Multiway, Squeezes and Limps

More players, different order

Our engine deals all the opponents at once and scores everybody on the same board — a three-way spot is computed three-way, never stitched together from head-to-head numbers. Against one, two and three random opponents:

• A♥A♦K♥K♦: 70.7% / 54.9% / 45.4%.
• A♥A♦8♣3♠: 61.6% / 41.2% / 29.3%.
• J♠T♠9♥8♥: 55.8% / 40.7% / 32.8%.

Look down the last column. Heads-up, bare aces lead the rundown by six points. Against three opponents they have fallen to 29.3% and the rundown is ahead at 32.8%. The order changed. With more players in, "can become the best hand" is worth more than "is the best hand right now" — because somebody is going to make something, and one pair of aces will not be it.

In multiway pots, tighten towards the nuts and raise your threshold for attacking.

Three or more players to the flop, around 100 big blinds.

Tightening is not "only play the nuts". J♠T♠9♥8♥ still has 32.8% against three random hands, while the price to call a squeeze spot is 22.7% (next page). Fold every non-nut holding and you have built exactly the systematic over-fold this rule exists to prevent. Borderline hands are genuinely borderline: play them, and let the flop tell you which ones to keep.

Method: the two- and three-opponent cells are 200000 simulations each, standard error about 0.11 points. The one-opponent cell is not a second run — it is the same 1000000-hand computation already quoted in C3 and C6 (standard error 0.05). One hand against one field gets one number in this book: an earlier version of this page ran that column separately and then had to explain away a first-decimal gap, which is a worse thing to print than the number itself.

The squeeze spot

Coverage note: our reference range still does not cover squeezes or multiway pots, and this one is not a scheduling gap — building a squeeze table needs a joint opponent range of "one opener plus one cold-caller", and no public source gives a cold-calling frequency we could anchor to. We do not model it rather than invent the caller. The amounts below come from the pot calculator and the equities from the engine; the actions are yours to reason out.

The cutoff opens to 7, the button calls, and it is on you in the big blind. The pot is 17 and it costs 5: break-even 22.7%, a much better price than the 33.3% you were getting heads-up, because there is an extra player's money in there. You can also squeeze: the full pot raise is to 29.

The price looks great. But you have to beat two players, not one — and the same trap from C8 is waiting. J♠T♠9♥8♥ is 40.7% against two random hands, far above 22.7%, so calling is right on price. Price was never the binding constraint. Three-way, out of position, first to act on every street, you will realise a smaller share of that 40.7% than the number suggests.

Multiway prices are good, so call more.

When you have position, or when the hand plays itself after the flop and is drawing to nut holdings rather than second-best ones.

A 22.7% threshold is cleared by almost anything — C8 already showed that this style of reasoning has no filtering power in Omaha. The binding question multiway is not "is the price right" but "when I get there, will I have the nuts". The good price is real; it is just not the thing to decide on.

The limped pot, where you cannot raise

Coverage note: limped pots stay outside our reference range, and we do not model them at all, rather than keep promising them. Modelling limps means three interlocking tables — who limps, with what, and how to raise into them — and the public record offers exactly one number to anchor any of it. One number cannot hold up three tables. The amounts here are from the pot calculator and the equity from the engine; the read on what to do with them is reasoning, not a lookup.

2 players limp and you have the option in the big blind. The pot is 7. Your maximum legal raise — the full pot — is to 9. That is the whole range of expression pot-limit gives you here.

Think about what that produces. You raise to 9, two players call because they are getting a wonderful price, and now you are in a small pot with a high stack-to-pot ratio against multiple opponents. That is the hardest combination in the game, and the betting cap handed it to you rather than the other way round.

And the hand strength does not bail you out: even A♥A♦K♥K♦ is only 45.4% against three opponents. Below half, with the best starting hand there is.

When players limp, raise to isolate.

When you have position and the raise size is actually large enough to make continuing wrong for them.

From the big blind here the cap is 9 — four and a half big blinds into a field that is already getting a discount. You will isolate almost nobody. The instinct comes from no-limit, where you can pick any size you like and price people out; pot-limit removes the tool, so the plan has to change with it. Raising is still often right — for value and for initiative — but "raise to get heads-up" is a plan this game will not sell you.

What to take away

• Multiway reorders hands. Bare aces fall to 29.3% against three opponents while a double-suited rundown sits at 32.8%.
• Even the best hand is under half — 45.4% four ways. Nobody is a favourite in a crowd.
• Tighten towards the nuts and raise your attacking threshold — and do not turn that into "only the nuts", or you have swapped one leak for another.
• Prices get better and matter less. 22.7% is easy to clear and tells you nothing.
• Pot-limit caps your isolation raise at 9 in a limped pot. Plan for a crowd, because you are getting one.

Folded into one habit: before you decide to enter, count how many players this pot is going to have, then ask what you will actually have when you get there — the nuts, or the second-best version of it. Answer both before you look at the price, because multiway the price always looks good and is never the thing to decide on.

Coverage reminder: squeezes, limped pots and multiway generally sit outside what our reference range covers, with no date on changing that. What this lesson gives you is mechanism and arithmetic, not a lookup — at the table you run the reasoning rather than copy a row.