Outs and dirty outs: count the cards that do the job

A card can complete your draw and still cost you the pot. That is the problem with counting a flush draw as nine winning outs before looking at the rest of the hand. The count of remaining hearts may be right. The conclusion about winning may be wrong. Start with the cards that improve your hand, then check which of those cards actually do the job against the hands you are considering.

This lesson uses ordinary high-only Texas Hold’em with a standard 52-card deck. The examples are fictional study hands. We first count cards that complete a straight or flush on the next street. Then we reveal an opponent’s hand on the turn to examine dirty outs with one river left. There is no further betting in that second fixture, so the final hands settle the result.

Amina: “Before you say how many outs you have, tell me what you need the next card to accomplish.” Ace: “Complete the flush?” Amina: “Good. Now check whether that flush wins. Those are two separate questions.”

Give the count a clear target

An out is an unseen card that improves your hand to a likely winner. During a calculation, it helps to call the first list your candidate cards. That makes room for the checks still to come. A queen that pairs your hole card, a seven that completes your straight and a heart that makes your flush are different candidates. They need a reason to belong in the final count.

Suppose you hold two hearts and there are two more hearts on the board. Four of the thirteen hearts are visible in those cards, leaving nine hearts outside them. That is a sound count of flush-completing candidates before other known cards are considered. It says nothing yet about a higher flush, a full house, or an opponent who can improve after you hit.

Avoid silently changing the target halfway through a calculation. If you counted cards that make at least a straight, do not describe the answer as cards that make the nuts. If you counted cards that beat one exact pair, do not carry the same number into a hand against a set. A useful count names the starting hand, board, target and opponent assumption together.

Count physical cards once

Hold Jh Th on a flop of 9h 8h 2c. Here T means ten, h hearts and c clubs. A seven or queen gives you a straight on the next card. Four sevens and four queens make eight straight candidates. Any remaining heart gives you a flush, making nine flush candidates. The two lists are easy to remember, but adding their sizes directly gives the wrong answer.

The 7h and Qh belong to both lists. Each is one physical card, even though it completes both draws. Combine eight straight candidates with nine heart candidates and remove the two duplicates: 8 + 9 − 2 = 15 unique cards. Another way to count is to start with nine hearts, then add only the six non-heart sevens and queens. Both routes reach the same pile.

Name the overlapping cards before subtracting them. A memorized correction of two works for this example because you can point to 7h and Qh. Other combinations of draws need their own check. Two descriptions of an improvement do not guarantee that they overlap in exactly two cards, and a board card or hole card cannot be counted as available again.

Fifteen is a draw-completion count for the next card. No opposing hand has been supplied. Some completions may lose against a particular holding, and a good turn can be overtaken on the river. Removing duplicate cards repairs the addition; it does not establish the chance of winning the hand. Keep that limitation attached to the result when you put it in your notes.

Check the whole hand when a draw arrives

Now consider a different hand on the turn. You have Kh Qh. Your opponent’s revealed cards are As Ad. The board is Ah 7h 2c 9d. Your opponent has three aces and you have a heart draw. The reveal is a teaching device: both private hands are known so we can test each possible river. It is not an assumption that a player normally knows an opponent’s cards.

Nine hearts remain outside the eight known cards. They are 2h, 3h, 4h, 5h, 6h, 8h, 9h, Th and Jh. Every one completes your flush. Seven also win against the revealed aces: 3h, 4h, 5h, 6h, 8h, Th and Jh. Those rivers leave the opponent with three aces while your flush is higher in the hand rankings. There is no later card in this fixture to change that result.

The 2h and 9h are the trap. They give you a flush, but each pairs a rank already on the board. On 2h, the opponent uses three aces and two twos for a full house. On 9h, the opponent uses three aces and two nines. A full house beats a flush. The cards did improve your hand. They improved the opponent’s hand enough to keep you behind.

Ace: “The heart arrived, so I hit.” Amina: “You made the flush. Lay out the opponent’s best five cards too.” Ace: “Aces full. Then that heart never belonged in my winning list.”

Against this exact hand, those two hearts are dirty outs. Subtract them from the nine flush completions to leave seven winning rivers. None of the non-heart rivers gives Kh Qh a hand that beats the aces. Some improve your pair or help the board, but that improvement is insufficient. The full fixture therefore has seven winning rivers, 37 losing rivers and no ties.

Keep the information set with the denominator

The turn board contains four cards, your hand contains two, and the revealed opponent holds two more. Eight distinct cards are known. Subtracting them from 52 leaves 44 legal rivers. The chance of a river in the seven-card winning set is 7/44, about 15.9%, if those legal rivers are equally likely under this study model. Flush completion alone is a different event, 9/44, about 20.5%.

The usual turn starting point of 46 unknown cards uses your two hole cards and the four board cards as the known information. It does not reveal an opponent’s exact holding. Once As Ad are explicitly removed in our fixture, they cannot remain among the possible rivers. Using the seven-card numerator from the revealed hand with a 46-card denominator would mix two information sets.

Unknown opponent cards are not simply available river cards in each exact physical deal. The 46-card view describes uncertainty from your information. A range model must keep each possible holding and its remaining deck consistent. Do not remove guessed folded cards as if you had seen them. Equally, a genuinely exposed card needs to be excluded wherever it affects the candidate list or remaining deck.

Dirty and discounted outs require an assumption

The 2h loses against As Ad in our example because it makes the full house. The same card is not automatically dirty against every other hand. A bare instruction to subtract two outs whenever a board can pair misses the reason for the adjustment. Identify what the opponent can make on that particular card, then compare the actual five-card hands.

With an unknown opponent hand, a candidate can be useful against some holdings and fail against others. For example, pairing an overcard may beat a smaller pair but lose to two pair or a set. Counting all three remaining cards of that rank as clean winning outs assumes too much. Removing all three without checking the range may also throw away useful outcomes.

A discounted-out estimate expresses uncertainty about usefulness. A player might treat a candidate as worth less than one effective out against an assumed mix of hands. There are still whole physical cards in the deck. The fractional value belongs to the model, and it needs a defensible range and weights. Do not turn a vague concern into a precise-looking fraction just to finish the arithmetic.

In a multiway hand, clearing one opponent is insufficient. The card must be checked against the other hands that can contest the pot. A straight may beat the bettor’s pair while losing to a third player’s flush. The extra player changes the showdown question even when your own draw looks identical. Count against the relevant field of hands and keep any side-pot issue separate.

Try the card-set workbench

The first preset contains the two lists from Jh Th on 9h 8h 2c. Inspect them and locate 7h and Qh in the overlap. The second preset contains the nine heart candidates from the revealed-aces fixture, with 2h and 9h supplied as exclusions. Compare the unique set with the retained set. Both exercises show the actual cards so you can challenge the arithmetic.

Edit the lists to test a counting claim. Remove Qh from the heart list in the first preset and notice that the unique total stays at fifteen: Qh remains in the straight list. Remove it from both lists and the total falls to fourteen. In the second preset, clearing the exclusions retains nine candidates, but it does not make the two losing hearts win. You changed a list, not the hand rankings.

The workbench performs set operations only. It rejects invalid card codes, known cards entered as candidates, repeated cards inside a single list and exclusions absent from both candidate lists. It allows a card in both lists because that is the overlap we need to inspect. It does not decide whether a candidate really improves your hand or whether your exclusions are strategically justified.

A good turn can still have a bad river

With two cards to come, a next-card out may give you a temporary lead. You can complete a flush on the turn and lose when the river pairs the board to make an opponent’s full house. A calculation for hitting a card from a fixed target set counts that event. It does not automatically subtract every runout on which the completed draw is overtaken.

Other runouts can help without hitting the draw you first counted. A pair can hold up, two later cards can make a different hand, or the board can produce a tie. Showdown equity includes wins and the appropriate shares of tied outcomes under the model. A quick outs count may leave some of those paths out entirely. State the narrow question it answers before using it for a broader one.

Backdoor draws need two suitable cards. Three hearts among your hole cards and the flop do not mean that any single heart completes a flush. The turn heart creates a direct river draw; a second heart is still needed. Count complete ordered runouts, or use an appropriate two-card model. Adding runner-runner possibilities to a one-card outs list makes unlike events look interchangeable.

Keep the probability separate from the price

Counting outs is one part of studying a decision. The amount to call and the final pot determine the immediate price. Further betting, rake, position, the opponent’s range and the chance of reaching showdown affect how a real hand plays out. None is supplied by a neat candidate total. Fifteen attractive cards does not by itself establish that a call, bet or raise earns money.

Even a correct estimate for making a draw by the river may assume access to both remaining cards. Calling on the flop often buys the turn only; another bet may follow. An all-in model removes that later decision, but it still needs correct final-hand outcomes and ties. Choose the time horizon first, then use a calculation suited to that horizon. Do not let the shortcut choose it for you.

Amina: “What belongs in your review note?” Ace: “The cards I knew, the cards I counted, and why I removed each one.” Amina: “Keep the street and opponent assumption beside them. Another reviewer should be able to reconstruct the count.”

Write a note you can check later

For the overlap hand, save both candidate lists and name the two duplicates. For the dirty-heart hand, record the nine flush completions, the two full-house rivers and the seven remaining winners. Label the opponent’s aces as revealed. These details preserve the logic. A note that says only “missed a big draw” or “had seven outs” leaves too much for memory to invent.

Check the reason for an exclusion before checking the total. A duplicate needs to be counted once, not removed entirely. A known card was never available. A dirty candidate can improve your hand and still lose. A discounted candidate expresses uncertainty across assumed hands. Those corrections answer different problems, so record them separately even when the final number happens to be the same.

Finish a practice hand by naming one retained card and one rejected card, then explaining both best five-card hands. In the revealed-aces example, compare 3h with 2h. The first gives the winning flush; the second gives a losing flush against aces full of twos. Once you can explain that difference, the count follows from the cards instead of from the draw’s name.

For more off-table study, return to Info & Tools and compare the draw-probability and hand-review lessons. These fictional examples are for adults studying poker. They do not supply an opponent range or recommend a live betting action.