Stackruns

Coherent 0.1.108: 3274 ± 55 over five anchors

Published Updated chess-reporter Project: Coherent Chess anchoringmeasurement

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.

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-0122 of 24 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 0.1.108 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 3274 ± 55 (anchored to CCRL Computer Chess Rating Lists, a public ranking of chess engines from games under standard conditions. blitz ratings), ± 19 of it statistical. From 24 September 2026 that is the figure at the head of this site; it replaces the anchoring of 0.1.107 at 3282 ± 55.

The runs

Coherent 0.1.108: the anchors and the published figure — Project measurement, not in the register Project measurement, not in the register. Five horizontal intervals, one per opponent, on a scale of anchored playing strength: Mess 0.3.0 at 3275 with a half-width of 64, Dumb 2.1 at 3275 with a half-width of 50, Renegade 0.12.0 at 3269 with a half-width of 40, Koivisto 3.0 at 3278 with a half-width of 39, Velvet 3.1.0 at 3272 with a half-width of 31. Below them the published figure for 0.1.108, 3274 with a band of 55, shaded behind every row. The plus-minus values are half-widths of nominal 95 % intervals; the published band also carries the transfer term. 3200 3220 3240 3260 3280 3300 3320 3340 Mess 0.3.0 3275 ± 64 Dumb 2.1 3275 ± 50 Renegade 0.12.0 3269 ± 40 Koivisto 3.0 3278 ± 39 Velvet 3.1.0 3272 ± 31 0.1.108 published 3274 ± 55 playing strength anchored to CCRL blitz ratings
Each row is one opponent: the value anchored through it, with its 95 % half-width. The bottom row is the published figure for 0.1.108, 3274 ± 55; the shaded strip carries its band, transfer term included, across all rows. Project measurement, not in the register.
OpponentListedGamesScore rateDifferenceFrom it
Mess 0.3.02773 ± 1236094.72 %+501.60 ± 62.733275 ± 64
Dumb 2.12828 ± 1336092.92 %+447.14 ± 47.963275 ± 50
Renegade 0.12.02935 ± 1536087.22 %+333.67 ± 37.233269 ± 40
Koivisto 3.03020 ± 1636081.53 %+257.91 ± 35.073278 ± 39
Velvet 3.1.03153 ± 1636066.53 %+119.33 ± 26.213272 ± 31

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 3273.4 with a statistical half-width of ± 18.2; the delivery rounds it to ± 19. With the measuring side's transfer term the band is ± 54.2; the delivery names 3274 ± 55, and this site publishes the delivery's figure. The anchored values are compatible with one another (chi-square 0.12 at 4 degrees of freedom, bound 9.49; half-widths read as 95 % half-widths).

Against 0.1.107 five anchors were played in both campaigns, Mess 0.3.0, Dumb 2.1, Renegade 0.12.0, Koivisto 3.0 and Velvet 3.1.0. Per anchor the difference of the two measured differences is −25.78 ± 89.57, +33.72 ± 66.86, −18.00 ± 56.69, +4.76 ± 47.65 and −32.48 ± 37.46; combined, −12 ± 23. The per-anchor values are compatible with one another (chi-square 3.55 at 4 degrees of freedom, bound 9.49). That is the measured change: the opponent's listed rating and the transfer term cancel in it. The 8 points between the two headline figures are not a measured gain; they carry both bands in full.

Coherent 0.1.108 against 0.1.107: the change per shared anchor — Project measurement, not in the register Project measurement, not in the register. Five horizontal intervals, one per anchor played in both campaigns, on a scale of Elo difference against 0.1.107: Mess 0.3.0 at −26 with a half-width of 90, Dumb 2.1 at +34 with a half-width of 67, Renegade 0.12.0 at −18 with a half-width of 57, Koivisto 3.0 at +5 with a half-width of 48, Velvet 3.1.0 at −32 with a half-width of 37. Below them the combined value, −12 ± 23. A dashed vertical line marks zero, no change. −150 −100 −50 0 +50 +100 +150 Mess 0.3.0 −26 ± 90 Dumb 2.1 +34 ± 67 Renegade 0.12.0 −18 ± 57 Koivisto 3.0 +5 ± 48 Velvet 3.1.0 −32 ± 37 combined −12 ± 23 difference of the two measured differences against 0.1.107, in Elo
Per anchor played in both campaigns, the difference of the two measured differences against 0.1.107; the bottom row combines them to −12 ± 23. The dashed line is zero: no change. The opponent's listed rating and the transfer term cancel in each row. Project measurement, not in the register.

What belongs to this figure

  1. Measured outside the formal registration procedure. Every result line of this measurement path is recorded twice, independently, with its checksum.
  2. The calibration of the total band is not confirmed from the complete raw data.
  3. The evidence for the scale that this book maps one-to-one to the list at 2+1 comes from a round robin with only three openings per pairing across ten colour-swapped rounds. The grouping 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 lab's own anchoring runs had a separate opening for each colour pair (180 out of 180 per anchor), leaving the statistical half-width of 18.2 unaffected.
  4. Anchor reservation: 5 out of 5 anchors (Mess 0.3.0, Dumb 2.1, Renegade 0.12.0, Koivisto 3.0, Velvet 3.1.0) lie 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.
  5. 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-percent level, which together with the ± 18.2 from games and list errors yields ± 55.

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 24 September 2026, 10:00 UTC.

Generated from the data delivery. The recalculation is the site's; every other figure is the measuring side's.