15bb bet = –15bb.
\[ EV_{bet} = (0.50 \times 20) + (0.35 \times 35) + (0.15 \times -15) \] \[ EV_{bet} = 10 + 12.25 - 2.25 = +20\ \text{bb} \]
Compare that to checking and giving up, or checking to win at showdown some fraction of the time. The bet’s +20bb is the number any alternative line must beat. That is the core of decision-making: you judge each play against its best alternative.
6.4 The EV of a Bluff
A pure bluff is the cleanest EV calculation in poker: with no showdown value, you win only when villain folds.
The formula for a bluff is:
\[ EV_{bluff} = (P_{fold} \times \text{pot}) - (P_{call} \times \text{bet}) \]
You win the current pot when he folds and lose your bet when he calls. (We are ignoring the rare cases where you bluff, get called, and hit a miracle card; that adds a small positive term if a card is to come.)
Pot is 12bb. You bluff 8bb on the river. How often must villain fold for this to break even? Set EV to zero:
\[ 0 = (P_{fold} \times 12) - ((1 - P_{fold}) \times 8) \] \[ 0 = 12\,P_{fold} - 8 + 8\,P_{fold} \] \[ 8 = 20\,P_{fold} \quad\Rightarrow\quad P_{fold} = 0.40 \]
You need villain to fold 40% of the time to break even. This is the breakeven bluff frequency, and the shortcut is worth memorizing:
\[ \text{Breakeven fold \%} = \frac{\text{bet}}{\text{bet} + \text{pot}} \]
Here: 8 / (8 + 12) = 8/20 = 40%. This is the risk fraction of your own bluff, sometimes called alpha, and it belongs to the bettor. Don’t confuse it with the price villain gets to call, a separate quantity: he is laid 8 / (12 + 8 + 8) = 8/28 ≈ 29%. A few reference points worth knowing cold:
| Bet size (relative to pot) | Fold % needed to break even |
|---|---|
| Half pot (0.5x) | 33% |
| Two-thirds pot (0.67x) | 40% |
| Three-quarters pot (0.75x) | 43% |
| Full pot (1.0x) | 50% |
| Overbet 1.5x pot | 60% |
Bigger bluffs need to work more often. A pot-sized river bluff must succeed 50% of the time just to break even; a half-pot bluff only 33%. Bet sizing is therefore a strategic lever: it directly sets the bar your read must clear. If a player folds 45% of the time, a half-pot bluff (needs 33%) is comfortably +EV, but a pot-sized bluff (needs 50%) is –EV against the very same player.
Let’s confirm one is +EV. Against an opponent you estimate folds 55% to that 8bb bet into 12bb:
\[ EV_{bluff} = (0.55 \times 12) - (0.45 \times 8) = 6.6 - 3.6 = +3.0\ \text{bb} \]
This is clearly profitable: 55% comfortably clears the 40% you needed.
Beginners bluff “because villain might fold” without ever asking how often he must fold. “Might fold” is not an analysis. The real question: does his folding frequency beat the breakeven number your bet size demands? A bluff that gets through 35% of the time is a disaster at pot-size but a small +EV play at quarter-pot, where it clears the 20% breakeven (per-bluff profit about +0.19 pots: 0.35 × 1 − 0.65 × 0.25 in pot units). Same fold rate, opposite verdict, because the sizing changed the math.
6.5 Comparing Candidate Lines by EV
Real hands are branching trees of decisions. The professional lays out the two or three realistic lines, estimates the EV of each, and takes the highest. You rarely have exact numbers (your inputs are educated estimates), but even rough EV comparison beats gut feeling: it makes your assumptions explicit and tests whether they hold up.
Consider a turn spot. The pot is 30bb and you hold a flush draw plus an overcard: nine outs (cards that give you the best hand) to the nut flush, with a chance your overcard is live. You face no bet, you are last to act on the turn, and you have 40bb behind. Your candidate lines:
- Check back and try to realize equity / hit on the river for free.
- Bet 20bb as a semi-bluff, hoping to win immediately but with equity as a backup when called.
Suppose you estimate about 18% equity from the nine flush outs (rule of 2: nine × 2 ≈ 18% on one card), call it ~20% once you add a live overcard, and you estimate the 20bb semi-bluff folds villain out 45% of the time.
Line 1 — check back. You see a free river and win the 30bb pot about 20% of the time when you hit (ignoring the times your unimproved hand is already best, to keep the model lean):
\[ EV_{check} \approx 0.20 \times 30 = +6\ \text{bb} \]
Line 2 — semi-bluff 20bb. Break it into branches:
- Villain folds (45%): win 30bb now → +30bb.
- Villain calls (55%), then you hit on the river (20%): you win the 30bb pot plus his 20bb call = +50bb. Combined probability 0.55 × 0.20 = 11%.
- Villain calls (55%), then you miss (80%): you lose your 20bb bet = –20bb. Combined probability 0.55 × 0.80 = 44%.
\[ EV_{semibluff} = (0.45 \times 30) + (0.11 \times 50) + (0.44 \times -20) \] \[ EV_{semibluff} = 13.5 + 5.5 - 8.8 = +10.2\ \text{bb} \]
The semi-bluff at +10.2bb beats the check-back at +6bb, so you bet. Be fair, though: the check-back estimate is deliberately conservative. It credits only the 20% hit-and-win branch and ignores both the extra river value you collect when you hit and the times your unimproved ace-high is best. Crediting those would shrink the gap to well under 4.2bb. The bet still wins, but it wins by adding a second route to the pot: fold equity now (the value of the chance villain folds) on top of your draw later, the two ways to win. That edge is invisible to results-oriented thinking, where a missed semi-bluff just looks like a loss.
This model is simplified: it ignores rivers where you miss but can bluff again and the times villain raises. A solver (software that computes near-optimal play) accounts for all of it. But the framework, enumerate the lines, branch each into outcomes, then weight and sum, is identical on a napkin or across a solver’s millions of nodes.
6.6 A Fully Worked Example Hand
Setup. $1/$2 online cash, 100bb effective. You are in the big blind with A♥5♥. The cutoff opens to 2.5bb, the button calls, and you call. Three players, 8.0bb in the pot (your 2.5 plus theirs, 2.5 + 2.5, plus the small blind’s dead 0.5).
Flop: K♥9♥4♣. You have the nut flush draw, nine clean outs to the best hand, plus a backdoor wheel possibility (a draw that needs two running cards) and an ace that may or may not be good. You check, the cutoff bets 5bb, the button folds. Pot is now 13.0bb (the 8.0bb plus the cutoff’s 5bb) and it costs 5bb to call.
The call decision. With nine outs you complete your flush by the river about 35% of the time, but you won’t always see both cards for free, so price the call against this turn card alone (about 19% to hit) plus the implied value of stacking a made hand. Your draw, the ace outs, and implied odds clear the 5 / (13.0 + 5) ≈ 28% pot odds comfortably. Call. (A check-raise semi-bluff is also a candidate line, but take the call to keep the example moving.)
Turn: 2♠. This card helps a little. The deuce gives you a live gutshot (an inside straight draw) to the wheel: holding A-2-4-5, any 3 completes A-2-3-4-5, and three of those (3♠, 3♣, 3♦) are extra clean outs beyond your flush draw, since the 3♥ was already one of your nine flush outs. (The wheel is effectively the nuts on those rivers: no flush is possible and a five-high straight beats any two pair or set.) That lifts you to roughly twelve outs. The pot entering the turn is 18.0bb (13.0bb + your 5bb call). The cutoff now bets 16bb, a big turn barrel (a continuation bet on the next street), leaving roughly 76bb behind. With the nut flush draw plus the wheel gutshot you have about 26% to improve on the river (twelve outs, 12/46 ≈ 26%).
Now the EV of calling 16bb. When you call and win, you drag the 18bb pot plus the cutoff’s 16bb bet, 34bb in all (the win amount includes the bettor’s bet, as before):
\[ EV_{call} = (0.26 \times 34) - (0.74 \times 16) = 8.8 - 11.8 = -3.0\ \text{bb} \]
On raw pot odds alone, this call is still –EV. You need 16 / (18 + 16 + 16) = 16/50 = 32% equity (your 16bb call has to win the 18bb pot plus both 16bb bets), and you have about 26%. So why call profitably? Implied odds: the extra money you expect to win on the river when you hit. The break-even extraction is only about 3.0 / 0.26 ≈ 12bb, modest against villain’s ~76bb stack. So if, on the ~26% of rivers where you improve, you can extract even 20bb more on average, you add roughly 0.26 × 20 ≈ +5.2bb to the call’s EV, covering the –3.0bb gap and turning the call clearly profitable (≈ +2.2bb). Whether it is correct depends on how much you actually get paid: against a station (a player who calls far too often) who never folds top pair, call all day; against a thinking player who shuts down when the flush comes in, fold.
River: 7♥. You hit the nut flush. Pot is 50.0bb (the 18.0bb entering the turn plus both 16bb turn bets). Villain checks. You hold the absolute nuts, so the only question is a value-betting one: what size maximizes (probability of being called) × (amount called)?
If you shove 76bb and get called 25% of the time, the bet adds 0.25 × 76 = 19bb of value. If you bet 35bb and get called 60% of the time, it adds 0.60 × 35 = 21bb. The smaller bet makes more here because it more than doubles your call rate while only roughly halving your size (and you collect the 50.0bb pot either way). Even with the nuts, sizing is an EV optimization, and the biggest number on the betting menu is often not the most profitable. The exception: against a pure calling station who never folds, the largest size (overbet or shove) genuinely can be the highest-EV value bet, because when the call rate barely drops, max sizing wins the most. The skill is matching the size to how the specific opponent calls.
Take the river spot above. Build a three-row table for bet sizes of 20bb, 35bb, and 76bb (shove). For each, estimate villain’s calling frequency, then compute (call% × bet size) to find the extra value each size captures. Which size wins? Now do the same for the turn call: write the call’s raw EV (you computed ≈ –3.0bb), then solve for the average river extraction you’d need to make the call breakeven. Repeat for five different hands to build the reflex.
6.7 Process Over Results: The Mindset Shift
The single most expensive habit in poker, and almost everyone has it to some degree, is judging a decision by whether the hand was won or lost. This is results-oriented thinking, and it is the main reason players stop improving.
Poker is a game of incomplete information and large variance. You can make a flawless, clearly +EV call and lose; you can make a reckless, –EV hero call and stack someone. The professional standard is to evaluate each decision using only the information available at the time, then accept whatever the cards do. A good decision that loses is still a good decision; a bad decision that wins is still a bad decision. What matters is whether, given what you knew, you chose the highest-EV action; who won the pot is beside the point. Let the result grade the decision and you learn the wrong lessons: you reinforce mistakes that happen to win and abandon good plays that happen to lose.
There is a one-word name for this error: resulting, equating decision quality with outcome quality. In review, never answer “did I play that hand well?” by looking at who won.
“I called and he had it, so I should have folded.” No. The hand he turned over is one sample from his entire range. If his range at that moment held enough bluffs and worse value hands that your call beat the whole range, the call was correct even though this specific instance lost. Judge the call against the range you faced, not the one hand you got shown. The same logic runs the other way: “I bluffed and he folded, so it was a great bluff”, maybe, but if he folds only 25% and your sizing needed 50%, you made a –EV play that got lucky. The fold doesn’t vindicate the bluff.
How do you internalize this? Three practices:
- Separate decision review from results. When studying hands, cover the result. Ask only: what were the available lines, and which had the highest EV given my reads? Reveal the outcome last, and only to check whether your inputs were calibrated, not to grade the decision.
- Track your decisions, not just your bankroll. Note spots where you knew the +EV play and made it, regardless of how they turned out. A session where you made twelve good decisions and lost five buy-ins to variance is a successful session by the only metric you control.
- Let the long run judge you. Variance is loud in the short term and silent in the long term. Over a single session, luck dominates. Over tens of thousands of hands, EV dominates and luck washes out. Keep making +EV decisions and let the sample size do its work.
You cannot control whether you win a hand, only whether you make the +EV decision. Detach “played well” from the outcome and attach it to the quality of your reasoning. Players who make this shift stop tilting (letting frustration drive their play), stop chasing losses, and start improving, because they are finally optimizing the one thing they can control.
6.8 Putting It Together
Expected value is the grammar of poker. Pot odds, implied odds, fold equity, bet sizing, bluff frequency, and value extraction are all applications of the same EV calculation. When you are unsure what to do, fall back to the fundamental procedure:
- List your realistic options (fold, call, raise; check, bet small, bet big).
- For each option, enumerate the outcomes and estimate their probabilities from your read of villain’s range.
- Weight each outcome by its probability, sum to get the EV of that option.
- Choose the highest.
- Judge yourself on steps 1–4, never on whether the river bricked.
You won’t run these calculations to three decimals in real time; nobody does. But drilling them away from the table builds the intuition to feel the EV live: this bluff is too big for how sticky he is; this call is a clear price; this half-pot value bet is too small for what the hand is worth. That intuition, grounded in arithmetic you have actually done, is what poker expertise looks like. Everything else in this book applies the principle you just learned: at every decision, find the +EV play, make it, and trust the long run.
- Compute EV from the current decision forward, in big blinds, weighting each outcome by its probability.
- Compare your realistic lines and take the highest EV.
- A bluff’s breakeven fold frequency is bet / (bet + pot); bigger bets must work more often.
- Implied odds can turn a draw that is priced out on raw pot odds into a profitable call.
- Judge decisions on the information you had, not the result. Catch yourself resulting.