Classical Rating List — Original UCI Track — Accumulated Official Update
This page is the updated official Classical rating surface of the IJCCRL Arena Series for the Original UCI Track. The previous official 180-game league surface is preserved as the reference list, and the newly audited 270-game regular season has now been added to it. The accumulated rating corpus is therefore 450 regular-season games: 90 games per engine, with exact 45 White / 45 Black balance per engine. The rating scale remains Elo base 3500.
Rating method
- Official prior reference: the published 180-game Classical Original UCI Track rating list.
- New evidence: the audited 270-game regular season, 54 games per engine.
- Scale: Elo base 3500, with the final rounded list mean preserved at exactly 3500.
- Update rule: game-count weighting of the official prior surface and the new 270-game Ordo-style performance surface: 36 old games + 54 new games per engine.
- White advantage was explicitly estimated for the new block because White scored 209.5 / 270 (77.59%).
Current regular-season audit scope
- Track: Original UCI Track.
- Control: Classical 40m+2s.
- Format: 10-engine round robin with fixed mirrored openings and pair ratio 0.
- Closure: 270 / 270 games completed; each engine finished with 27 White games and 27 Black games.
- Scheduler closure: openingIndex 270, pending null.
- Current phase winner: Stockfish 18.
| # | Engine | Official Elo | Accumulated score | W-D-L | W/B | Next phase |
|---|---|---|---|---|---|---|
| 1 | Stockfish 18 Previous official label: Stockfish dev-20260218-a6d055d7 | 3658 Δ +47 | 60.5 / 90 current phase: 36.5 / 54 | 45-31-14 phase: 27-19-8 | 45 / 45 phase: 27 / 27 | Q#1 |
| 2 | Obsidian dev-16.15 Previous official label: Obsidian 16.0 | 3600 Δ +28 | 55.0 / 90 current phase: 33.0 / 54 | 38-34-18 phase: 23-20-11 | 45 / 45 phase: 27 / 27 | Q#2 |
| 3 | Reckless 0.10.0-dev-6abcff05 | 3580 Δ +44 | 52.5 / 90 current phase: 32.5 / 54 | 36-33-21 phase: 23-19-12 | 45 / 45 phase: 27 / 27 | Q#3 |
| 4 | PlentyChess 7.0.1 Previous official label: PlentyChess 7.0.55 | 3513 Δ -5 | 46.5 / 90 current phase: 27.5 / 54 | 31-31-28 phase: 17-21-16 | 45 / 45 phase: 27 / 27 | Q#5 |
| 5 | Berserk 14 Previous official label: Berserk 13 | 3511 Δ +29 | 45.5 / 90 current phase: 28.5 / 54 | 33-25-32 phase: 22-13-19 | 45 / 45 phase: 27 / 27 | Q#4 |
| 6 | Caissa 1.24 POPCNT Previous official label: Caissa 1.24 AVX2 | 3490 Δ -1 | 44.0 / 90 current phase: 26.5 / 54 | 27-34-29 phase: 16-21-17 | 45 / 45 phase: 27 / 27 | Q#6 |
| 7 | Dragon 3.3 by Komodo Chess 64-bit | 3479 Δ -12 | 43.0 / 90 current phase: 25.5 / 54 | 29-28-33 phase: 17-17-20 | 45 / 45 phase: 27 / 27 | Q#8 |
| 8 | Clover 9.1 | 3474 Δ +10 | 42.0 / 90 current phase: 26.0 / 54 | 24-36-30 phase: 15-22-17 | 45 / 45 phase: 27 / 27 | Q#7 |
| 9 | RubiChess 20240817 (popcnt) Previous official label: RubiChess 20240817 AVX2 | 3426 Δ -20 | 37.5 / 90 current phase: 22.5 / 54 | 18-39-33 phase: 10-25-19 | 45 / 45 phase: 27 / 27 | ELIMINATED |
| 10 | Integral v7.0.0 | 3269 Δ -120 | 23.5 / 90 current phase: 11.5 / 54 | 11-25-54 phase: 5-13-36 | 45 / 45 phase: 27 / 27 | ELIMINATED |
Regular Season Forensic Closure — AUDIT PASS
The audited package is accepted as a complete IJCCRL regular-season block. PGN, results JSON and scheduler state are coherent: 270 games were completed, no game is pending, no result is open, and the colour contract remains exact at engine, pairing and global level.
Integrity controls
| Control | Value | Status |
|---|---|---|
| PGN game count | 270 | PASS |
| results.json game count | 270 | PASS |
| PGN ↔ results.json White / Black / Result differences | 0 | PASS |
| scheduler_state.json closure | openingIndex 270; pending null | PASS |
| Engine colour balance | 27 White / 27 Black for every engine | PASS |
| Pairing colour balance | 3 White / 3 Black per engine in each pairing | PASS |
| Open or unfinished PGN results | 0 | PASS |
| Crash, illegal move, disconnect or readyok-timeout marker | 0 | PASS |
Forensic note on flags and White score
- The 20 flag terminations are concentrated on Integral v7.0.0 and remain valid competition results.
- No readyok-timeout, crash, illegal move, disconnect, aborted game or open result marker was detected.
- The large White score, 209.5 / 270, does not invalidate the event because every engine has exact colour balance and every pairing is mirrored.
- The rating surface preserves the explicit White-advantage correction already stated in the official Elo methodology above.
| Seed | Engine | Score | W-D-L | W/B | Result |
|---|---|---|---|---|---|
| 1 | Stockfish 18 | 36.5 / 54 | 27-19-8 | 27 / 27 | PHASE WINNER |
| 2 | Obsidian dev-16.15 | 33.0 / 54 | 23-20-11 | 27 / 27 | QUALIFIED |
| 3 | Reckless 0.10.0-dev-6abcff05 | 32.5 / 54 | 23-19-12 | 27 / 27 | QUALIFIED |
| 4 | Berserk 14 | 28.5 / 54 | 22-13-19 | 27 / 27 | QUALIFIED |
| 5 | PlentyChess 7.0.1 | 27.5 / 54 | 17-21-16 | 27 / 27 | QUALIFIED |
| 6 | Caissa 1.24 POPCNT | 26.5 / 54 | 16-21-17 | 27 / 27 | QUALIFIED |
| 7 | Clover 9.1 | 26.0 / 54 | 15-22-17 | 27 / 27 | QUALIFIED |
| 8 | Dragon 3.3 by Komodo Chess 64-bit | 25.5 / 54 | 17-17-20 | 27 / 27 | QUALIFIED |
| 9 | RubiChess 20240817 (popcnt) | 22.5 / 54 | 10-25-19 | 27 / 27 | ELIMINATED |
| 10 | Integral v7.0.0 | 11.5 / 54 | 5-13-36 | 27 / 27 | ELIMINATED |
Quarter-final Pairings by Regular-season Seeding
Q8 pairing doctrine follows the audited sporting seed order from the 270-game regular season: first seed against eighth seed, second against seventh, third against sixth and fourth against fifth.
Stockfish 18 vs Dragon 3.3 by Komodo Chess 64-bit
Regular-season scores: Stockfish 18 36.5 / 54; Dragon 3.3 25.5 / 54.
Obsidian dev-16.15 vs Clover 9.1
Regular-season scores: Obsidian 33.0 / 54; Clover 26.0 / 54.
Official Q8 result: Obsidian dev-16.15 wins 14.0–10.0. Match closed at 24 / 24 games, 12 mirrored units, 12W / 12B per engine, pair ratio 0. No tiebreak required.
Reckless 0.10.0-dev-6abcff05 vs Caissa 1.24 POPCNT
Regular-season scores: Reckless 32.5 / 54; Caissa 26.5 / 54.
Official Q8 result: Reckless 0.10.0-dev-6abcff05 wins 14.0–10.0. Match closed at 24 / 24 games, 12 mirrored units, 12W / 12B per engine, pair ratio 0. No tiebreak required.
PlentyChess 7.0.1 defeats Berserk 14
12.5–11.5
Regular-season scores: Berserk 28.5 / 54; PlentyChess 27.5 / 54.
Official Q8 result: PlentyChess 7.0.1 wins 12.5–11.5. Match closed at 24 / 24 games, 12 mirrored units, 12W / 12B per engine, pair ratio 0. No tiebreak required.
Quarter-final Audit Update — Q8_M2, Q8_M3 and Q8_M4
Three additional quarter-final matches have been audited under IJCCRL Rules+. All accepted packages are complete 24-game fixed mirrored opening matches: 12 mirrored units, exact 12 White / 12 Black balance for each engine, pair ratio 0, scheduler closed and no pending game state.
Obsidian dev-16.15 defeats Clover 9.1
14.0–10.0
Official closure: 24 / 24 games, 12 mirrored units, 12 White / 12 Black for each engine, pair ratio 0, scheduler closed with pending null. Obsidian finished with 12 wins, 4 draws and 8 losses; Clover finished with 8 wins, 4 draws and 12 losses.
Terminations: 20 checkmates, 3 fifty-move-rule draws and 1 threefold-repetition draw. Forensic note: the block had 20 White wins, 4 draws and 0 Black wins, but the result remains valid because the colour and mirrored-opening contract is exact.
Reckless 0.10.0-dev-6abcff05 defeats Caissa 1.24 POPCNT
14.0–10.0
Official closure: 24 / 24 games, 12 mirrored units, 12 White / 12 Black for each engine, pair ratio 0, scheduler closed with pending null. Reckless finished with 10 wins, 8 draws and 6 losses; Caissa finished with 6 wins, 8 draws and 10 losses.
Terminations: 16 checkmates, 4 threefold-repetition draws, 2 fifty-move-rule draws, 1 stalemate and 1 insufficient-material draw. Forensic note: the phase label still mentions Reckless 0.9.0, but the effective engine identity in the audit package is Reckless 0.10.0-dev-6abcff05.
PlentyChess 7.0.1 defeats Berserk 14
12.5–11.5
Official closure: 24 / 24 games, 12 mirrored units, 12 White / 12 Black for each engine, pair ratio 0, scheduler closed with pending null. PlentyChess finished with 9 wins, 7 draws and 8 losses; Berserk finished with 8 wins, 7 draws and 9 losses.
Terminations: 17 checkmates, 4 insufficient-material draws, 2 threefold-repetition draws and 1 fifty-move-rule draw. Forensic note: all decisive games were White wins, but the result remains valid because the colour and mirrored-opening contract is exact.
| Match | Engines | Official result | W-D-L winner | Structure | Verdict |
|---|---|---|---|---|---|
| Q8_M2 | Obsidian dev-16.15 vs Clover 9.1 #2 vs #7 | Obsidian 14.0–10.0 Clover | 12-4-8 | 24 games; 12 mirrored units; 12W / 12B per engine; pair ratio 0 | AUDIT PASS |
| Q8_M3 | Reckless 0.10.0-dev-6abcff05 vs Caissa 1.24 POPCNT #3 vs #6 | Reckless 14.0–10.0 Caissa | 10-8-6 | 24 games; 12 mirrored units; 12W / 12B per engine; pair ratio 0 | AUDIT PASS |
| Q8_M4 | Berserk 14 vs PlentyChess 7.0.1 #4 vs #5 | PlentyChess 12.5–11.5 Berserk | 9-7-8 | 24 games; 12 mirrored units; 12W / 12B per engine; pair ratio 0 | AUDIT PASS |
SF-M2 Forensic Closure — Obsidian dev-16.15 defeats Reckless 0.10.0-dev-6abcff05
The second semifinal has been audited under IJCCRL Rules+. The accepted package is a complete 48-game Classical 40m+2s fixed mirrored opening match: 24 mirrored units, exact 24 White / 24 Black balance for each engine, pair ratio 0, scheduler closed, no pending game state and no technical incident marker affecting the sporting result.
Official SF-M2 result: Obsidian dev-16.15 24.5–23.5 Reckless 0.10.0-dev-6abcff05
Obsidian advances to the Classical Original UCI final by the minimum possible match margin. This is a valid sporting result, but it should be published with a forensic note: the margin is narrow, 21 of 24 mirrored units were balanced, and all decisive games were White wins. The result remains accepted because the colour and mirrored-opening contract is exact.
Obsidian dev-16.15 defeats Reckless 0.10.0-dev-6abcff05
24.5–23.5
Official closure: 48 / 48 games, 24 mirrored units, 24 White / 24 Black for each engine, pair ratio 0, scheduler closed with pending null. Obsidian finished with 8 wins, 33 draws and 7 losses; Reckless finished with 7 wins, 33 draws and 8 losses.
Terminations: 22 threefold-repetition draws, 15 checkmates, 8 insufficient-material draws, 2 fifty-move-rule draws and 1 stalemate. No timeout, crash, illegal move, readyok-timeout, disconnect or aborted-game marker was detected in the accepted audit package.
Mirror-unit evidence
2–21–1
Across 24 mirrored units, Obsidian won 2 units, Reckless won 1 unit and 21 units were balanced at 1–1. The final +1 match point for Obsidian comes from that net one-unit advantage.
All 15 decisive games were White wins. This does not invalidate the match because each engine had exactly 24 games with White and 24 games with Black, and every mirrored unit was closed.
| Match | Engines | Official result | W-D-L winner | Structure | Verdict |
|---|---|---|---|---|---|
| SF-M2 | Obsidian dev-16.15 vs Reckless 0.10.0-dev-6abcff05 #2 vs #3 | Obsidian 24.5–23.5 Reckless | 8-33-7 | 48 games; 24 mirrored units; 24W / 24B per engine; pair ratio 0 | AUDIT PASS |
Final Forensic Closure — Stockfish 18 defeats Obsidian dev-16.15 — AUDIT PASS
The Classical Original UCI Track final has been normalized under IJCCRL Rules+ at the mirrored-unit level.
The raw event archive contains 64 scheduler records, but opening units #18 and #19 are excluded in full from
the official sporting corpus because three of their four records ended at UCI readiness with
readyok timeout (12000ms). Removing the complete affected units preserves pair ratio 0,
exact colour balance and the statistical identity of every accepted opening pair.
Classical Original UCI Track Champion: Stockfish 18
Official corrected result: Stockfish 18 35.0–25.0 Obsidian dev-16.15. Stockfish closes the accepted 60-game final with 28 wins, 14 draws and 18 losses. Obsidian closes with 18 wins, 14 draws and 28 losses. The ten-point margin is decisive and no tiebreak is required. The accumulated regular-season Elo table above remains unchanged because knockout evidence is appended for sporting closure and is not used as a rating input.
Stockfish 18
35.0
Accepted final record: 28 wins, 14 draws and 18 losses from 60 games. Stockfish scored 29.0 / 30 with White and 6.0 / 30 with Black. Across the 30 accepted mirrored units, Stockfish won 11 units, drew 18 units and lost 1 unit.
Obsidian dev-16.15
25.0
Accepted final record: 18 wins, 14 draws and 28 losses from 60 games. Obsidian scored 24.0 / 30 with White and 1.0 / 30 with Black. The runner-up result is preserved without any technical loss from the excluded UCI-readiness incidents.
Canonical correction
- Raw scheduler package: 64 records and 32 nominal opening units.
- Excluded unit #18: raw opening indexes 35–36; one played game plus one
proxy_fatalreadiness timeout. - Excluded unit #19: raw opening indexes 37–38; two
proxy_fatalreadiness timeouts. - The played companion game at raw index 35 is excluded with unit #18; it is not counted independently.
- Official canonical package: 60 games, 30 complete opening units and no accepted technical adjudication.
- The raw 64-record archive remains preserved as incident evidence; the 60-game ledger is the official sporting result.
Forensic result profile
- Accepted decisive games: 46, all won by White; accepted draws: 14.
- Terminations: 45 checkmates, 8 threefold repetitions, 3 fifty-move-rule draws, 3 insufficient-material draws and 1 valid flag.
- The strong White bias does not invalidate the final because both engines have exact 30 White / 30 Black balance and all accepted opening units are mirrored.
- Average accepted game length: 184.45 plies; shortest 109 plies; longest 417 plies.
| Place | Engine | Score | W-D-L | White / Black | Status |
|---|---|---|---|---|---|
| 1 | Stockfish 18 #1 seed | 35.0 / 60 | 28-14-18 | 30 / 30 | CHAMPION |
| 2 | Obsidian dev-16.15 #2 seed | 25.0 / 60 | 18-14-28 | 30 / 30 | RUNNER-UP |
| Control | Official value | Status |
|---|---|---|
| Canonical game count | 60 accepted games | PASS |
| Complete mirrored opening units | 30 / 30 accepted units | PASS |
| Engine colour balance | 30 White / 30 Black for each engine | PASS |
| Pair ratio | 0; every accepted unit contains both colour directions | PASS |
| PGN ↔ results JSON sporting data | 0 White / Black / Result mismatches in the accepted corpus | PASS |
| Open or unfinished accepted results | 0 | PASS |
| Crash, illegal move, disconnect or readyok timeout in accepted corpus | 0 | PASS |
| Administrative exclusion | Raw indexes 35–38 removed as complete units #18 and #19 | PASS |
| Final winner | Stockfish 18 by 10 match points | CHAMPION |
Current Regular Season — 270 / 270 Games Audited
The new regular-season block is accepted as a valid IJCCRL rating input. The event closed with 10 engines, 45 pairings, 6 games per pairing, exact colour balance and no pending scheduler state.
| Seed | Engine | Current score | W-D-L | W/B | Status |
|---|---|---|---|---|---|
| 1 | Stockfish 18 | 36.5 / 54 | 27-19-8 | 27 / 27 | QUALIFIED |
| 2 | Obsidian dev-16.15 | 33.0 / 54 | 23-20-11 | 27 / 27 | QUALIFIED |
| 3 | Reckless 0.10.0-dev-6abcff05 | 32.5 / 54 | 23-19-12 | 27 / 27 | QUALIFIED |
| 4 | Berserk 14 | 28.5 / 54 | 22-13-19 | 27 / 27 | QUALIFIED |
| 5 | PlentyChess 7.0.1 | 27.5 / 54 | 17-21-16 | 27 / 27 | QUALIFIED |
| 6 | Caissa 1.24 POPCNT | 26.5 / 54 | 16-21-17 | 27 / 27 | QUALIFIED |
| 7 | Clover 9.1 | 26.0 / 54 | 15-22-17 | 27 / 27 | QUALIFIED |
| 8 | Dragon 3.3 by Komodo Chess 64-bit | 25.5 / 54 | 17-17-20 | 27 / 27 | QUALIFIED |
| 9 | RubiChess 20240817 (popcnt) | 22.5 / 54 | 10-25-19 | 27 / 27 | ELIMINATED |
| 10 | Integral v7.0.0 | 11.5 / 54 | 5-13-36 | 27 / 27 | ELIMINATED |
Eight Engines Qualified for the Next Phase
The table below preserves the sporting seed order from the current 270-game regular season. This seed order is distinct from the updated accumulated Elo order.
| Seed | Engine | Current score | W-D-L | W/B | Status |
|---|---|---|---|---|---|
| #1 | Stockfish 18 Updated official Elo: 3658 (+47) | 36.5 / 54 | 27-19-8 | 27 / 27 | QUALIFIED |
| #2 | Obsidian dev-16.15 Updated official Elo: 3600 (+28) | 33.0 / 54 | 23-20-11 | 27 / 27 | QUALIFIED |
| #3 | Reckless 0.10.0-dev-6abcff05 Updated official Elo: 3580 (+44) | 32.5 / 54 | 23-19-12 | 27 / 27 | QUALIFIED |
| #4 | Berserk 14 Updated official Elo: 3511 (+29) | 28.5 / 54 | 22-13-19 | 27 / 27 | QUALIFIED |
| #5 | PlentyChess 7.0.1 Updated official Elo: 3513 (-5) | 27.5 / 54 | 17-21-16 | 27 / 27 | QUALIFIED |
| #6 | Caissa 1.24 POPCNT Updated official Elo: 3490 (-1) | 26.5 / 54 | 16-21-17 | 27 / 27 | QUALIFIED |
| #7 | Clover 9.1 Updated official Elo: 3474 (+10) | 26.0 / 54 | 15-22-17 | 27 / 27 | QUALIFIED |
| #8 | Dragon 3.3 by Komodo Chess 64-bit Updated official Elo: 3479 (-12) | 25.5 / 54 | 17-17-20 | 27 / 27 | QUALIFIED |
Official Elo Delta Ledger
The updated official Elo is calculated from the previous official Elo reference and the current regular-season performance surface. The calculation preserves the global 3500 base after integer rounding.
| Engine | Previous Elo | Current phase performance | Updated official Elo | Delta | Accumulated score |
|---|---|---|---|---|---|
Stockfish 18 previous label: Stockfish dev-20260218-a6d055d7 | 3611 | 3689.5 | 3658 | +47 | 60.5 / 90 |
Obsidian dev-16.15 previous label: Obsidian 16.0 | 3572 | 3619.1 | 3600 | +28 | 55.0 / 90 |
Reckless 0.10.0-dev-6abcff05 previous label: Reckless 0.10.0-dev-6abcff05 | 3536 | 3609.1 | 3580 | +44 | 52.5 / 90 |
PlentyChess 7.0.1 previous label: PlentyChess 7.0.55 | 3518 | 3509.9 | 3513 | -5 | 46.5 / 90 |
Berserk 14 previous label: Berserk 13 | 3482 | 3529.7 | 3511 | +29 | 45.5 / 90 |
Caissa 1.24 POPCNT previous label: Caissa 1.24 AVX2 | 3491 | 3490.2 | 3490 | -1 | 44.0 / 90 |
Dragon 3.3 by Komodo Chess 64-bit previous label: Dragon 3.3 by Komodo Chess 64-bit | 3491 | 3470.6 | 3479 | -12 | 43.0 / 90 |
Clover 9.1 previous label: Clover 9.1 | 3464 | 3480.4 | 3474 | +10 | 42.0 / 90 |
RubiChess 20240817 (popcnt) previous label: RubiChess 20240817 AVX2 | 3446 | 3411.9 | 3426 | -20 | 37.5 / 90 |
Integral v7.0.0 previous label: Integral v7.0.0 | 3389 | 3189.6 | 3269 | -120 | 23.5 / 90 |
Terminations in the 270-game block
| Termination | Games |
|---|---|
| checkmate | 155 |
| threefold repetition | 55 |
| flag | 20 |
| fifty move rule | 19 |
| insufficient material | 18 |
| stalemate | 3 |
Forensic note
- No readyok-timeout, crash, illegal move, disconnect or infrastructure adjudication marker was found in the audited 270-game block.
- The 20 flag results are preserved as valid competition results.
- The heavy White score is neutralized in the rating calculation through an explicit White-advantage parameter before the new phase is merged into the official list.
- The updated rating surface is an accumulated regular-season surface only.