Run it twice: one hand, two endings, separate payouts

The all-in is called on the flop. Someone asks, “Twice?” The flop stays where it is. If the room permits the agreement and everyone required agrees, the dealer completes two different turn-and-river runouts. Each decides its own share of the eligible pot. Winning the first does not settle the second.

The useful questions are precise. Has betting actually finished? Which cards are shared? Who can win each pot? What happens to one leftover chip? Mara and the HPD spade work through those questions with practice chips, then examine a nine-out river draw where the second result depends on the first.

Use this as an away-from-table lesson. It does not advise on a live hand, require an account or place wagers.

Jump to consent and eligibility, the exact-card runouts, the side-pot ledger, the payout and variance lab, or the final check. The examples use Hold’em high-hand rankings throughout.

Choose the rule profile before saying yes

The named live reference here is the 2026 WSOP Official Live Action Rules, rules 208–217 and 152–153. This is a documented event profile, not a claim that the event is still running or that every casino uses its rules.

• It permits multiple runouts in time-raked flop games and $5–$10 no-limit Hold’em, with agreement from everyone remaining in the pot
• At least one player must be all-in, with no action pending. Players cannot arrange to check down just to qualify
• Complete one runout before dealing the next. Existing board cards remain shared
• The extra chip between runouts goes to the first, or main, board. Within a tied button-game share, the first tied hand clockwise from the button gets the odd chip

The same profile includes special deck-shortage procedures, including possible reuse of burns or muck, and side-pot-only arrangements. Those require the dealer or floor. Our worked hands have enough cards and unanimous agreement for every pot shown; they do not model those exceptions.

PokerStars’ current help page describes an online setting: any active player with Run It Twice disabled makes the board run once, and it states there is no additional rake for the feature. Its Never, Always and per-table settings are preferences; “Always” still depends on availability. Do not transfer that rake statement to a different room.

Partypoker’s published eligibility terms specifically require a no-limit or pot-limit communal-board cash game, an undealt river, no possible further betting and the feature enabled for everyone still involved. These examples show why a game’s label alone cannot establish whether the option will appear.

Decision check: A is all-in, but B can still raise C. Ready to run twice?

No. B and C still have a betting decision. Finish the legal action first. The existence of one all-in player does not erase a live side-pot decision, and the quoted WSOP profile does not permit an agreement to check down just to activate the option.

A short conversation should identify the number of runouts and the pots covered before any additional community cards are exposed. A player may decline. A spectacular first result is not a reason to renegotiate the second half afterward. If the agreement is unclear, stop and ask the dealer instead of inventing a payout.

Keep the shared cards; replace only the future

In the ordinary same-deck construction used below, the first runout’s new cards stay out of the deck. They cannot reappear in the second runout. The printed shared flop or turn may appear in both descriptions because it is one physical set used by both results. That is different from dealing a duplicate card.

Follow the room’s burn procedure. A burn consumes a physical card but its identity is normally unknown; an odds example must distinguish unknown cards from the dealer’s actual stub. Never put exposed runout cards back to “restore the odds” unless a specific house irregularity ruling directs it.

Case 1: the flop is shared, and each player wins once

A holds A♥ A♦; B holds K♣ K♦. The shared flop is Q♠ 7♥ 2♣. All betting has finished, both agree, and the distributable pot is 400 chips. There are no fees or odd chips in this illustration.

K♥ and 3♠ belong only to the first ending. A♣ and 4♦ belong only to the second. Neither player can borrow a card from the other runout to build a better hand. The final 200/200 payment comes from two independently evaluated hands, even though the draws themselves are not statistically independent.

Case 2: a tie on one runout does not chop the whole pot

A holds A♠ 9♣; B holds A♥ 8♣. The shared four-card board is K♦ Q♥ J♠ 2♦. The pot is 600 chips. Only a river remains.

• First river: 10♣. Both make A-K-Q-J-10, so the first 300-chip share pays 150 each
• Second river: 9♥. A makes a pair of nines; B has ace-high. A takes the second 300-chip share
• Final award: A gets 450 and B gets 150. A has three quarters of the original pot

Try changing the 600-chip pot to 800. What does A receive?

600 chips. The first runout is worth 400, of which A receives 200 for the tie. A then wins all 400 from the second runout. B receives the remaining 200. A’s fraction is three quarters because the pot divides once between boards and again inside the tied share.

Case 3: keep main-pot and side-pot eligibility separate

In this constructed hand, A contributes 100 chips, B 250 and C 250. Nobody has folded, there are no earlier dead contributions, and all amounts are matched. The main pot is 100 × 3 = 300. B and C each add a further 150, making a separate 300-chip side pot. A cannot win that side pot. All three agree to run every pot twice.

A holds A♥ A♦; B holds K♥ K♦; C holds Q♥ Q♦. The shared flop is 7♠ 5♣ 2♥. The first ending is A♣ 4♠; the second is Q♣ 9♠.

On the first runout, A has the strongest hand at the table. That still gives A no claim to the side pot. A’s 150-chip gross award is a 50-chip gain relative to the 100 contributed. B receives 150 after contributing 250, a 100-chip loss. C receives 300 after contributing 250, a 50-chip gain. These net changes sum to zero.

Use the two three-player presets in the payout lab to reproduce the ledger, one pot at a time. If a bet was never matched, return it before building the pots. The separate side-pot and uncalled-bet lesson covers that earlier accounting step.

The odd chip needs two decisions

Take a distributable 101-chip pot in one-chip units. Under the named profile, the first runout receives 51 and the second receives 50. If A wins first and B wins second, A gets 51. Reverse the winners and A gets 50. “One each” can differ by one chip.

Now tie the first runout between A and B, with A first clockwise from the button. Its 51 chips split 26 to A and 25 to B. If B wins the second runout outright, the final totals are A 26 and B 75. The initial extra chip between runouts and the later extra chip between tied players are separate operations.

Payout and variance lab

Both controls run locally in your browser. They send and store no choices. Use Tab to move, arrow keys in menus, and Enter or Space on buttons. Reset affects only its own control. The payout tool accepts already-separated pots; it does not decide card winners or reconstruct contributions.

Why the second river has a different conditional chance

Here is the exact hand behind the model’s default. A holds A♥ A♦; B holds K♠ 9♠. The turn board is 2♠ 7♠ J♦ 3♣. B wins on any of the nine remaining spades and loses on every other river. No river produces a tie. Check this before using “nine outs”: another hand may have dirty outs, a higher flush draw or redraws.

Eight cards are exposed, leaving 44 unseen identities. The calculation conditions only on those eight cards. Unknown burns or folded cards are not extra known dead cards; their hidden identities are averaged over. The 44 is an information set, not a claim that the physical dealer’s stub contains 44 cards. The two river positions are a uniform sample without replacement under the model.

Suppose the first river is 5♠. B wins that half, but 5♠ is now unavailable: eight winning cards remain among 43 unseen identities, so the next-river chance is 8/43, about 18.605%. If the first river is 8♥ instead, all nine spade outs remain: 9/43, about 20.930%. Before either is exposed, each river position has the same 9/44 marginal chance. After the first is exposed, the conditional chance changes.

Try N = 2 and W = 1. Can the two results be independent?

No. With only one winning and one losing card, two draws without replacement must produce exactly one win. Running twice always returns half the pot in this abstract model. The independent-draw comparison would wrongly allow two wins or two losses. With a 400-chip pot, both methods have a 200-chip expected payout, but the true twice-run payout has zero variance.

Expected payout and variance answer different questions

Let P be a fixed distributable pot and let X be the fraction won on one runout: 0 for a loss, 1 for a win, and 1/2 for a heads-up tie. If both runouts have the same marginal expected fraction e, then the twice-run gross award is P × (X₁ + X₂) / 2. Its expectation is P × e. That calculation does not require independent draws.

It does require the same pot and equal board weights. Extra charges, different reward conditions, odd-chip allocation and changes to the underlying wager need their own accounting. An expected gross payment is not profit; subtract the player’s contribution and relevant costs to obtain a net result.

The lab’s simpler no-tie, one-card model lets you calculate the entire distribution exactly. Put p = W/N. The three possible twice-run awards are 0, P/2 and P. Their probabilities are:

• Zero wins: (N − W)(N − W − 1) / [N(N − 1)]
• One win: 2W(N − W) / [N(N − 1)]
• Two wins: W(W − 1) / [N(N − 1)]

For N = 44 and W = 9, there are 44 × 43 = 1,892 ordered pairs of different river cards. Exactly 72 win twice, 630 win once, and 1,190 lose twice. Counted this way, a winning first river cannot be recycled into the second.

One-run variance is P²p(1 − p). Two-run variance in this particular finite-pool model is P²p(1 − p) × (N − 2) / [2(N − 1)]. Standard deviation is the square root of variance, so it is expressed in chips rather than chips squared. For the default 400-chip pot, the variances are 26,033.06 and 12,713.82 chips squared; the mean gross award is 81.82 chips for both.

The independence shortcut would assign two wins probability p², about 4.184%, rather than the correct 3.805%. It would also give exactly half the one-run variance. The model’s actual multiplier is 42/86, about 0.4884. Do not copy that multiplier into a preflop or flop calculation: multiple future cards, ties and shared hand patterns change the joint outcome distribution.

More generally, for two equally distributed runout fractions, Var[P(X₁ + X₂)/2] = P² × [Var(X₁) + Cov(X₁, X₂)] / 2. The covariance records their dependence. The lab’s non-constant one-card model has correlation −1/(N − 1); it is not a universal correlation for poker runouts. A lower variance is not a profit guarantee and does not remove the possibility of losing both shares.

Two runouts are different from a double-board bomb pot

A run-it-twice agreement changes how unfinished community cards resolve an already-played hand after the relevant betting has ended. The cards already on the board are shared by both outcomes. A double-board bomb pot starts with the bomb-pot contribution and two boards as part of its format; its players can still face betting decisions across those boards. It does not need an all-in agreement to become a two-board hand.

The repeated-runout hand also preserves its original hand-construction rule. This page uses Hold’em, where the best five cards can use any number of hole cards. If an eligible Omaha game runs twice, each evaluated Omaha hand still needs exactly two hole cards and three board cards. More boards never grant permission to mix community cards between them.

Finish with a five-question check

Practice check, not a saved record

I can identify who must consent and whether any action remains I can mark the shared cards and the cards unique to each runout I can award every main and side pot only to eligible players I can distinguish a board split from a tie within that board I can explain why equal expected payout does not mean independent results

Does running twice mean I get my contribution back if I win one board?

Only in a particular accounting setup. If two players alone contributed equal amounts and there are no fees, ties or odd chips, one whole board returns half the pot, equal to either contribution. Dead money, unequal contributions across separate pots, fees and ties can change that result. Count the actual pot and the actual award.

Can I accept after seeing the first result?

Treat the run count and affected pots as a settled agreement before the additional cards are exposed. Seeing the first result gives information about the remaining deck. It is not a fresh invitation to change the deal. Ask the dealer about a disputed or unclear agreement.

For another paper exercise, change the three-player main pot to 303 chips and tie its first runout between all three players. Keep A first clockwise. Its first share is 152, paying A 51, B 51 and C 50; if C wins the second 151 outright, the main-pot totals are A 51, B 51 and C 201. Then settle the B/C side pot separately.

Return to HPD’s poker tools and rules lessons when you want a different exercise. The useful habit is to write the eligible pot, shared cards and board share before counting a winner’s chips.