Blitz 5+2 — Original UCI Track • Grand Final Audit Closed • Stockfish 18 Champion
The current IJCCRL Arena Series 2026 — Original UCI Track — Blitz 5m+2s cycle is now closed. The Regular Season remains forensically validated on 520 accepted games and the knockout phase has concluded with the Grand Final between Stockfish 18 and Obsidian dev-16.15. The raw final package closed at 600 games with exact 300 White / 300 Black games per engine. Five proxy_fatal records caused by readyok timeout (12000ms) contaminated four mirrored opening units; under IJCCRL Rules+ those four two-game units are removed in full. The official final corpus is therefore 592 valid games, 296 complete mirrored units, exact 296 White / 296 Black games per engine, pair ratio 0 and zero accepted proxy_fatal results. Stockfish 18 wins the canonical final 403.0–189.0 with a 282-242-68 W-D-L record and is the IJCCRL Arena Series 2026 — Blitz 5+2 — Original UCI Track champion. Stockfish author: The Stockfish developers. Historical knockout evidence from the previous completed cycle is retained further down the page as archive material.
Current Regular Season method note
- Raw scheduler package: 540 games, 10 engines, 108 games per engine and exact 54 White / 54 Black before audit correction.
- Invalid technical results: 6 proxy_fatal adjudications caused by readyok timeout (12000ms).
- Rules+ correction: remove complete opening-index cycles 61–70 and 221–230 rather than deleting isolated games.
- Official accepted corpus: 520 games, 260 complete mirrored units, 104 games per engine and exact 52 White / 52 Black.
- Pair ratio remains 0; no accepted mirrored unit is incomplete.
- Phase Elo is computed only from the 520 accepted games with draws scored as 0.5, using a standard logistic Elo / Bradley-Terry maximum-likelihood fit and constraining the ten-engine pool mean to IJCCRL Base 3700. Displayed ratings are rounded to the nearest Elo point.
Qualification and seeding note
- The first eight engines of the audited Regular Season qualify for Q8.
- Reckless and Obsidian are tied at 59.0 points; Berserk and PlentyChess are tied at 51.5 points.
- No new tiebreak rule is invented in this update. For Q8 seed numbering, the relative order already present in the source standings is retained: Reckless before Obsidian, and Berserk before PlentyChess.
- Q8 pairing rule: #1 vs #8, #2 vs #7, #3 vs #6 and #4 vs #5.
- Qualified: Stockfish, Reckless, Obsidian, Caissa, Dragon, Berserk, PlentyChess and Clover.
- Eliminated after the Regular Season: RubiChess and Integral.
- Q8-M1 audit closure: Stockfish 18 defeated Clover 9.1 by 136.0–64.0 over 200 valid games and advances.
- Q8-M2 is closed with Reckless 0.10.0-dev-6abcff05 advancing over PlentyChess 7.0.1. The sporting decision is retained here without inventing unavailable forensic detail.
- Q8-M3 audit closure: Obsidian dev-16.15 defeated Berserk 14 by 103.0–85.0 on the canonical 188-game corpus after removing six complete mirrored units affected by seven readyok-timeout proxy_fatal records.
- Q8-M3 canonical structure: 188 valid games, 94 complete mirrored units, exact 94 White / 94 Black per engine, pair ratio 0 and zero accepted proxy_fatal games.
- Q8-M4 audit closure: Caissa 1.24 POPCNT defeated Dragon 3.3 by Komodo Chess 64-bit by 103.5–94.5 on the canonical 198-game corpus after removing the complete mirrored A88 unit containing the single readyok-timeout proxy_fatal record.
- Q8-M4 canonical structure: 198 valid games, 99 complete mirrored units, exact 99 White / 99 Black per engine, pair ratio 0 and zero accepted proxy_fatal games.
- Q4 SF-M1 audit closure: Stockfish 18 defeated Caissa 1.24 POPCNT by 117.5–68.5 on the canonical 186-game corpus after removing seven complete mirrored units affected by nine readyok-timeout proxy_fatal records.
- SF-M1 canonical structure: 186 valid games, 93 complete mirrored units, exact 93 White / 93 Black per engine, pair ratio 0 and zero accepted proxy_fatal results. Stockfish 18 advances to the final.
- Q4 SF-M2 is closed with Obsidian dev-16.15 advancing over Reckless 0.10.0-dev-6abcff05. This page records the sporting consequence required to establish the audited final pairing without inventing unsupported SF-M2 forensic detail.
- Grand Final pairing: Stockfish 18 vs Obsidian dev-16.15. The final is audit-closed under IJCCRL Rules+ on 592 canonical games; Stockfish 18 wins 403.0–189.0 and is champion.
Official Regular Season Rating List — 520 Accepted Games
This is the current statistical surface used for the Regular Season closure and Q8 qualification. GP counts accepted games only. Every engine has 104 valid games with exact 52 White / 52 Black balance. The Elo column is recalculated exclusively from the accepted 520-game corpus on the Base 3700 scale; excluded proxy-fatal games do not contribute to score, W-D-L, GP or Elo.
| # | Engine | Phase Elo | Score | Valid GP | W-D-L | W/B | Status |
|---|---|---|---|---|---|---|---|
| 1 | Stockfish 18 Key: stockfish18 | 3871 | 77.0 / 104 74.04% | 104 | 52-50-2 | 52 / 52 | #1 • QUALIFIED |
| 2 | Reckless 0.10.0-dev-6abcff05 Key: reckless-windows-generic | 3748 | 59.0 / 104 56.73% | 104 | 33-52-19 | 52 / 52 | #2 • QUALIFIED |
| 3 | Obsidian dev-16.15 Key: Obsidian | 3746 | 59.0 / 104 56.73% | 104 | 32-54-18 | 52 / 52 | #3 • QUALIFIED |
| 4 | Caissa 1.24 POPCNT Key: caissa-1.24-x64-sse4-popcnt | 3733 | 57.5 / 104 55.29% | 104 | 31-53-20 | 52 / 52 | #4 • QUALIFIED |
| 5 | Dragon 3.3 by Komodo Chess 64-bit Key: dragon-3.3-64bit | 3701 | 52.0 / 104 50.00% | 104 | 20-64-20 | 52 / 52 | #5 • QUALIFIED |
| 6 | Berserk 14 Key: berserk-14-sse41 | 3698 | 51.5 / 104 49.52% | 104 | 29-45-30 | 52 / 52 | #6 • QUALIFIED |
| 7 | PlentyChess 7.0.1 Key: PlentyChess-7.0.0-windows-generic | 3699 | 51.5 / 104 49.52% | 104 | 26-51-27 | 52 / 52 | #7 • QUALIFIED |
| 8 | Clover 9.1 Key: Clover.9.1-old | 3694 | 50.0 / 104 48.08% | 104 | 23-54-27 | 52 / 52 | #8 • QUALIFIED |
| 9 | RubiChess 20240817 (popcnt) Key: RubiChess-20240817_x86-64-sse3-popcnt | 3640 | 42.5 / 104 40.87% | 104 | 13-59-32 | 52 / 52 | ELIMINATED |
| 10 | Integral v7.0.0 Key: integral_sse41_popcnt | 3471 | 20.0 / 104 19.23% | 104 | 5-30-69 | 52 / 52 | ELIMINATED |
Q8 Bracket Status — Current Knockout Phase
Q8 is generated directly from the audited Regular Season seed order: #1 vs #8, #2 vs #7, #3 vs #6 and #4 vs #5. All four Q8 matches are now closed. Q8-M1 is audit-closed with Stockfish 18 advancing over Clover 9.1 by 136.0–64.0 in 200 valid fixed mirrored games. Q8-M2 is closed with Reckless advancing over PlentyChess; no unsupported forensic score is added here. Q8-M3 is audit-closed after the canonical mirrored cut, with Obsidian defeating Berserk by 103.0–85.0 over 188 valid games and advancing. Q8-M4 is audit-closed after removing one complete mirrored unit affected by a readyok-timeout proxy_fatal record; Caissa defeats Dragon by 103.5–94.5 over the canonical 198-game corpus and advances. The resulting Q4 field is Stockfish, Reckless, Obsidian and Caissa. Where the Regular Season audit cut created equal point totals, the source standings order is retained solely to assign the seed number; no additional tiebreak formula is introduced here.
| Q8 Match | Seed A | Seed B | Status |
|---|---|---|---|
| Q8-M1 | #1 Stockfish 18 Regular Season: 77.0/104 • Phase Elo 3871 | #8 Clover 9.1 Regular Season: 50.0/104 • Phase Elo 3694 | AUDIT PASS Stockfish wins 136.0–64.0. 200/200 valid • Stockfish advances. |
| Q8-M2 | #2 Reckless 0.10.0-dev-6abcff05 59.0/104 • Phase Elo 3748 | #7 PlentyChess 7.0.1 51.5/104 • Phase Elo 3699 | CLOSED Reckless advances. Sporting result retained; no unsupported forensic score added in this update. |
| Q8-M3 | #3 Obsidian dev-16.15 59.0/104 • Phase Elo 3746 | #6 Berserk 14 51.5/104 • Phase Elo 3698 | AUDIT PASS Obsidian wins 103.0–85.0. 188 canonical valid games • 94 mirrored units • Obsidian advances. |
| Q8-M4 | #4 Caissa 1.24 POPCNT 57.5/104 • Phase Elo 3733 | #5 Dragon 3.3 by Komodo Chess 64-bit 52.0/104 • Phase Elo 3701 | AUDIT PASS Caissa wins 103.5–94.5. 198 canonical valid games • 99 mirrored units • Caissa advances. |
Q4 Semifinals — Closed • Finalists Confirmed
The Q8 phase and both Q4 semifinals are closed. SF-M1 — Stockfish 18 vs Caissa 1.24 POPCNT — is audit-closed under IJCCRL Rules+ on 186 canonical games, 93 complete mirrored units, exact 93 White / 93 Black per engine and pair ratio 0; Stockfish 18 wins 117.5–68.5 and advances. SF-M2 — Reckless 0.10.0-dev-6abcff05 vs Obsidian dev-16.15 — is closed with Obsidian advancing to the final. No unsupported SF-M2 forensic score is invented in this update. The resulting Grand Final pairing is Stockfish 18 vs Obsidian dev-16.15, and that final is now audit-closed with Stockfish 18 champion by 403.0–189.0 on 592 canonical games.
| Q4 Match | Engine A | Engine B | Status |
|---|---|---|---|
| SF-M1 | #1 Stockfish 18 Regular Season winner • Q8-M1 winner | #4 Caissa 1.24 POPCNT Q8-M4 winner • 103.5–94.5 vs Dragon | AUDIT PASS Stockfish wins 117.5–68.5. 186 canonical games • 93 mirrored units • Stockfish advances to the final. |
| SF-M2 | #2 Reckless 0.10.0-dev-6abcff05 Q8-M2 winner | #3 Obsidian dev-16.15 Q8-M3 winner • 103.0–85.0 vs Berserk | CLOSED Obsidian advances to the final. Sporting decision retained here; no unsupported SF-M2 forensic score added. |
Q4 Audit Update — SF-M1 — Stockfish 18 vs Caissa 1.24 POPCNT
SF-M1 is accepted under IJCCRL Rules+ after the minimum canonical mirrored correction. The raw scheduler closed at 200 games with exact 100 White / 100 Black per engine, openingIndex 200 and pending null. Nine proxy_fatal records were identified, all caused by readyok timeout (12000ms). They affect seven mirrored opening units. IJCCRL publication doctrine removes every affected two-game unit in full rather than preserving isolated infrastructure forfeits. Fourteen games are therefore excluded, producing the official 186-game corpus. Stockfish 18 wins 117.5–68.5 and advances to the final.
Official SF-M1 result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Stockfish 18 Key: stockfish18 | 117.5 / 186 63.17% | 70-95-21 | 93 / 93 |
Caissa 1.24 POPCNT Key: caissa-1.24-x64-sse4-popcnt | 68.5 / 186 36.83% | 21-95-70 | 93 / 93 |
Rules+ audit controls
- Audit verdict: AUDIT PASS after canonical mirrored cut.
- Match: Q4 SF-M1 — Regular Season #1 Stockfish 18 vs #4 Caissa 1.24 POPCNT.
- Format: Blitz 5+2, fixed mirrored opening match.
- Raw scheduler closure: 200 games, openingIndex 200 and pending null.
- Raw colour balance: exact 100 White and 100 Black games per engine.
- Technical incidents: 9 proxy_fatal results, all readyok timeout (12000ms).
- Affected mirrored units: 7.
- Excluded games: 14 total — 9 direct proxy_fatal records plus 5 mirror counterparts removed solely to preserve complete units.
- Accepted corpus: 186 games.
- Accepted colour balance: exact 93 White and 93 Black games per engine.
- Mirrored structure: 93/93 complete two-game units.
- Pair ratio: 0.
- Accepted proxy_fatal results: 0.
- Tiebreak required: no.
- Official decision: Stockfish 18 advances to the final.
Canonical result structure
- Checkmate: 88 games.
- Threefold repetition: 50 games.
- Insufficient material: 25 games.
- 50-move rule: 16 games.
- Stalemate: 4 games.
- Flag: 3 games.
- Proxy_fatal accepted: 0.
- The three sporting flag results remain valid because they followed competitive play and are not infrastructure/proxy failures.
Forensic evidence
- Affected OpeningIndex units removed: 7–8, 9–10, 93–94, 101–102, 103–104, 139–140 and 141–142.
- PGN and results.json contain 200 aligned raw records; scheduler_state.json confirms 200 games, 100 White / 100 Black per engine and pending null.
- 100 opening lines are present in the raw match as two-game colour-reversed units.
- games PGN SHA256: 5A509D3BB0DBBA6C0F6C0335C34B67E5D863B585AB0E4138A24D948D503B73CC.
- results.json SHA256: 5D255DCA5AE0F86CAB5E176C4C5F38A72096017D41977365475691D631596340.
- scheduler_state.json SHA256: 103DA9CB36BE606B0121425AFC39214D283344C571E593FB5532A3827EBB86BE.
- Official knockout record: Stockfish 18 advances over Caissa 1.24 POPCNT by 117.5–68.5.
Q8 Audit Update — Q8-M1 — Stockfish 18 vs Clover 9.1
Q8-M1 is accepted as a complete IJCCRL Rules+ fixed mirrored match. The audited PGN, results and scheduler-state package contains 200/200 valid games, exact 100 White / 100 Black colour balance for both engines, 100 closed mirrored opening units, 100 distinct opening lines and pair ratio 0. No crash, timeout/flag forfeit, illegal move, disconnect, aborted game, engine error or external/manual adjudication was found. Stockfish 18 wins 136.0–64.0 and advances to the next knockout round.
Official Q8-M1 result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Stockfish 18 Key: stockfish18 | 136.0 / 200 68.00% | 87-98-15 | 100 / 100 |
Clover 9.1 Key: Clover.9.1-old | 64.0 / 200 32.00% | 15-98-87 | 100 / 100 |
Rules+ audit controls
- Audit verdict: AUDIT PASS.
- Match: Q8-M1 — Regular Season #1 Stockfish 18 vs #8 Clover 9.1.
- Format: Blitz 5+2, fixed mirrored opening match.
- Games: 200/200 valid; no forensic cut required.
- Scheduler closure: openingIndex 200, final pairGameNo 2 and pending null.
- Colour balance: exact 100 White and 100 Black games per engine.
- Mirrored structure: 100/100 two-game opening units closed.
- Opening integrity: 100 distinct lines, each used exactly twice with colours reversed.
- Pair ratio: 0.
- PGN / results.json / scheduler_state.json parity: verified across all 200 games.
- Tiebreak required: no.
- Official decision: Stockfish 18 advances.
Result and termination structure
- Decisive games: 102; drawn games: 98.
- Checkmate: 102 games.
- Insufficient material: 39 games.
- Threefold repetition: 37 games.
- 50-move rule: 20 games.
- Stalemate: 2 games.
- All 102 decisive games ended by checkmate.
- No incomplete mirrored unit and no duplicated match ID were found.
Colour and technical notes
- White wins: 99; Black wins: 3; draws: 98.
- Aggregate White score: 148.0/200 (74.00%).
- Stockfish by colour: 92.0/100 as White and 44.0/100 as Black.
- Clover by colour: 56.0/100 as White and 8.0/100 as Black.
- The strong White bias does not invalidate the match because every one of the 100 opening lines is played with reversed colours.
- TB reporting is present for both engines in all games and TBHitsWhite + TBHitsBlack = TBHitsTotal throughout the package.
- No crash, timeout/flag forfeit, readyok timeout, illegal move, disconnect, aborted game, engine error or external/manual adjudication was found.
Q8 Audit Update — Q8-M3 — Obsidian dev-16.15 vs Berserk 14
Q8-M3 is accepted under IJCCRL Rules+ after the minimum canonical mirrored correction. The raw scheduler closed at 200 games with exact 100 White / 100 Black per engine, openingIndex 200 and pending null, but seven proxy_fatal records were identified, all caused by readyok timeout (12000ms) and all ending at 16 plies, exactly at the end of the forced opening segment. Because IJCCRL publication doctrine does not preserve an isolated technical forfeit inside a mirrored opening unit, every affected two-game unit is removed in full. Six mirrored units are therefore excluded, producing the official 188-game corpus. Obsidian wins 103.0–85.0 and advances to the next knockout round.
Official Q8-M3 result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Obsidian dev-16.15 Key: Obsidian | 103.0 / 188 54.79% | 53-100-35 | 94 / 94 |
Berserk 14 Key: berserk-14-sse41 | 85.0 / 188 45.21% | 35-100-53 | 94 / 94 |
Rules+ audit controls
- Audit verdict: AUDIT PASS after canonical mirrored cut.
- Match: Q8-M3 — Regular Season #3 Obsidian dev-16.15 vs #6 Berserk 14.
- Format: Blitz 5+2, fixed mirrored opening match.
- Raw scheduler closure: 200 games, openingIndex 200 and pending null.
- Raw colour balance: exact 100 White and 100 Black games per engine.
- Technical incidents: 7 proxy_fatal results, all readyok timeout (12000ms), all at 16 plies.
- Affected mirrored units: 6.
- Excluded games: 12 total — the 7 proxy_fatal records plus 5 otherwise playable mirror counterparts.
- Accepted corpus: 188 games.
- Accepted colour balance: exact 94 White and 94 Black games per engine.
- Mirrored structure: 94/94 complete two-game units.
- Pair ratio: 0.
- Accepted proxy_fatal results: 0.
- Tiebreak required: no.
- Official decision: Obsidian dev-16.15 advances.
Canonical result structure
- Checkmate: 81 games.
- Threefold repetition: 51 games.
- Insufficient material: 36 games.
- 50-move rule: 9 games.
- Flag: 7 games.
- Stalemate: 4 games.
- Proxy_fatal accepted: 0.
- The seven sporting flag results remain valid because they followed normal competitive play and are not infrastructure/proxy failures.
Forensic and publication notes
- The 200 PGN records and 200 JSON records remain structurally aligned in the raw package.
- The 100 opening lines occur as colour-reversed two-game units in the raw match.
- All seven technical failures occur at the same readyok boundary and cannot be treated as normal played losses.
- Removing complete affected units preserves colour equality and the IJCCRL pair-ratio doctrine.
- Canonical score difference: +18.0 points for Obsidian.
- Official knockout record: Obsidian dev-16.15 advances over Berserk 14 by 103.0–85.0.
Q8 Audit Update — Q8-M4 — Caissa 1.24 POPCNT vs Dragon 3.3
Q8-M4 is accepted under IJCCRL Rules+ after the minimum mirrored correction. The raw scheduler closed at 200 games with exact 100 White / 100 Black per engine, openingIndex 200 and pending null. One technical record was identified at OpeningIndex 29: a proxy_fatal caused by a readyok timeout (12000ms) on the Caissa side after 16 plies. Under the IJCCRL pair-ratio doctrine, its otherwise playable colour-reversed counterpart at OpeningIndex 30 is removed with it because both games belong to the same A88 mirrored unit. The official corpus is therefore 198 games, 99 complete mirrored units, exact 99 White / 99 Black per engine and pair ratio 0. Caissa wins 103.5–94.5 and advances to Q4.
Official Q8-M4 result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Caissa 1.24 POPCNT Key: caissa-1.24-x64-sse4-popcnt | 103.5 / 198 52.27% | 47-113-38 | 99 / 99 |
Dragon 3.3 by Komodo Chess 64-bit Key: dragon-3.3-64bit | 94.5 / 198 47.73% | 38-113-47 | 99 / 99 |
Rules+ audit controls
- Audit verdict: AUDIT PASS after canonical mirrored cut.
- Match: Q8-M4 — Regular Season #4 Caissa 1.24 POPCNT vs #5 Dragon 3.3.
- Format: Blitz 5+2, fixed mirrored opening match.
- Raw scheduler closure: 200 games, openingIndex 200 and pending null.
- Raw colour balance: exact 100 White and 100 Black games per engine.
- Technical incidents: 1 proxy_fatal result caused by readyok timeout (12000ms), at OpeningIndex 29 and 16 plies.
- Affected mirrored unit: A88, OpeningIndex 29–30.
- Excluded games: 2 total — the proxy_fatal record plus its mirror counterpart.
- Accepted corpus: 198 games.
- Accepted colour balance: exact 99 White and 99 Black games per engine.
- Mirrored structure: 99/99 complete two-game units.
- Pair ratio: 0.
- Accepted proxy_fatal results: 0.
- Tiebreak required: no.
- Official decision: Caissa 1.24 POPCNT advances to Q4.
Canonical termination structure
- Checkmate: 85 games.
- Threefold repetition: 98 games.
- Insufficient material: 8 games.
- 50-move rule: 6 games.
- Stalemate: 1 game.
- Proxy_fatal accepted: 0.
- Total TB hits in the accepted corpus: 36,002,152.
Forensic and publication notes
- The raw PGN and results package contain 200 aligned records; the scheduler closes on the second game of the final mirrored unit.
- OpeningIndex 29 is excluded because it is an infrastructure/proxy failure rather than a played chess result.
- OpeningIndex 30 is excluded solely to preserve the complete mirrored A88 unit and exact colour equality.
- Removing the complete affected unit preserves pair ratio 0 and changes the raw 104.0–96.0 score to the canonical 103.5–94.5 score.
- The identity of the winner is unchanged by the correction.
- Official knockout record: Caissa 1.24 POPCNT advances over Dragon 3.3 by 103.5–94.5.
Grand Final Audit Update — Stockfish 18 vs Obsidian dev-16.15
The Grand Final is accepted under IJCCRL Rules+ after the minimum canonical mirrored correction. The raw scheduler package closed at 600 games, 300 mirrored opening units, exact 300 White / 300 Black games per engine, openingIndex 600 and pending null. Five proxy_fatal records were identified, all caused by readyok timeout (12000ms). Those incidents affect four mirrored units. IJCCRL publication doctrine removes each contaminated two-game unit in full, so eight games are excluded from the sporting corpus. The official final therefore contains 592 valid games, 296 complete mirrored units, exact 296 White / 296 Black games per engine, pair ratio 0 and zero accepted proxy_fatal results. Stockfish 18 wins 403.0–189.0 and is the IJCCRL Arena Series 2026 — Blitz 5+2 — Original UCI Track champion.
Official Grand Final result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Stockfish 18 Key: stockfish18 Author: The Stockfish developers | 403.0 / 592 68.07% | 282-242-68 | 296 / 296 |
Obsidian dev-16.15 Key: Obsidian | 189.0 / 592 31.93% | 68-242-282 | 296 / 296 |
Rules+ audit controls
- Audit verdict: AUDIT PASS after canonical mirrored cut.
- Match: Grand Final — Stockfish 18 vs Obsidian dev-16.15.
- Format: Blitz 5+2, fixed mirrored opening match.
- Raw scheduler closure: 600 games, 300 mirrored units, openingIndex 600 and pending null.
- Raw colour balance: exact 300 White and 300 Black games per engine.
- Technical incidents: 5 proxy_fatal results, all readyok timeout (12000ms).
- Affected mirrored units: 4.
- Excluded games: 8 total — five direct technical records plus their required mirror counterparts where necessary to preserve complete units.
- Accepted corpus: 592 games.
- Accepted colour balance: exact 296 White and 296 Black games per engine.
- Mirrored structure: 296/296 complete two-game units.
- Pair ratio: 0.
- Accepted proxy_fatal results: 0.
- Tiebreak required: no.
- Official decision: Stockfish 18 is champion.
Canonical result structure
- Checkmate: 270 games.
- Threefold repetition: 153 games.
- Flag: 80 games.
- Insufficient material: 48 games.
- 50-move rule: 29 games.
- Stalemate: 12 games.
- Proxy_fatal accepted: 0.
- The 80 sporting flag results remain valid because they followed competitive play and are not infrastructure/proxy failures.
Forensic evidence and champion record
- Affected complete mirrored ranges removed: OpeningIndex 21–22, 23–24, 199–200 and 251–252.
- The raw package contains 600 unique game IDs and closes with exact 300 / 300 colour balance per engine.
- After the mirrored cut, all 296 accepted opening units remain complete with colours reversed and pair ratio 0.
- PGN / results.json / scheduler_state.json parity was verified across the canonical corpus.
- Raw score: Stockfish 407.5–192.5 Obsidian.
- Canonical score after the Rules+ mirrored cut: Stockfish 403.0–189.0 Obsidian.
- Stockfish 18 canonical W-D-L: 282-242-68.
- Obsidian dev-16.15 canonical W-D-L: 68-242-282.
- Stockfish author: The Stockfish developers.
- games.pgn SHA256: 4000831BFCF795A8741788307FF18BD6238C827126F97DCBAEBDFF33774C72C5.
- results.json SHA256: F6C3562F6444B6161163686502FEAA7BCA1D8E53946BE0744F6043DA12E67EC6.
- scheduler_state.json SHA256: 67A5855CB974B46BEE8D41AA24BA11CB46F69569502C4F328655EC68AAA18127.
- Official champion: Stockfish 18.
Historical Knockout Archive — Previous Completed Cycle
The detailed semifinal, final and earlier knockout audit blocks below are retained as historical evidence from the previous completed Original UCI Blitz cycle. They are not part of the current 520-game Regular Season score, the completed current Q8 bracket or the active Q4 semifinal phase above.
Semifinal Pairing Map — Four Classified Engines
The semifinal phase is fully audit-closed and the final is now audit-closed. The field was ordered by the accumulated classification: #1 Stockfish, #2 Reckless, #3 Obsidian and #4 Alexandria. SF-M1 closed with Stockfish defeating Alexandria by 30.0–20.0. SF-M2 closed with Reckless defeating Obsidian by 29.5–20.5. The final closed with Stockfish dev-20260426-1a882efc defeating Reckless 0.10.0-dev-8ba888c6 by 55.5–44.5 over 100 clean fixed mirrored games.
| Semifinal | Classified seed A | Classified seed B | Source / status |
|---|---|---|---|
| SF-M1 | #1 Stockfish dev-20260426-1a882efc First classified • accumulated Elo 3864 • audit result 30.0 / 50 | #4 Alexandria-9.0.9 Fourth classified • accumulated Elo 3770 • audit result 20.0 / 50 | AUDIT PASS Stockfish wins 30.0–20.0 and advances to the final. |
| SF-M2 | #2 Reckless 0.10.0-dev-8ba888c6 Second classified • accumulated Elo 3834 • semifinal result 29.5 / 50 | #3 Obsidian dev-16.15 Third classified • accumulated Elo 3787 • semifinal result 20.5 / 50 | AUDIT PASS Reckless wins 29.5–20.5 and advances to the final. |
Semifinal Audit Update — SF-M1 — Stockfish vs Alexandria
SF-M1 is accepted as a clean IJCCRL fixed mirrored semifinal match. The audited package contains 50 games, exact 25 White / 25 Black colour balance for both engines, openingIndex 50, pending null and a clean watchdog stop at the 50-game target. The result is appended as semifinal evidence only; it remains historical evidence from the previous 1,056-game accumulated regular-season surface and does not alter the current 520-game Regular Season table above.
Official result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Stockfish dev-20260426-1a882efc Key: stockfish18-avx2 | 30.0 / 50 60.00% | 23-14-13 | 25 / 25 |
Alexandria-9.0.9 Key: alexandria-avx2 | 20.0 / 50 40.00% | 13-14-23 | 25 / 25 |
Audit controls
- Audit verdict: AUDIT PASS.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — Semifinal SF-M1 — Stockfish vs Alexandria.
- Format: Blitz 5+2, fixed mirrored match.
- Games: 50/50 closed.
- Scheduler state: openingIndex 50 and pending null.
- Colour balance: exact 25 White and 25 Black games per engine.
- Mirrored structure: 25/25 UHO opening units closed.
- Pair ratio: 0.
- Opening authority: UHO_2024_8mvs_big_+100_+119.pgn, lines #1–#25.
- PGN/results/scheduler/watchdog consistency: verified.
- Tiebreak required: no.
- Official semifinal decision: Stockfish advances to the final.
Result structure
- PGN result distribution: 36 wins for White, 14 draws and 0 wins for Black.
- Colour score: White 43.0/50, Black 7.0/50.
- Stockfish by colour: 24.0/25 as White and 6.0/25 as Black.
- Alexandria by colour: 19.0/25 as White and 1.0/25 as Black.
- Mirrored-unit distribution: 11 units won by Stockfish, 1 unit won by Alexandria and 13 units split 1.0–1.0.
- No incomplete mirrored unit was found.
Technical notes
- Termination distribution: 36 checkmates, 6 threefold repetitions, 4 stalemates, 2 games by the 50-move rule and 2 insufficient-material results.
- Total recorded TB hits: 10,373,896.
- Telemetry note: TB hits are recorded on the Stockfish side in the audited package; Alexandria reports 0 TB hits in the same records.
- This TB telemetry difference is documented as a reporting/configuration caveat and does not affect the match result.
- Official knockout record: Stockfish wins SF-M1 by 30.0–20.0 and reaches the final.
Semifinal Audit Update — SF-M2 — Reckless vs Obsidian
SF-M2 is accepted as a clean IJCCRL fixed mirrored semifinal match. The audited package contains 50 games, exact 25 White / 25 Black colour balance for both engines, openingIndex 50, pending null and 25 closed UHO mirrored opening units. The result is appended as semifinal evidence only; it remains historical evidence from the previous 1,056-game accumulated regular-season surface and does not alter the current 520-game Regular Season table above.
Official result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Reckless 0.10.0-dev-8ba888c6 Key: reckless-avx2 | 29.5 / 50 59.00% | 24-11-15 | 25 / 25 |
Obsidian dev-16.15 Key: obsidian-avx2 | 20.5 / 50 41.00% | 15-11-24 | 25 / 25 |
Audit controls
- Audit verdict: AUDIT PASS.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — Semifinal SF-M2 — Reckless vs Obsidian.
- Format: Blitz 5+2, fixed mirrored match.
- Games: 50/50 closed.
- Scheduler state: openingIndex 50 and pending null.
- Colour balance: exact 25 White and 25 Black games per engine.
- Mirrored structure: 25/25 UHO opening units closed.
- Pair ratio: 0.
- PGN/results/scheduler consistency: verified.
- Tiebreak required: no.
- Official semifinal decision: Reckless advances to the final.
Result structure
- PGN result distribution: 39 wins for White, 11 draws and 0 wins for Black.
- Colour score: White 44.5/50, Black 5.5/50.
- Reckless by colour: 24.5/25 as White and 5.0/25 as Black.
- Obsidian by colour: 20.0/25 as White and 0.5/25 as Black.
- Mirrored-unit structure: 25 closed two-game opening units.
- No incomplete mirrored unit was found.
Technical notes
- Termination distribution: 39 checkmates, 4 insufficient-material results, 3 games by the 50-move rule, 3 threefold repetitions and 1 stalemate.
- No timeout, readyok-timeout, crash, illegal move, forfeit, disconnect or engine-error evidence was found in the audited package.
- Official knockout record: Reckless wins SF-M2 by 29.5–20.5 and reaches the final.
- The final pairing is therefore Stockfish dev-20260426-1a882efc vs Reckless 0.10.0-dev-8ba888c6.
Final Audit Update — Stockfish vs Reckless
The Original UCI Track Blitz final is accepted as a clean IJCCRL fixed mirrored final. The audited package contains 100 games, exact 50 White / 50 Black colour balance for both engines, openingIndex 100, pending null and 50 closed UHO mirrored opening units. The final closed without timeout, readyok-timeout, forfeit, crash, illegal move, engine-error or exception evidence. Stockfish dev-20260426-1a882efc is the IJCCRL Arena Series 2026 — Blitz 5+2 — Original UCI Track champion.
Official final result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Stockfish dev-20260426-1a882efc Key: stockfish18-avx2 | 55.5 / 100 55.50% | 47-17-36 | 50 / 50 |
Reckless 0.10.0-dev-8ba888c6 Key: reckless-avx2 | 44.5 / 100 44.50% | 36-17-47 | 50 / 50 |
Final audit controls
- Audit verdict: AUDIT PASS.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — Blitz 5+2 — Final — Stockfish vs Reckless.
- Format: fixed mirrored final.
- Games: 100/100 closed.
- Scheduler state: openingIndex 100 and pending null.
- Colour balance: exact 50 White and 50 Black games per engine.
- Mirrored structure: 50/50 UHO opening units closed.
- Pair ratio: 0.
- Time control: 300+2.
- PGN/results/scheduler consistency: verified.
- Tiebreak required: no.
- Official final decision: Stockfish is champion.
Result structure
- PGN result distribution: 83 wins for White, 17 draws and 0 wins for Black.
- Colour score: White 91.5/100, Black 8.5/100.
- Stockfish by colour: 48.5/50 as White and 7.0/50 as Black.
- Reckless by colour: 43.0/50 as White and 1.5/50 as Black.
- Mirrored-unit distribution: 37 units split 1.0–1.0, 12 units won by Stockfish 1.5–0.5 and 1 unit won by Reckless 1.5–0.5.
- No incomplete mirrored unit was found.
Technical notes
- Termination distribution: 83 checkmates, 7 threefold repetitions, 7 insufficient-material results, 2 games by the 50-move rule and 1 stalemate.
- Total recorded TB hits: 46,921,416.
- No timeout, readyok-timeout, forfeit, crash, illegal move, disconnect, engine error or exception evidence was found in the audited final package.
- ZIP package SHA256: 0915631152d8472610973a30086947fe01d3973d30ff3fd4a6dbb26dc2f5f029.
- games.pgn SHA256: 0750db18e6a62edbcdd6ac08d97b2b20dd578daf1773b249aaad13f5bd816cc0.
- Official champion: Stockfish dev-20260426-1a882efc.
Knockout Bracket Record — R8M-M1 — Stockfish vs PlentyChess
R8M-M1 is recorded as a Stockfish win for bracket continuity, but the PGN/results/scheduler package is currently unavailable on the server and therefore this row is not presented as a full IJCCRL audit pass. The disclosure is kept explicit so the public page does not claim evidence that is not currently present in the archived package set.
Bracket record
- Match: IJCCRL Arena Series 2026 — Original UCI Track — R8M-M1 — Stockfish vs PlentyChess.
- Recorded sporting decision: Stockfish dev-20260426-1a882efc advances.
- Opponent: PlentyChess 7.0.66.
- Audit status: documentary bracket record only.
- Reason: the PGN/results/scheduler package could not be located on the server for full forensic validation.
Publication rule
- The result is used only to prepare the next bracket phase.
- No 50-game score, colour balance claim, mirrored-unit claim or pair-ratio claim is published for M1 until the package is recovered.
- If the archive is later found, the page should be updated with the full audit block.
- Current next-phase consequence: Stockfish is included as the highest qualified seed.
Knockout Audit Update — R8M-M2 — Reckless vs Horsie
R8M-M2 is accepted as a closed and coherent IJCCRL fixed mirrored match. The result is appended as knockout audit evidence only; it remains historical evidence from the previous 1,056-game accumulated regular-season surface and does not alter the current 520-game Regular Season table above.
Official result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Reckless 0.10.0-dev-8ba888c6 Key: reckless-avx2 | 32.5 / 50 65.00% | 25-15-10 | 25 / 25 |
Horsie 1.1.0 Key: horsie-v1 | 17.5 / 50 35.00% | 10-15-25 | 25 / 25 |
Audit controls
- Audit verdict: AUDIT PASS.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — R8M-M2 — Reckless vs Horsie.
- Format: Blitz 5+2, fixed mirrored match.
- Games: 50/50 closed.
- Scheduler state: openingIndex 50 and pending null.
- Colour balance: exact 25 White and 25 Black games per engine.
- Mirrored structure: 25/25 mirrored opening units closed.
- Pair ratio: 0.
- Opening authority: UHO_2024_8mvs_big_+100_+119.pgn, lines #1–#25.
- PGN/results/scheduler consistency: verified.
Result structure
- PGN result distribution: 35 wins for White, 15 draws and 0 wins for Black.
- Colour score: White 42.5/50, Black 7.5/50.
- Reckless by colour: 25.0/25 as White and 7.5/25 as Black.
- Horsie by colour: 17.5/25 as White and 0.0/25 as Black.
- Mirrored-unit distribution: 15 units won by Reckless 1.5–0.5 and 10 units split 1.0–1.0.
- No incomplete mirrored unit was found.
Technical notes
- Termination distribution: 35 checkmates, 9 games by the 50-move rule, 4 threefold repetitions and 2 insufficient-material results.
- Total recorded TB hits: 15,530,988.
- Telemetry note: recorded TB hits appear on the Reckless side; Horsie reports 0 TB hits in the audited records.
- This TB telemetry difference is documented as a reporting/configuration caveat and does not affect the match result.
- Official knockout record: Reckless advances over Horsie by 32.5–17.5.
Knockout Audit Update — R8M-M3 — Obsidian vs RubiChess
R8M-M3 is accepted as a closed and coherent IJCCRL fixed mirrored match after package normalization to the official 50-game contract. The two surplus games from the same final opening block were removed from the publication package so that the audited archive remains at 50 games, exact 25 White / 25 Black per engine and pair ratio 0. The result is appended as knockout audit evidence only; it remains historical evidence from the previous 1,056-game accumulated regular-season surface and does not alter the current 520-game Regular Season table above.
Official result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Obsidian dev-16.15 Key: obsidian-avx2 | 35.0 / 50 70.00% | 28-14-8 | 25 / 25 |
RubiChess 20260308 (avx2) Key: rubichess-avx2 | 15.0 / 50 30.00% | 8-14-28 | 25 / 25 |
Audit controls
- Audit verdict: AUDIT PASS after 50-game package normalization.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — R8M-M3 — Obsidian vs RubiChess.
- Format: Blitz 5+2, fixed mirrored match.
- Games: 50/50 closed in the corrected publication package.
- Colour balance: exact 25 White and 25 Black games per engine.
- Pair ratio: 0.
- PGN/results/scheduler consistency: verified after removing the two surplus games.
- Tiebreak required: no.
- Official knockout record: Obsidian advances over RubiChess by 35.0–15.0.
Result structure
- PGN result distribution: 32 wins for White, 14 draws and 4 wins for Black.
- Colour score: White 39.0/50, Black 11.0/50.
- Obsidian by colour: 24.5/25 as White and 10.5/25 as Black.
- RubiChess by colour: 14.5/25 as White and 0.5/25 as Black.
- Termination distribution: 36 checkmates, 13 threefold repetitions and 1 game by the 50-move rule.
- Total recorded TB hits: 21,566,007.
Normalization note
- The downloaded raw package contained 52 games.
- The corrected audit package removes the two surplus games from the same final C25 opening block to restore the official 50-game contract.
- The corrected package keeps exact 25 White and 25 Black games per engine.
- The corrected score is Obsidian 35.0–15.0 RubiChess.
- The sporting decision is unchanged: Obsidian wins the match without tiebreak.
Knockout Audit Update — R8M-M4 — Alexandria vs Clover
R8M-M4 is accepted as a closed and coherent IJCCRL fixed mirrored match after mirrored package correction to the official 50-game contract. The raw archive contained 52 games, so the audit package removed one complete mirrored unit instead of taking a simple first-50 truncation. This restored exact 25 White / 25 Black per engine and pair ratio 0. The result is appended as knockout audit evidence only; it remains historical evidence from the previous 1,056-game accumulated regular-season surface and does not alter the current 520-game Regular Season table above.
Official result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Alexandria-9.0.9 Key: alexandria-avx2 | 27.5 / 50 55.00% | 19-17-14 | 25 / 25 |
Clover 9.1 Key: clover-avx2 | 22.5 / 50 45.00% | 14-17-19 | 25 / 25 |
Audit controls
- Audit verdict: AUDIT PASS after mirrored 50-game correction.
- Match: IJCCRL Arena Series 2026 — Original UCI Track — R8M-M4 — Alexandria vs Clover.
- Format: Blitz 5+2, fixed mirrored match.
- Games: 50/50 closed in the corrected publication package.
- Colour balance: exact 25 White and 25 Black games per engine.
- Pair ratio: 0.
- PGN/results/scheduler consistency: verified after removing one surplus mirrored unit from the 52-game raw archive.
- Tiebreak required: no.
- Official knockout record: Alexandria advances over Clover by 27.5–22.5.
Result structure
- PGN result distribution: 33 wins for White, 17 draws and 0 wins for Black.
- Colour score: White 41.5/50, Black 8.5/50.
- Alexandria by colour: 22.0/25 as White and 5.5/25 as Black.
- Clover by colour: 19.5/25 as White and 3.0/25 as Black.
- All decisive games in the corrected PGN were won by White.
- The sporting decision is clear: Alexandria wins the match by 5 points without tiebreak.
Normalization note
- The downloaded raw package contained 52 games.
- A simple first-50 truncation was rejected because it produced Alexandria 26W / 24B and Clover 24W / 26B.
- The corrected audit package removes one complete mirrored unit so that both engines remain at 25 White and 25 Black games.
- The corrected score is Alexandria 27.5–22.5 Clover.
- Operational note for the next scripts: stop exactly at 50 games and avoid inherited Source labels from previous match launchers.
Knockout Audit Update — R8M-M5 — Viridithas vs Caissa
R8M-M5 is accepted for publication only as a clean balanced 48-game subset. The raw VPS package contained 50 games and the package hash was valid, but the raw match failed the strict 50-game mirrored contract because the colour balance was Viridithas 26 White / 24 Black and Caissa 24 White / 26 Black. The first 48 games restore exact 24 White / 24 Black per engine and are therefore used as the corrected bracket evidence. This block is deliberately labelled as an accepted subset rather than a full 50-game audit pass.
Accepted subset result
| Engine | Score | W-D-L | W/B |
|---|---|---|---|
Viridithas 20.0.0-dev Key: viridithas-avx2 | 26.5 / 48 55.21% | 21-11-16 | 24 / 24 |
Caissa 1.25 AVX2 Key: caissa-avx2 | 21.5 / 48 44.79% | 16-11-21 | 24 / 24 |
Audit controls
- Accepted publication status: corrected 48-game balanced subset.
- Raw package integrity: SHA256 verified.
- Raw package: 50 PGN games, results.json present and scheduler_state.json present.
- Raw 50-game issue: colour balance failed at 26/24 instead of 25/25.
- Accepted subset: first 48 games only.
- Accepted subset colour balance: exact 24 White and 24 Black games per engine.
- Tiebreak required: no.
- Official bracket consequence: Viridithas advances over Caissa by 26.5–21.5 on the accepted subset.
Result structure
- Accepted subset PGN result distribution: 37 wins for White, 11 draws and 0 wins for Black.
- Colour score in accepted subset: White 42.5/48, Black 5.5/48.
- Viridithas by colour: 22.5/24 as White and 4.0/24 as Black.
- Caissa by colour: 20.0/24 as White and 1.5/24 as Black.
- Termination distribution in accepted subset: 37 checkmates, 6 threefold repetitions and 5 insufficient-material results.
- Total recorded TB hits in accepted subset: 2,588,019.
Correction note
- The raw 50-game result was Viridithas 28.5–21.5 Caissa.
- The raw 50-game result is not used as the official bracket result because the final mirrored cycle was incomplete.
- The excluded games are the last two games of the raw package, both from the incomplete final C25 block with Viridithas as White.
- This restores balanced engine colours but does not claim the originally intended 25 mirrored opening units.
- Operational action: the Ubuntu Node launcher must be corrected so it advances the opening only after the complete colour-reversed mirrored unit and stops on a contract that preserves colour balance.
Current Regular Season Audit Update — 520 Accepted Games
The raw 540-game scheduler package is accepted only after the minimum Rules+ forensic correction. Six proxy_fatal / readyok-timeout adjudications were found. To preserve complete mirrored units, equal games per engine and exact colour balance, two complete 10-game global cycles are excluded. The resulting 520-game corpus is the sole official source for the current standings and phase Elo table.
Audit controls
- Raw scheduler closure: 540 games, openingIndex 540, pending null.
- Raw per-engine balance: 108 games, 54 White and 54 Black.
- Technical exclusions: 6 proxy_fatal results caused by readyok timeout (12000ms).
- Excluded complete cycles: opening indexes 61–70 and 221–230.
- Accepted corpus: 520 games.
- Accepted per-engine state: 104 games, 52 White and 52 Black for every engine.
- Mirrored structure: 260 complete accepted units.
- Pair ratio: 0.
- Accepted proxy_fatal results: 0.
- PGN/results structural parity across the accepted corpus: verified.
Accepted result structure
- Checkmate: 222 games.
- Threefold repetition: 160 games.
- Insufficient material: 59 games.
- Flag: 42 games.
- 50-move rule: 32 games.
- Stalemate: 5 games.
- Sporting flags are retained; only infrastructure/proxy-fatal adjudications are excluded.
- Regular Season winner: Stockfish 18 with 77.0/104.
- Top 8 qualify for Q8; RubiChess and Integral are eliminated from this stage.
Historical Rating / Knockout Archive Note
The previous 1,056-game accumulated rating surface and its completed knockout cycle remain part of the historical archive. The current public competition state is the separately audited 520-game Regular Season followed by the completed current Q8, Q4 and Grand Final phases. Historical match evidence is preserved below without being mixed into the current 520-game standings or the current championship result.
Previous archive summary
- Previous accumulated regular-season surface: 1,056 games.
- Previous completed knockout evidence remains preserved in the detailed blocks on this page.
- Previous final: Stockfish dev-20260426-1a882efc defeated Reckless 0.10.0-dev-8ba888c6 by 55.5–44.5 over 100 audited games.
- Previous cycle champion: Stockfish dev-20260426-1a882efc.
Current page doctrine
- The current rating table uses the accepted 520-game Regular Season only.
- Every current engine is shown with 104 valid games and 52 / 52 colour balance.
- The completed current Q8 bracket uses the audited Regular Season seed order; Q4 is closed and the Grand Final is audit-closed with Stockfish 18 champion over Obsidian dev-16.15 by 403.0–189.0 on 592 canonical games.
- The previous knockout blocks are historical evidence and do not alter current Regular Season scores, phase Elo or the current Grand Final championship record.