RIVER POKER LAB · DEVICE TEST 1

Test solving on this phone

Both engines solve the same small games locally. No results are uploaded. Keep Safari open while testing.

Phone model (optional) Include flop through river (may take several minutes) Repeat three times for more stable timings

Ready. Start with the quick river and turn tests.

Preparing offline files…

What this measures

Single-thread WebAssembly in dedicated workers. Every run uses a fresh worker. Solve time includes game setup and result extraction; total time also includes loading and compiling the engine. Accuracy target is the sum of both players’ possible improvements, at most 0.5% of the starting pot. This is a comparison threshold, not a promise of full-game accuracy.

Memory is WebAssembly allocation, not all memory used by Safari. Core-storage estimates from different engines cover different things. A 128 MiB core-storage guard and two-minute timeout limit each run. These limits cannot guarantee Safari will retain the tab.

The rake sanity test should report zero possible improvement. The newer engine has a known negative diagnostic on this test; that is a failed diagnostic, not exceptional accuracy. The timing games have no rake or raises and use small ranges. Large ranges, full size menus on every street and general raises still need testing.

Sources and licenses · WASM Postflop source snapshot · Other engine source snapshot