TCEC14: the 14th top chess engine championship Article Accepted Version The official report Haworth, G. and Hernandez, N. (2019) TCEC14: the 14th top chess engine championship. ICGA Journal, 41 (3). pp. 143- 151. ISSN 1389-6911 doi: https://doi.org/10.3233/ICG-190117 Available at http://centaur.reading.ac.uk/82052/ It is advisable to refer to the publisher’s version if you intend to cite from the work. See Guidance on citing . Published version at: https://content.iospress.com/articles/icga-journal/icg190117 To link to this article DOI: http://dx.doi.org/10.3233/ICG-190117 Publisher: The International Computer Games Association All outputs in CentAUR are protected by Intellectual Property Rights law, including copyright law. Copyright and IPR is retained by the creators or other copyright holders. Terms and conditions for use of this material are defined in the End User Agreement . www.reading.ac.uk/centaur CentAUR Central Archive at the University of Reading Reading’s research outputs online TCEC14: the 14th Top Chess Engine Championship Guy Haworth and Nelson Hernandez1 Reading, UK and Maryland, USA TCEC Season 14 started on November 12th 2018 and introduced a number of changes from TCEC 13 (Haworth and Hernandez, 2019b). An enlarged Division 4 featured twelve engines and seven newcomers to accommodate the increasing interest in computer chess and this competition in particular. The other divisions remained eight strong. The five divisions played two or more double round-robins (‘DRR’) each with promotions and relegations following. Tempi gradually lengthened from ‘Rapid’ to ‘Classical’, and the Premier division’s top two engines played a 100-game match to determine the Grand Champion. The trio of STOCKFISH, KOMODO and HOUDINI have dominated the TCEC medals for several seasons and a key point of interest was whether others would reach the podium. LEELA CHESS ZERO and ETHEREAL were certainly expected to perform well in Division P, having shown remarkable improve- ment in the previous few months. KOMODO MCTS was a dark horse. DP D1 D2 D3 D4 Fig. 1. Logos for the TCEC 14 engines (CPW, 2019) as in their original divisions. There were a few nudges to TCEC’s adjudication rules. Draw adjudication could be invoked after move 35 (rather than move 40) and the two engines had to both evaluate within 0.08 (rather than 0.05) for 1 Corresponding author: [email protected] eight consecutive plies and with plycount 0. While draw-adjudication requirements were relaxed, win-adjudication requirements were tightened. Engine evaluations had to be outside 10 (rather than 6.5) for ten consecutive plies (rather than eight); plycount was not a factor. This change was welcomed by those of us who wanted to see a clearer demonstration of superiority on the board: it will be interest- ing to see how long it prolongs the decisive games and what mysteries remain. The common platform for TCEC14 consisted of two computers. One was the established, formidable 44-core server of TCEC11-13 (Intel, 2017) with 64GB of DDR4 ECC RAM and a Crucial CT250M500 240 GB SSD for the EGTs. The ‘GPU server’, a Quad Core i5 2600k, was sporting Nvidia (2018) GeForce RTX 2080 Ti and 2080 GPUs for those engines which could exploit them. 1 The engines Season 13 competitors BOBCAT, DEUS X, HANNIBAL and SENPAI rested for this TCEC season. TCEC welcomed first appearances for engines DEMOLITO, KOMODO MCTS, PIRARUCU, ROFCHADE, SCHOONER, SCORPIONN and WINTER, see Fig. 1 and Table 1. Table 1. The TCEC14 engines (CP, 2019), details, authors and progress. Engine Initial proto- Hash Countr Final # thr. EGTs Authors ab Name Version ELO Div. col Kb y Codes Div. 01 An Andscacs 095 3391 P 43 uci 8,192 — Daniel José Queraltó AD ↘ 1 02 Ar Arasan CCC3 3204 3 43 xboard 16,384 Syz. Jon Dart US → 3 03 Bo Booot 6.3.1 3246 2 32 uci 16,384 — Alex Morozov UA → 2 04 Cb ChessbrainVB 3.72 3258 2 43 xboard 1,200 — Roger Zuehlsdorf DE → 2 05 c22 chess22k 1.11 3072 4 16 uci 4,096 — Sander Maassen vd Brink NL → 4 06 Ch Chiron S14 3354 1 43 uci 16,384 Syz. Ubaldo Andrea Farina IT → 1 07 Dm Demolito 20181029 2900 4 43 uci 2,048 Syz. Lucas Braesch FR → 4 08 Et Ethereal 11.14 3377 P 43 uci 16,384 Syz. Andrew Grant US → P 09 Fi Fire 7.1 3452 P 43 uci 16,384 Syz. Norman Schmidt US → P 10 Fz Fizbo 2 3265 1 43 uci 16,384 Syz. Youri Matiounine US → 1 11 Fr Fritz 16.10 3244 1 43 uci 16,384 Nal? Vasik Rajlich CZ/US ↘ 2 12 Gi Ginkgo 2.18b 3340 1 43 uci 16,384 Syz. Frank Schneider DE ↘ 2 13 Gu Gull 180521 3217 2 43 uci 16,384 — Vadim Demichev RU ↘ 3 14 Ha Hannibal 20181202 3169 3 16 uci 8,192 — Sam Hamilton, Edsel Apostol US/PH ↘ 4 15 Ho Houdini 6.03 3527 P 43 uci 16,384 Syz. Robert Houdart BE → P 16 Jo Jonny 8.1 3250 1 43 uci 16,384 Syz. Johannes Zwanzger DE → 1 Don Dailey, Larry Kaufman, 17 Ko Komodo 2227.00 3565 P 43 uci 16,384 Syz. US → P Mark Lefler 18 Km Komodo MCTS 2180.00 3475 4 43 uci 16,384 Syz. Mark Lefler US ↗↗↗↗↘ 1 19 La Laser 181205 3241 1 43 uci 16,384 Syz. Jeffrey An, Michael An US → 1 20 Lc LCZero v19-TP-11248 3247 3 4 uci — Syz. UCT/NN AI Community — ↗↗↗ P 21 Ne Nemorino 5.05 3181 4 43 uci 16,384 Syz. Christian Günther US ↗↘ 4 22 Ni Nirvana 2.4 3160 2 32 uci 16,384 — Thomas Kolarik US → 2 23 Pe Pedone 191118 3197 3 43 uci 16,384 Syz. Fabio Gobbato IT → 3 24 pi pirarucu 2.6.9 2900 4 43 uci 4,096 — Raoni Campos BR → 4 25 Ro Rodent III 1.0.171 3030 4 16 uci 4,096 — Pawel Koziol PL → 4 26 rf rofchade 1.019T 2900 4 43 uci 16,384 — Ronald Friederich NL ↗ 3 27 Sc Schooner 2.0.33 2900 4 16 xboard 1,024 — Dennis Sceviour CA → 4 28 Sn ScorpioNN 2.8.9 2600 4 — xboard — — Daniel Shawl ET ↘ – Tord Romstad, Marco Costalba, NO/IT/ 29 St Stockfish 181224 3612 P 43 uci 16,384 Syz. → P Joona Kiiski, Gary Linscott FI/CA 30 Te Texel 1.08a13 3197 2 43 uci 16,384 Syz. Peter Österlund SE ↘ 3 31 Tu Tucano 7.06 2919 4 43 xboard 1,024 — Alcides Schulz BR → 4 32 Va Vajolet2 2.6.1 3184 3 43 uci 16,384 Syz. Marco Belli IT → 3 33 Wa Wasp 3.37 2964 4 43 uci 8,192 Syz. John Stanback US → 4 34 Wi Winter 181107 2900 4 43 uci 16,384 — FM Jonathan Rosenthal CH ↘ – 35 Xi Xiphos 0.4.14 3245 2 43 uci 16384 — Milos Tatarevic RS → 2 2 Division 4: 2 DRR Phases, 4 round robins, 264 games, 30+10/m As for TCEC12/13, each engine played both White and Black from four-ply openings defined by the second author here. The results are as in Table 2: ‘P%’ is the %-score and ‘ELO’ is the change to the engine’s nominal ELO based on its performance. Generic stats are in Tables 9 and 10. Online interest naturally focused on the newcomers, especially KOMODO MCTS (Chessdom, 2018), a further innovation from the Lefler/Kaufman camp. The engines had a wide range of ability leading to only 34.1% of games being drawn: those given a default ‘TCEC-entry ELO’ of 2900 ranged across the field. WINTER was always headed for a demotion spot. SCORPIONN clearly was not ready for the contest and even though it disconnected eight times, it did not impact the ranking elsewhere. The bottom three missed TCEC Cup 2 (Haworth and Hernandez, 2019c). The three engines promoted were clearly ahead: KOMODO MCTS, ROFCHADE and NEMORINO. Table 2. The TCEC14 Division 4 cross-table: two DRR phases, 44 rounds, 264 games. # Engine ELO Pts P% Elo ± SB Km rf Ne Sc pi Wa Dm Tu Ro c22 Wi Sn 01 Komodo MCTS 2180.00 3475 37.5 85.2 -133 745.75 ==01 111= 11== 1111 1111 11=1 =11= =111 1=11 111= 1111 02 rofChade 1.019T 2900 32.5 73.9 +461 627.75 ==10 0111 ==== =10= ==1= 11=1 =111 1111 1=== 11=1 1111 03 Nemorino 5.05 3181 30.0 68.2 -96 541.00 000= 1000 ===1 11== ==1= 1=1= 1111 =1=1 1101 111= 1111 04 Schooner 2.0.33 2900 28.0 63.6 +317 507.25 00== ==== ===0 01=1 ==== =0== 1=== 1111 =111 111= 1111 05 pirarucu 2.6.9 2900 24.5 55.7 +205 431.25 0000 =01= 00== 10=0 =1=1 ==01 1=== 1=== 1==1 1=1= 1111 06 Wasp 3.37 2964 24.5 55.7 +88 409.25 0000 ==0= ==0= ==== =0=0 0=0= =1=1 =1=1 11=1 1111 1111 07 Demolito 20181029 2900 23.0 52.3 +157 409.50 00=0 00=0 0=0= =1== ==10 1=1= 0=11 0==1 ===1 =011 1111 08 Tucano 7.06 2919 18.0 40.9 -38 298.50 =00= =000 0000 0=== 0=== =0=0 1=00 01== 01=1 01=1 1111 09 Rodent III 1.0.171 3030 17.5 39.8 -256 266.50 =000 0000 =0=0 0000 0=== =0=0 1==0 10== ==10 =111 1111 10 chess22k 1.11 3072 17.0 38.6 -344 279.75 0=00 0=== 0010 =000 0==0 00=0 ===0 10=0 ==01 1==1 1111 11 Winter 181107 2900 11.0 25.0 -227 177.50 000= 00=0 000= 000= 0=0= 0000 =100 10=0 =000 0==0 1=11 12 ScorpioNN 2.8.9 2600 0.5 01.1 -135 5.50 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0=00 3 Division 3: two DRR phases, 14 rounds, 112 games, tempo 30+10/m Again, the eight engines involved played both sides of 14 prescribed four-ply openings.