logaritmisk

logaritmisk pushed tag v0.4.2 to logaritmisk/trueskill-tt

2026-09-07 20:47:03 +00:00

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 20:43:58 +00:00

8116fd081f test: localise the erfc_inv tail residual to the caller's argument

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 20:33:41 +00:00

17d072b2ae fix: route every transcendental through libm, and combine sigmas with hypot

logaritmisk commented on issue logaritmisk/trueskill-tt#41

2026-09-07 20:25:37 +00:00

erfc is a 1.2e-7 approximation and sets the accuracy floor for the whole crate

Done in 3dd6593. The answer to "is it worth paying for" turned out to be that there is nothing to pay.

Accuracy

libm (the Rust port of FDLIBM), measured against an independent incomplete-g…

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 20:25:12 +00:00

3dd659307a fix: replace the erfc approximation with libm, for free

logaritmisk closed issue logaritmisk/trueskill-tt#41

2026-09-07 20:25:11 +00:00

erfc is a 1.2e-7 approximation and sets the accuracy floor for the whole crate

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 19:49:14 +00:00

7e289ee834 chore: Release trueskill-tt version 0.4.1

logaritmisk pushed tag v0.4.1 to logaritmisk/trueskill-tt

2026-09-07 19:49:14 +00:00

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 19:46:54 +00:00

564969ee5d test: pin quality()'s N-group closed form, closing the README cross-check

683813ec10 fix: correct erfc_inv's sign error and keep evidence in log space

Compare 2 commits »

logaritmisk opened issue logaritmisk/trueskill-tt#42

2026-09-07 14:02:29 +00:00

The Schedule trait is public API that the engine never uses — wire it into run_chain or make it private

logaritmisk pushed tag v0.4.0 to logaritmisk/trueskill-tt

2026-09-07 13:49:20 +00:00

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 13:49:20 +00:00

2a48d10aa9 chore: Release trueskill-tt version 0.4.0

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 13:48:39 +00:00

d4f91fd221 fix: reject convergence options that silently disable inference

logaritmisk commented on issue logaritmisk/trueskill-tt#41

2026-09-07 13:43:20 +00:00

erfc is a 1.2e-7 approximation and sets the accuracy floor for the whole crate

Independent confirmation and concrete replacement candidates, from a literature/ecosystem survey run separately from the measurements above.

The measurement reproduces

The crate's erfc was…

logaritmisk commented on issue logaritmisk/trueskill-tt#20

2026-09-07 13:42:57 +00:00

Error API: two variants are never constructed, converge() never returns Err, and MismatchedShape is misused for type errors

Item 1 is done in 8c087ad. Most of the rest turned out to be already resolved. Leaving this open for the one thing that isn't.

Item 1 — done

ConvergenceFailed and NegativePrecision…

logaritmisk closed issue logaritmisk/trueskill-tt#10

2026-09-07 13:42:41 +00:00

Member::with_prior is silently ignored for competitors the history already knows

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 13:42:41 +00:00

8c087ad015 fix!: apply competitor configuration whenever it is supplied

logaritmisk opened issue logaritmisk/trueskill-tt#41

2026-09-07 13:29:12 +00:00

erfc is a 1.2e-7 approximation and sets the accuracy floor for the whole crate

logaritmisk pushed to main at logaritmisk/trueskill-tt

2026-09-07 13:28:43 +00:00

7341669d1a fix: stop destroying tail precision in evidence and truncation

logaritmisk closed issue logaritmisk/trueskill-tt#40

2026-09-07 13:13:35 +00:00

History consumes its Observer with no way to read it back, so every observer needs interior mutability