Coherent 0.1.105: 3129 ± 54 over five anchors
Listen to this post7 min
Created by AI from the text of this post and read by a synthetic voice; not a word-for-word version, and it may contain errors.
Loading…
Created by AI from the text of this post and read by a synthetic voice; not a word-for-word version, and it may contain errors.Position and speed stay in this browser and are sent nowhere.
The recording could not be played. Open the audio file
Project measurement, not in the register: an anchoring against foreign engines with a listed rating. The figures come from the data delivery (entry E-NNNN One Elo measurement: a build anchored against foreign engines with a listed rating, on a measuring machine of its own. Numbered in the order delivered; the register run it measured is named with it. E-0001 to E-0012; from the thirteenth entry on the prefix is M, the count running on.); the build measured is the one from run T-0108 of 22 September 2026. The reservations below are those delivered by the day of publication; the full and current account is on the playing-strength page.
Coherent 0.1.105 played 1,800 games against five foreign engines with a published rating, Mess 0.3.0, Dumb 2.1, Renegade 0.12.0, Koivisto 3.0 and Velvet 3.1.0, 360 against each at 120+1. The result is 3129 ± 54 (anchored to CCRL Computer Chess Rating Lists, a public ranking of chess engines from games under standard conditions. blitz ratings), ± 16 of it statistical. From 23 September 2026 that is the figure at the head of this site; it replaces the anchoring of 0.1.100 at 2989 ± 58.
The runs
| Opponent | Listed | Games | Score rate | Difference | From it |
|---|---|---|---|---|---|
| Mess 0.3.0 | 2773 ± 12 | 360 | 89.86 % | +379.03 ± 43.24 | 3152 ± 45 |
| Dumb 2.1 | 2828 ± 13 | 360 | 83.89 % | +286.63 ± 37.56 | 3115 ± 40 |
| Renegade 0.12.0 | 2935 ± 15 | 360 | 75.42 % | +194.73 ± 30.07 | 3130 ± 34 |
| Koivisto 3.0 | 3020 ± 16 | 360 | 66.67 % | +120.41 ± 28.87 | 3140 ± 33 |
| Velvet 3.1.0 | 3153 ± 16 | 360 | 44.44 % | −38.76 ± 25.34 | 3114 ± 30 |
All half-widths at the 95 % level, as the delivery states them; the column From it is the anchor plus the measured difference. The listed ratings are from the CCRL Blitz Rating List as of 2026-09-12.
How the figure is formed
Recalculated from the anchors: the five anchored values, averaged with weights inverse to their variance, give 3128.2 with a statistical half-width of ± 15.7; the delivery rounds it to ± 16. With the measuring side's transfer term the band is ± 53.4; the delivery names 3129 ± 54, and this site publishes the delivery's figure. The anchored values are compatible with one another (chi-square 2.87 at 4 degrees of freedom, bound 9.49; half-widths read as 95 % half-widths).
Against 0.1.100 one anchor was played in both campaigns, Renegade 0.12.0. The difference of the two measured differences is +139 ± 47. That is the measured change: the opponent's listed rating and the transfer term cancel in it. The 140 points between the two headline figures are not a measured gain; they carry both bands in full.
What belongs to this figure
- Measured outside the formal registration procedure. Every result line of this measurement path is recorded twice, independently, with its checksum.
- The calibration of the total band is not confirmed from the complete raw data.
- The evidence for the scale that this book maps one-to-one onto the list at 2+1 comes from a round robin with only three openings per pairing across ten colour-swapped rounds. The clustering is not accounted for in the calculation provided (intraclass correlation about 0.30, factor 1.93), so the evidence is about half as strong as reported. This is a reservation against the model assumption, not against the error calculation: the anchor runs themselves had a distinct opening for each colour pair (180 out of 180 per anchor), leaving the statistical half-width of 15.7 unaffected.
- Anchor reservation: 4 of 5 anchors (Mess 0.3.0, Dumb 2.1, Renegade 0.12.0, Koivisto 3.0) are more than 100 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. below the version measured.
- Statistical half-width plus transfer uncertainty. A round robin among the five anchors showed that an individual published rating deviates under external conditions; averaged across five anchors, ± 51 of this remains at the 95-per-cent level, which together with the ± 15.7 from games and list errors yields ± 54.
The anchor binaries are documented in the delivery as the authors' own builds: Mess 0.3.0 (Apache-2.0), Dumb 2.1 (MIT) and Renegade 0.12.0 (MIT). Without a provenance record in the delivery: Koivisto 3.0 and Velvet 3.1.0.
Conditions
120+1, book 8moves_v3 with 34,700 positions, each colour pair with an opening of its own, 64 MB hash, one thread, 30 concurrent games, no adjudication, a fixed number of games without SPRT Sequential probability ratio test. Games are played until the accumulated evidence reaches one of two bounds; then the test stops.; 0 losses on time. The measurement is dated 23 September 2026, 04:50 UTC.
Generated from the data delivery. The recalculation is the site's; every other figure is the measuring side's.