Benchmarks
Time bridge-solver in your browser, on fixed work that is the same on every device, so your result can be set beside anyone else's and beside the native tools below. There are two suites: the double-dummy analysis every hand starts with, and the claim checker. One runs at a time.
Double-dummy analysis
Each item is one deal's full double-dummy table: the tricks each of the four players takes as declarer in each of the five strains. The deals are the reference solver's own 200-deal test set.
Loading the engine…
| Deals | Nodes | ms | ms/deal | Native |
|---|
Claim checker
When a player claims, the checker decides whether the claim holds however the unseen cards lie, not just with every card on view. Each item is one claim, generated from a fixed seed. Try claims of your own.
Loading the engine…
| Workload | Cards | V | X | ? | Nodes | ms | ms/claim | Verdicts | Native |
|---|
The claim workloads
- rest-dd
- "The rest are mine." Declarer's side has the high cards and is on lead, and double dummy takes every trick; declarer claims them all. This is the claim players make.
- strong-dd
- Declarer's side has the high cards and claims the double-dummy number of tricks.
- decl-dd, def-dd
- Random hands, with declarer or the opening leader claiming exactly the double-dummy number. The hard case: these fail only when the claimer would have to guess.
- decl-all
- Random hands, declarer on lead claiming every trick. Nearly all are bad claims.
Each claim is V (valid), X (invalid) or ? (the search ran out of its node budget, so the claim is rejected, as BBO's checker does). Only the checking is timed, not generating the claims.
Next: every core, natively
This page uses one thread of your browser. The native tools run the same deals and claims
on every core, faster per core too. Download them for your system from the
latest release:
bridge-solver-<system> for analysis and
bridge-claim-<system> for claims, where the system is
macos-aarch64, macos-x86_64, linux-x86_64 or
windows-x86_64.exe. Then run them from a terminal:
bridge-solver --bench # all 200 deals, all cores bridge-solver --bench quick # the same deals as this page's Quick run bridge-claim bench # every claim workload, 7 to 13 cards, all cores bridge-claim bench --preset quick # the same claims as this page's Quick run bridge-claim check claims.txt # claims of your own; --help gives the format
Tables, verdicts and node counts never depend on the device, so each run here is checked
against the native tools' and only the times should differ. The downloads are not signed.
On a Mac, clear the download flag first with
xattr -d com.apple.quarantine bridge-claim-macos-aarch64 and
chmod +x it; on Windows, choose "More info", then "Run anyway".
Everything runs on this device. When a run finishes, the page records an anonymous
summary: which suite, how long it took, how much work, and your browser and system family,
with no identifier of any kind. Add ?nolog to the address to turn that off.
Privacy · Statistics