Run T-0100 passed

2026-09-21 · Evidence: SPRT Sequential probability ratio test. Games are played until the accumulated evidence reaches one of two bounds; then the test stops.

Hypothesis

Incorporating game outcomes into the training target with a WDL ratio of 0.3 improves evaluation compared to relying purely on teacher scores.

Wording edited for publication from the German entry; the register carries the author’s own text.

Stages

Point estimate with the reported 95-percent half-width, per stage. The dashed line is zero. Estimates from a stopped sequential test are biased by the stopping rule; the interval is nominal.

Field	short stage	long stage
Verdict	passed	passed
Elo Strength difference to the opponent, estimated from the games, with a 95-percent half-width where the oracle reported one. Never an absolute rating.	+39.13 ± 18.05	+49.16 ± 19.68
nElo Normalised Elo: the Elo difference divided by the spread of the game results, so that the SPRT bounds mean the same at different time controls.	+54.79 ± 25.03	+72.82 ± 28.72
Games	740	562
Wins / draws / losses	270 / 283 / 187	210 / 221 / 131
pentanomial Games are played in pairs with swapped colours; the five counts are the pairs scoring 0, ½, 1, 1½ and 2 points.	`16 / 70 / 136 / 111 / 37`	`9 / 45 / 111 / 90 / 26`
LLR Log-likelihood ratio, the running evidence of an SPRT. It starts at 0; a stage passes at the upper bound and fails at the lower one.	+2.97	+2.99
Bounds	−2.94 … +2.94	−2.94 … +2.94
SPRT Sequential probability ratio test. Games are played until the accumulated evidence reaches one of two bounds; then the test stops.	[0.0, 10.0]	[0.0, 10.0]
Error rates α / β	0.05 / 0.05	0.05 / 0.05
Model	normalized	normalized
time control Base time plus increment per move in seconds, e.g. 8+0.08: eight seconds per game plus 0.08 seconds per move.	8+0.08	40+0.4
Book	A	B
Opponent	vs predecessor	vs predecessor
Duration	6 min	23 min

Provenance

Date: 2026-09-21
Started: 2026-09-21 15:29:19 UTC
Duration: 29 min
Author: Agent A
Epoch: 5
Gatekeeper: passed
Bench nodes: 3,627

Playing strength carried forward

3218 ± 71

An estimate, not a measured Elo. This build has not played a single listed opponent — anchored is 0.1.100 alone, and this figure is carried on from there. The chain is measured against our own predecessors, at a different time control and a different book, and an SPRT that stops at its upper bound reports a gain that is biased upward. The figure is therefore more likely too high than too low. Only the anchoring itself is a measurement.

How it is worked out

The anchoring measures Coherent 0.1.100, the build of run T-0092: 2989 ± 58, transferred over foreign engines with a published rating at 120+1.
The register carries the passed long stages after T-0092: +11.29 (T-0098) +168.60 (T-0099) +49.16 (T-0100), together +229.05, each measured against its own predecessor at 40+0.4.
2989, plus +229.05, gives 3218. The half-widths are combined in quadrature — the root of the sum of the squares: ± 58 and ± 40.96 make ± 71.