Discounted and Tainted Outs
Some cards that complete your draw also complete something better for your opponent — and counting them at full value is how good draws lose stacks. This lesson teaches the full-out / half-out / dead-out discipline and applies it to tainted straight outs, dominated flush draws, and drawing against sets.
Assumptions: All examples use a 100bb-deep 6-max online cash game at $0.50/$1 with no rake unless a different setup is stated inline.
The out counts you've drilled so far carry a silent assumption: that every card which improves your hand wins you the pot. On dry boards against one made hand, that's roughly true. Everywhere else, some of your outs are tainted — the same card that completes your straight completes his flush, or pairs the board and fills his set, or makes you a flush that's second-best by one card. An out that improves you into a still-losing hand isn't an out at all. It's bait.
The fix isn't a new formula. It's a third step inserted into the routine you already have: count the outs by name, then interrogate each one — "when this card comes, do I actually win?" — and only then multiply. The interrogation has three possible verdicts:
- Full out (count 1): the card wins you the pot against everything you realistically face.
- Half out (count ½): the card wins against part of his range and loses (or splits) against the rest — or it wins now but leaves him meaningful re-draws.
- Dead out (count 0): the card "improves" you into the second-best hand against the range you're facing.
Half-counting feels imprecise, and it is — deliberately. You're averaging over his possible hands, and ½ is almost always closer to the truth than the 1 you'd lazily assign or the 0 you'd assign in a panic. Let's make all three verdicts concrete.
Tainted straight outs on a two-tone board
You defend the big blind with 8♣7♣ and the flop comes 9♣6♦2♦ — an open-ended straight draw, eight raw outs: the four tens and the four fives. The cutoff, a solid regular, bets hard, representing overpairs and strong top pairs, plenty of them holding a diamond — and sometimes a straight-up flush draw.
Interrogate the eight cards. The T♠, T♥, T♣, 5♠, 5♥, 5♣ complete your straight and complete nothing on the board: six full outs. But the T♦ and 5♦ put a third diamond out there. Against the part of his range that holds two diamonds, those cards make your straight and his flush in the same instant — the worst card combination in poker, because you improve enough to pay him off. Against his no-diamond hands they're clean wins. Verdict: half outs at best, dead against a known flush draw. Your honest count is roughly 6 to 7 outs, not 8.
The difference shows up brutally in the computed matchups:
Against A♥9♥ — top pair, no diamonds — your 8♣7♣ computes to 37%: eight clean outs plus backdoor clubs and runner two-pair. Against A♦9♦ — the same top pair holding the nut flush draw — you compute to 23%, almost exactly the six-out figure (six outs over two cards is 24%). Fourteen points of equity vanished, and nothing about your hand changed. The board's second suit reached into your pocket.
The table-speed version: on two-tone boards, knock your straight draw down by the one or two outs that wear the flush suit, weighted by how likely he is to actually have that draw. Six-and-a-half outs by the rule of 4 is ~26%; eight outs is 32%. Against a pot-size all-in (needing 33%), that adjustment alone flips the call to a fold.
The dominated flush draw: nine outs that belong to someone else
Now the harshest lesson in the discounting catalog. You complete from the small blind with 9♥6♥ and the flop comes K♥Q♥4♠. Nine hearts in the deck, you counted them, the rule says ~35%. The big blind — who 3-bet you and got called — starts piling chips in.
The question is no longer "how many hearts are there?" It's "when a heart comes, is my flush good?" Your flush is nine-high. Every opponent holding the A♥ or J♥+T♥ has you drawing into a colder deck than you think — and a 3-betting range on a K♥Q♥ board is dense with exactly those cards: A♥x, A♥K, J♥T♥, K♥Q♥-type hands.
The computed spread is the whole story in three numbers:
- vs K♦J♣ (top pair, no heart): 38% — your hearts are full outs, business as usual.
- vs A♥J♥ (nut flush draw + gutshot): 15% — every heart that "saves" you crowns him. Your outs are his outs, only better.
- vs J♥T♥ (straight flush draw): 9% — close to drawing dead while "holding a flush draw."
Averaged across a realistic aggressive range here, your hearts are worth maybe half their face value — call it 4–5 effective outs, not 9. The general principle: a flush draw's value is set by its highest card, not its existence. Nut flush draws get counted in full almost always. Second-nut draws lose a little. Baby flush draws on high, coordinated boards — where big-heart hands love to put in money — should be counted at half strength the moment serious chips go in, and the re-draw matters too: even against his one-heart hands, your made nine-high flush on the turn can still lose to a fourth heart on the river.
This is also why position-and-range context belongs in the count. The same 9♥6♥ draw against a loose button's wide stab is mostly facing no-heart hands: closer to full value. Against a 3-bettor's stack-off range on this exact board: half value. Identical cards, different arithmetic.
Drawing against a set: live outs, dying board
Third taint family: your outs are genuinely yours, but the opponent's hand keeps re-drawing against your made hand — and one of your outs fills him up immediately. You call a button raise in the big blind with J♦T♦ and see a Q♦7♦2♣ flop: flush draw (nine diamonds) plus a backdoor royal-ish straight (Q-J-T needs runner A-K, K-9, or 9-8). The button, it turns out, flopped a set of sevens.
Interrogate the nine diamonds one at a time and two separate discounts appear:
- One out is simply dead. The 2♦ completes your flush and pairs the board: sevens full of deuces beats any flush. Cross it off entirely. (The Q♦ would do the same, but it's already on the flop — this is why you name cards.) Eight live diamonds remain.
- The live ones are only ~80% live. Say the 9♦ hits the turn. You have a flush; he has a set with ten outs to a full house on the river — any queen (3), any deuce (3), any seven (1), plus pairing the turn card itself (3): 10/44 ≈ 23% of your "wins" get clawed back. Each "out" is really worth about three-quarters to four-fifths of a full out.
Stack the discounts: 8 live × ~0.78 ≈ 6.2, plus a little backdoor-straight equity ≈ about 7 effective outs. The computation agrees: J♦T♦ versus 7♥7♠ here is 27% — right at the seven-out two-card figure (28%), and a long way from the 35% the raw count suggested. Against a pot-size all-in needing 33%, the raw counter calls and the discounter folds, and the discounter is right.
This pattern — outs net of his re-draw — is the standard adjustment whenever you suspect a set, two pair, or any made hand with boat potential. You don't need the precise 10/44 at the table. The heuristic "against a probable set, knock 20% off your live outs and kill any out that pairs the board" gets you within an out.
The most-discounted out in poker: the overcard
The catalog's "two overcards = 6 outs" entry deserves its own warning label, because overcard outs fail the interrogation more often than any other type. A flush card completes a flush, full stop. An overcard merely makes you a pair — and a pair only wins if his hand stays a single weaker pair and your kicker plays.
Take A♥K♦ on a T♥7♦2♣ flop against a player who calls your continuation bet. Against a plain top pair like Q♠T♠, your six outs are honest: the computation gives you 23%, right at the six-out book value of 24%. But shift his hand to A♠T♠ — top pair with an ace — and watch: now your three "ace outs" make him aces-up while making you one pair. Only the kings still rescue you, and the computation collapses to 14%, barely above a three-out figure (three outs over two cards is about 12%, with backdoor straights making up the small difference). Half the draw was never there.
This is why experienced players talk about overcards in halves by default: against a calling range that contains Ax-top-pair, two-pair, and slowplayed sets — which real calling ranges always do — six raw overcard outs are typically worth three to four. The interrogation question for every overcard out is specific: "if this card pairs me, does it also improve hands he'd play this way?" An ace is the most dangerous card in the deck to count as an out, because aces live in everyone's strong range. Kings and queens are usually cleaner. Undercard kickers (the five-out pair-plus-kicker draw) face the same audit: your trips outs are near-full value, but the kicker-pairs-to-two-pair outs die against his better two pair or a set.
The same logic, run in reverse, is how you protect made hands: when you hold the top pair and he's calling with overcards or draws, every discount in this lesson is equity flowing back to you. Knowing his ace is only half an out is exactly why thin value bets work.
A quick reference: full, half, dead
The verdicts you'll assign most often, collected in one place:
- Usually full: nut flush draw cards; straight outs in offsuit ranks on rainbow boards; overcards against ranges capped at weak pairs; set-mining outs against overpairs.
- Usually half: non-nut flush outs when bigger suited hands are plausible; straight outs in the flush suit on two-tone boards; overcards against uncapped calling ranges; any out against a player whose line says "set" but who sometimes shows up with one pair.
- Usually dead: flush outs when the obvious bigger flush draw is in his range and he's piling money in; board-pairing cards against probable sets or two pair; outs to the same straight he already has with better cards (the "dumb end" — your 7♣6♣ on T-9-8 hitting a jack just deals you the second-best straight against QJ); ace outs against AK-heavy 3-betting ranges.
None of these verdicts are laws — they're priors you adjust by opponent and action. A passive fish check-calling twice has a very different range from a regular triple-barreling, and the same out can be full against one and dead against the other. The constant is the question itself: when this card comes, who really wins?
The discounting routine, assembled
Everything above compresses into a four-line procedure that runs in about ten seconds:
- Count raw outs by name (previous lessons' discipline).
- Kill the dead ones: outs that complete his likely better draw, pair the board against a probable set, or make you the second-best version of the same hand class.
- Halve the doubtful ones: outs that win against some of his range and lose to the rest; flush draws below the ace on boards his big suited hands attack; straight outs in the flush suit when a flush draw is plausible but not certain.
- Multiply the discounted total by 2 or 4 (with the big-draw correction) exactly as before.
Worked at speed, the three boards from this lesson: 8♣7♣ on 9♣6♦2♦ versus a barreling range — "ten and five, eight raw, the two diamonds are half at best, call it six and a half, times four, about 26." 9♥6♥ on K♥Q♥4♠ versus a 3-bettor's stack-off — "nine hearts but my flush is nine-high into a range full of A♥ and J♥T♥, half value, four-ish outs, maybe 17–18%." J♦T♦ on Q♦7♦2♣ when the jam screams set — "nine diamonds, 2♦ dead, knock the rest down a fifth, about seven, times four minus the correction... 27%."
Two closing cautions. First, don't over-discount: paranoia that assigns every opponent the nut draw is just as costly as optimism that assigns him none. Discount against his plausible range, not against the single scariest hand in it — that's the subject of the next lesson, and discounting is the bridge to it. Second, notice that every number we checked by computation landed within a point or two of the heuristic count. That's the recurring promise of this module: the shortcuts, applied with discipline, track the real math closely enough to bet your stack on.