What Double Dummy Analysis Tells You (And What It Doesn't)

What a double dummy solver actually computes

"Double dummy" describes a form of the game nobody plays: every card is face up. Declarer knows where all 26 defensive cards are, and each defender knows the other three hands as well. Under those conditions the deal has an exact value — a number of tricks that neither side can improve on against best play. That number is what a double dummy solver returns.

The computation is a search, not a simulation. The solver explores the tree of legal plays and evaluates each position assuming both sides continue optimally; alpha-beta pruning and transposition tables bring an otherwise astronomical search down to milliseconds. Because it searches a fully specified deal, the answer is deterministic: the same 52 cards always give the same table, and there is no sampling error to allow for.

The result is a 20-cell table — five strains by four possible declarers. Reading down a column tells you how the deal plays for one declarer in every strain. Reading across a row tells you how the same contract fares from each side of the table, which matters more often than people expect: a contract can be cold from one hand and down one from the other, purely because of which defender is on lead.

A worked example: 24 high card points and no notrump game

North-South hold 24 high card points between them and a four-four fit in both majors:

A pair using high card points as the primary guide to game will often land in 3NT here. The solver's verdict:

3NT is down one against best defence, while four spades and four hearts both make exactly. The two extra tricks in a suit contract are not conjured from nothing. In notrump, North-South have no long suit to establish and no way to stop the defenders cashing out; in a major, the small cards in the short hand ruff, and the defenders' club and diamond winners stop being winners.

The point is not that the deal is exotic. It is that high card points measure one thing and trick-taking measures another. The double dummy table separates them, because it prices every strain on the same 52 cards.

Notice also that both minors produce six tricks for either side. When the two partnerships can take the same number of tricks in a strain, the deal is close to a pure race, and the table tells you at a glance that neither pair has a cheap sacrifice available.

Why the number is not what you would make at the table

The most common misuse of a double dummy result is to treat it as a prediction. It is not, and the reasons cut in both directions.

• The solver's defenders see everything. They find the only lead that beats the contract, the only switch that stops a discard, and the duck that severs communication — with no clue from the auction or from partner's carding. A defender leading blind at trick one is playing a genuinely harder game.
• The solver's declarer also sees everything. It never misguesses a two-way finesse, never falls for a falsecard, and never takes a safety play it did not need. Contracts that hinge on locating a queen are cold double dummy and coin flips in practice.
• The two effects do not cancel deal by deal. Where a contract depends on finding a blind opening lead, the double dummy result understates declarer's real chances; where it depends on a two-way guess, it overstates them.

So a double dummy result is best read as the value of the deal rather than the expectation for a table. If you want to know how a contract will fare when only your own cards are known, that is a different computation — a single dummy simulation.

What it is genuinely good for

• Choosing between strains. This is the strongest use. The table prices every denomination on the same deal, so a question like "was the four-four major fit really better than the five-three?" gets an exact answer.
• Settling sacrifice and par questions. Reading both sides of the table shows whether five of a minor doubled was cheaper than letting the opponents play their game.
• Separating a wrong contract from bad play. If the contract you reached is a trick short double dummy, no technique was going to bring it home, and the post-mortem belongs in the auction.
• Calibrating judgment over many deals. Run the same class of hand repeatedly and you build a feel for what a fifth trump, a side-suit singleton, or an honour wasted opposite shortness is really worth.

Double dummy or single dummy?

The two analyses answer different questions, and choosing between them is mostly a matter of what you are allowed to see.

Use double dummy when the deal is over and you want the truth about it. Use single dummy while you are still deciding — bidding a close game, choosing between two contracts, judging whether a suit is worth playing for — because it prices the decision across every layout consistent with what you actually know.

Frequently asked questions

Is a double dummy result an upper bound on declarer's tricks?

Only against perfect defence. Real defenders miss the killing lead often enough that declarers regularly take more tricks at the table than the double dummy figure; on other deals they take fewer, because declarer faces a guess the solver never has to make.

Why can the same contract be worth different tricks depending on who declares it?

Because the opening lead comes from a different hand. Leading through a holding like K J x is far more attractive for the defence than leading round to it, so a tenace that is protected from the lead in one direction is exposed in the other. The two columns for each partnership make this visible at a glance.

Do I have to enter all 52 cards?

For double dummy analysis, yes — the computation is only defined on a fully specified deal. If you know only your own side, use the single dummy solver, which fills in the rest many times over and reports the distribution.

Which opening lead does the solver assume?

The best one, chosen with knowledge of all four hands. That is exactly the assumption a defender cannot make at the table, and it is the single largest source of difference between double dummy results and table results.