logaritmisk
  • Joined on 2023-10-06
logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 23:28:54 +00:00
71554fd944 feat: add UnknownKeys::Prior, and explain why there is no Skip
logaritmisk closed issue logaritmisk/trueskill-tt#44 2026-09-07 23:28:54 +00:00
UnknownKey reports indices, not the key — and rejecting the whole matchup is easy to swallow into a plausible constant
logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 23:22:33 +00:00
2cf21a753d docs: record that the event log is the source of truth, and why
logaritmisk closed issue logaritmisk/trueskill-tt#45 2026-09-07 23:22:33 +00:00
Design question: can a fitted History be persisted, or must every consumer replay its whole event log?
logaritmisk commented on issue logaritmisk/trueskill-tt#45 2026-09-07 23:21:36 +00:00
Design question: can a fitted History be persisted, or must every consumer replay its whole event log?

Answering with measurements rather than opinion. Short version: your premise is right that a snapshot is delicate, but wrong about where the delicacy is — and Q1's answer is better than you…

logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 23:20:51 +00:00
35d7512557 fix(test): the ingestion-order property was comparing two truncated fits
logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 23:19:04 +00:00
e1bddf2474 style: factor the event-pair type out of the reconvergence fixture
logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 23:18:33 +00:00
5d36fa1008 test: pin that re-convergence is path-independent
logaritmisk commented on issue logaritmisk/trueskill-tt#44 2026-09-07 21:41:30 +00:00
UnknownKey reports indices, not the key — and rejecting the whole matchup is easy to swallow into a plausible constant

Asks 1 and 2 are done in c12bc83. Ask 3 is deliberately still open — see below.

Thank you for the write-up. "No exception, no log line, no failing test — only a parameter sweep that produced…

logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 21:41:11 +00:00
c12bc830a5 feat!: name the unknown key, expose tail probabilities, flag short fits
logaritmisk closed issue logaritmisk/trueskill-tt#43 2026-09-07 21:41:11 +00:00
No public way to ask "what is P(skill below x)?" — cdf is pub(crate), so callers approximate it with a z-multiplier
logaritmisk opened issue logaritmisk/trueskill-tt#50 2026-09-07 21:27:19 +00:00
ITERATIONS = 30 is far too low for a dense history, and overrunning it fails quietly
logaritmisk opened issue logaritmisk/trueskill-tt#49 2026-09-07 21:27:06 +00:00
expected_information_gain is discrete-outcome only, so scored consumers cannot ask "which matchup next?"
logaritmisk opened issue logaritmisk/trueskill-tt#48 2026-09-07 21:26:50 +00:00
Every predict_* is a ranked outcome — there is no predictive distribution for a SCORE
logaritmisk opened issue logaritmisk/trueskill-tt#47 2026-09-07 21:26:32 +00:00
A node's reported sigma is ~10x wider than the estimate's measured reproducibility
logaritmisk opened issue logaritmisk/trueskill-tt#46 2026-09-07 21:26:12 +00:00
No joint posterior: anything derived from two or more competitors has to add marginals and overstate its width
logaritmisk opened issue logaritmisk/trueskill-tt#45 2026-09-07 21:20:47 +00:00
Design question: can a fitted History be persisted, or must every consumer replay its whole event log?
logaritmisk opened issue logaritmisk/trueskill-tt#44 2026-09-07 21:20:27 +00:00
UnknownKey reports indices, not the key — and rejecting the whole matchup is easy to swallow into a plausible constant
logaritmisk opened issue logaritmisk/trueskill-tt#43 2026-09-07 21:20:05 +00:00
No public way to ask "what is P(skill below x)?" — cdf is pub(crate), so callers approximate it with a z-multiplier
logaritmisk pushed to main at logaritmisk/trueskill-tt 2026-09-07 20:47:03 +00:00
901f60972e chore: Release trueskill-tt version 0.4.2