Coherent 0.2.0: 3600 ± 52 over 8 anchors
Listen to this post8 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 M-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. The series continues the entries E-0001 to E-0012 and starts at M-0013.); the build measured is the one from run T-0153 of 2 October 2026. The reservations below are those delivered by the day of publication; the full and current account is on the playing-strength page.
Coherent 0.2.0 played 3,600 games against 8 foreign engines with a published rating, akimbo 1.0.0, Black Marlin 9.0, PlentyChess 0.3.0, Minke 6.0.0, PZChessBot 7.1, Stormphrax 8.0.0, Cinder 0.6.1 and Reckless 0.9.0, 450 against each at 120+1. The result is 3600 ± 52 (anchored to CCRL Computer Chess Rating Lists, a public ranking of chess engines from games under standard conditions. blitz ratings), ± 7 of it statistical. From 3 October 2026 that is the figure at the head of this site; it replaces the anchoring of 0.1.118 at 3455 ± 52.
The runs
Project measurement, not in the register. Eight horizontal intervals, one per opponent, on a scale of anchored playing strength: akimbo 1.0.0 at 3593 with a half-width of 17, Black Marlin 9.0 at 3609 with a half-width of 18, PlentyChess 0.3.0 at 3600 with a half-width of 19, Minke 6.0.0 at 3596 with a half-width of 18, PZChessBot 7.1 at 3605 with a half-width of 20, Stormphrax 8.0.0 at 3593 with a half-width of 21, Cinder 0.6.1 at 3620 with a half-width of 20, Reckless 0.9.0 at 3585 with a half-width of 21. Below them the published figure for 0.2.0, 3600 with a band of 52, shaded behind every row. The plus-minus values are half-widths of nominal 95 % intervals; the published band also carries the transfer term.
| Opponent | Listed | Games | Score rate | Difference | From it |
|---|---|---|---|---|---|
| akimbo 1.0.0 | 3567 ± 10 | 450 | 53.78 % | +26.30 ± 14.02 | 3593 ± 17 |
| Black Marlin 9.0 | 3590 ± 10 | 450 | 52.78 % | +19.32 ± 14.73 | 3609 ± 18 |
| PlentyChess 0.3.0 | 3592 ± 12 | 450 | 51.11 % | +7.72 ± 14.17 | 3600 ± 19 |
| Minke 6.0.0 | 3594 ± 13 | 450 | 50.33 % | +2.32 ± 12.93 | 3596 ± 18 |
| PZChessBot 7.1 | 3712 ± 12 | 450 | 35.11 % | −106.69 ± 15.59 | 3605 ± 20 |
| Stormphrax 8.0.0 | 3746 ± 12 | 450 | 29.33 % | −152.74 ± 16.95 | 3593 ± 21 |
| Cinder 0.6.1 | 3763 ± 13 | 450 | 30.56 % | −142.62 ± 15.84 | 3620 ± 20 |
| Reckless 0.9.0 | 3766 ± 11 | 450 | 26.11 % | −180.70 ± 18.35 | 3585 ± 21 |
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 2026-09-21, 2026-09-26.
How the figure is formed
Recalculated from the anchors: the 8 anchored values, averaged with weights inverse to their variance, give 3600.2 with a statistical half-width of ± 6.8; the delivery rounds it to ± 7. With the measuring side's transfer term the band is ± 51.4; the delivery names 3600 ± 52, and this site publishes the delivery's figure.
Against 0.1.118 no anchor was played in both campaigns; the 145 points between the two headline figures are therefore not a measured gain.
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 to the list at 2+1 comes from a round robin with only three openings per pairing in ten colour-swapped rounds. The clustering is not accounted for in the supplied calculation (intraclass correlation about 0.30, factor 1.93), so the evidence is about half as strong as reported. This is a reservation regarding the model assumption, not regarding the error calculation: Coherent's own anchoring runs had a separate opening for each colour pair (225 of 225 per anchor), the statistical half-width of 6.8 remains unaffected by this.
- Anchor reservation: 4 of 8 anchors (PZChessBot 7.1, Stormphrax 8.0.0, Cinder 0.6.1, Reckless 0.9.0) 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. above the version measured.
- Anchor reservation: scale by exception with anchor E=3670. Host offset since the upgrade of 30.09. is not measured, as the overlap was revoked on 03.10.2026. A new anchor set is in place, without a bridge to version 0.1.118.
- Statistical half-width plus transfer uncertainty: a given band with open calibration, without confirmed 95-percent coverage. A round robin among the five anchors showed that a single published rating diverges under unfamiliar conditions. Of that, ± 51 remains (1.96 times 36.9 divided by root 2, deliberately the value for two anchors); a shared error does not shrink with the number of anchors (here eight). Together with the ± 6.8 from games and list errors, this results in ± 52.
Without a provenance record in the delivery: akimbo 1.0.0, Black Marlin 9.0, PlentyChess 0.3.0, Minke 6.0.0, PZChessBot 7.1, Stormphrax 8.0.0, Cinder 0.6.1 and Reckless 0.9.0.
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 3 October 2026, 19:00 UTC.
Generated from the data delivery. The recalculation is the site's; every other figure is the measuring side's.