Comparing Lines: Call vs Raise vs Fold on a Draw
Fold, call, and raise are three separate EVs, not a yes/no question. Work one nut-flush-draw decision all three ways and watch the best line change when fold equity dries up.
Assumptions: All examples assume 100bb effective stacks in a 6-max online cash game with no rake, unless a different stack size is stated in the example.
When a bet lands in front of you, the question is never "is calling profitable?" It's "which of my options has the highest EV?" Fold, call, and raise are three different investments with three different return profiles, and a call that makes money can still be a mistake if the raise makes more.
Nowhere is this clearer than with a strong draw facing a c-bet. The semi-bluff raise is the canonical case where the aggressive line can beat both passive options — because it gets paid two ways: sometimes everyone folds, and sometimes you hit. This lesson works one such decision completely, all three ways, then breaks the answer by changing a single read.
The spot
The button opens 3.5bb, you call in the small blind with A♣4♣ (the site's SB-versus-BTN chart actually prefers a 3-bet with A4s — you flatted this one, and now you have a decision to make), and the big blind comes along. Pot: 10.5bb — call it 10bb to keep the arithmetic clean. The flop is J♣8♣3♥, giving you the nut flush draw plus an overcard. You check, the big blind checks, and the button c-bets 6bb. The big blind is a passive player who has shown he check-folds this texture; treat the decision as heads-up against the button.
Three options. Three EVs. Compute all of them.
EV(fold) = 0
Folding ends your investment in the hand. You win nothing more and lose nothing more from this point: EV(fold) = 0, always and exactly.
This isn't a throwaway line — it's the measuring stick. The chips you've already put in are gone regardless ("the pot's money isn't yours"), so folding never has negative EV and never has positive EV. Every other line must beat zero to be worth taking, and when calling and raising are both losers, "the least bad option" is fold at exactly 0. Fold being best doesn't require your hand to be hopeless; it requires the alternatives to be worse than nothing.
EV(call)
Calling 6bb closes the flop action. The pot is 16bb after his bet, so your price is 6 ÷ (16 + 6) = 27% required equity for the turn card alone.
What do you make on the turn card alone? Nine flush outs from 47 unseen cards = 19%. By direct odds the call fails: 19% is short of 27%.
But the river call from the first lesson had no future; this call does. When the third club lands, the betting isn't over — you hold the nut flush and the button often has top pair or an overpair that pays at least one more bet. Put a number on that: estimate you average +12bb of future profit in the hands where the flush arrives on the turn (sometimes a big turn bet gets called, sometimes he shuts down — 12bb is the average, a read-based estimate, not a computed quantity). The simplified model — you win pot plus future money when the club comes, you check-fold the turn otherwise:
EV(call) = 0.19 × (16 + 12) − 0.81 × 6 = 5.4 − 4.9 ≈ +0.5bb
This model is deliberately conservative: it ignores your ace outs (an ace gives top pair, though against AJ/overpairs it can also cost you money — we zero it out), and it ignores the times the turn checks through and you see a free river. It also ignores the times he barrels you off the draw. Call it +0.5bb with error bars. The point stands: the call only climbs above zero on the back of implied odds, not on its direct price.
EV(raise)
Now raise to 20bb. Two branches, with your read on this button putting his fold rate at 35% — he c-bets this flop with his whole range but continues against a check-raise only with jacks, overpairs, big draws, and better:
Branch 1 — he folds (35%): you win the 16bb sitting in the middle. Contribution: 0.35 × 16 = +5.6bb.
Branch 2 — he calls (65%): he puts in 14 more; the pot is 10 + 20 + 20 = 50bb, of which you invested 20bb at this decision. What's your equity when he continues? Against a realistic continue range — JJ+, AJ, KJs, QJs, plus sets of eights and threes — A♣4♣ on J♣8♣3♥ has 38% equity by simulation (nine clubs plus an ace that's sometimes good). Stipulate, to keep this a one-street calculation, that once he calls the raise the rest of the money stays out and you realize your raw equity:
when-called EV = 0.382 × 50 − 20 = 19.1 − 20 = −0.9bb
Read that twice, because it's the engine of the whole semi-bluff: getting called costs you almost nothing. You put in 20bb with 38% of a 50bb pot — a nearly break-even gamble on its own. So the fold branch is almost pure profit:
EV(raise) = 0.35 × 16 + 0.65 × (−0.9) = 5.6 − 0.6 = +5.0bb
The ranking
- EV(raise) ≈ +5.0bb
- EV(call) ≈ +0.5bb
- EV(fold) = 0
The raise isn't slightly better — it's worth ten times the call. And notice the anatomy of why: it's not that he folds a lot — 35% is nowhere near the 56% a pure bluff of this size would need (risking 20 to win 16, breakeven = 20/36). It's that your 38% equity subsidizes the call branch down to roughly zero cost, so any meaningful fold equity becomes nearly pure profit. This is the formal version of "semi-bluffs need fewer folds": the draw pays for the disaster branch.
A hand with no equity raising here would be torching money; a hand with this much equity raising here is printing it. Same raise, same fold rate — the difference is entirely in what happens when called.
Break the read, break the ranking
Same hand, different button: a calling station who treats a check-raise as a personal challenge. Your fold-rate estimate drops from 35% to 15%. Recompute — the when-called branch is unchanged at −0.9bb:
EV(raise) = 0.15 × 16 + 0.85 × (−0.9) = 2.4 − 0.8 = +1.6bb
Still positive — the draw's equity keeps the raise from ever being a disaster here — but the engine has stalled: two-thirds of the line's value just evaporated with the fold equity.
Meanwhile the call got better against this player, because the same stickiness that ruins your fold equity fattens your implied odds. A station with top pair pays off the flush far more reliably; bump the future-profit estimate from 12bb to 20bb:
EV(call) = 0.19 × (16 + 20) − 0.81 × 6 = 6.9 − 4.9 ≈ +2.0bb
New ranking:
- EV(call) ≈ +2.0bb
- EV(raise) ≈ +1.6bb
- EV(fold) = 0
The flip is the lesson. Nothing about your cards or the board changed — one read moved, and it moved two EVs in opposite directions, because fold equity and implied odds are two claims on the same tendency. A player who folds too much gives your raises value and your calls little payoff; a player who never folds does the reverse. Aggression isn't "better" or "worse" in general. It's better or worse against someone.
Two honest caveats on the station math. His check-raise-calling range is wider (more middling Jx), which nudges your when-called equity above 38% and helps the raise a little; but a station also keeps betting turns, which taxes the raise line in ways our frozen-pot model doesn't capture. The numbers above are estimates with simplifications stated — which is exactly what a by-hand EV comparison is supposed to be. The decision they point to (raise the folder, call the station) is robust even if every figure is off by a bb.
Sizing is a fourth option hiding inside the third
"Raise" isn't one line — every size is its own branch with its own EV, and the structure you just built prices them instantly. Take the original 35%-folder and compare the 20bb check-raise against a smaller raise to 14bb. The smaller raise threatens less, so suppose it only folds him out 25% of the time. When he calls 8 more, the pot is 10 + 14 + 14 = 38bb:
when-called EV = 0.382 × 38 − 14 = 14.5 − 14 = +0.5bb
EV(raise to 14) = 0.25 × 16 + 0.75 × 0.5 = 4.0 + 0.4 = +4.4bb
Notice what changed and what didn't. The smaller raise buys fewer folds (25% vs 35%) but gets a better price on its equity — the when-called branch actually flips positive, because you're putting in 14bb for 38% of a 38bb pot. The two effects nearly cancel: +4.4bb vs +5.0bb. Against this villain the bigger raise wins, but the gap is small enough that either size is fine — and against an opponent whose fold rate barely responds to sizing, the small raise can come out ahead. The deeper point: fold equity and pot-share price trade against each other through sizing, exactly the way fold equity and implied odds traded against each other through player type. Every knob you turn moves two numbers in opposite directions, and only the EV comparison tells you which effect wins.
(A raise size that would be a clear error: tiny, like a min-click to 12bb. It folds out almost nothing, so you've spent your aggression budget to inflate a pot you're 38% in, without the fold-equity engine that justified raising at all.)
The mistakes this framework kills
Four errors show up constantly in beginner line selection, and all four dissolve once you insist on writing three numbers before acting:
- Binary thinking. "Is this a call?" The honest answer to that question is sometimes "yes, and it's also the second-best option by 4.5bb." Always price all three lines, even if two of them get ten seconds of thought and one gets two.
- Raising on equity alone. "I have a flush draw, so I'm raising" skips the half of the EV that comes from fold equity. Against the 15%-folder, the raise survived only because the draw was strong; a weaker draw — say a gutshot with 17% when called — raising into a station is just lighting chips: nearly no folds, terrible pot share when called.
- Calling on stubbornness. The reverse error: "I can't fold the nut flush draw" is true here (both aggressive lines beat folding comfortably), but it's not true because the draw is pretty — it's true because the price, implied odds, and equity say so. Change the spot to a 15bb bet into 10bb and the direct price collapses; recompute before reciting slogans.
- Ignoring that EVs move together. When you re-read a player as stickier, don't just downgrade your bluffs — upgrade your value bets and implied-odds calls in the same breath. One read shifts every line's EV at once, and the player who only updates half their numbers ends up with a strategy at war with itself.
How to run this at the table
You won't produce +5.0 vs +0.5 vs 0 in real time. You can absolutely produce the structure:
- Fold is 0. Free.
- Call: direct price versus one-card odds, then ask whether implied odds plausibly cover the gap. Here: need 27%, have 19%, nut draw against a range that pays off — probably slightly above water.
- Raise: two questions. Does he fold enough to matter? and Am I okay when called? With 38% equity when called, "okay when called" is nearly automatic, so any real fold equity makes the raise the front-runner.
- Pick the biggest number. Re-pick when the player type changes.
The deeper habit this builds: stop grading lines in isolation. "Calling is +EV" is an incomplete sentence in a world where raising might be +5. Leaving 4.5bb on the table every time you take the second-best line is invisible in your results — you'll win the pot sometimes either way — and it is exactly the kind of leak that separates a 2bb/100 winner from a breakeven reg. The full tree-building method for spots with more branches is the last lesson of this module; the three-line comparison you just ran is its foundation.