{"id":17084424,"url":"https://github.com/niklasf/fishnet-tuning","last_synced_at":"2026-05-01T02:33:00.836Z","repository":{"id":139222483,"uuid":"254832774","full_name":"niklasf/fishnet-tuning","owner":"niklasf","description":"Experiments to find good parameters for reproducible Stockfish analysis","archived":false,"fork":false,"pushed_at":"2020-12-15T18:02:05.000Z","size":80,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-01-28T20:12:59.495Z","etag":null,"topics":["chess","lichess","stockfish"],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/niklasf.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2020-04-11T09:21:22.000Z","updated_at":"2020-12-15T18:02:10.000Z","dependencies_parsed_at":"2024-07-14T15:02:06.982Z","dependency_job_id":null,"html_url":"https://github.com/niklasf/fishnet-tuning","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/niklasf%2Ffishnet-tuning","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/niklasf%2Ffishnet-tuning/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/niklasf%2Ffishnet-tuning/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/niklasf%2Ffishnet-tuning/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/niklasf","download_url":"https://codeload.github.com/niklasf/fishnet-tuning/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245111928,"owners_count":20562511,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["chess","lichess","stockfish"],"created_at":"2024-10-14T13:07:07.037Z","updated_at":"2026-05-01T02:32:55.813Z","avatar_url":"https://github.com/niklasf.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"Fishnet tuning\n==============\n\nHash size\n---------\n\nSetup: In reproducible analysis, Stockfish uses 1 thread and the hashtable is\ncleared after every move. A fixed number of nodes (1M) is searched at\neach position. Hardware is irrelevant. Playing each position from\n`noob_3moves_sample.epd` with rematch with reversed colors. No adjudication.\n\n```\nHASH(1): 335\ndraw: 1305\nHASH(512): 360\n```\n\nElo diff estimate: -4\n\n```\nHASH(16): 334\ndraw: 1316\nHASH(512): 350\n```\n\nElo diff estimate: -3\n\nForgetfulness\n-------------\n\nSetup: A fixed number of nodes (1M) is searched at each position. Both\nengines have 512M hash, but FORGET clears its hashtable after every move.\n\n```\nFORGET: 88\ndraw: 862\nREMEMBER: 1050\n```\n\nElo diff estimate: -182\n\nCPU comparisons\n---------------\n\nhttps://docs.google.com/spreadsheets/d/1RjL0xcZC94qFtkgd_lDN2jgwHF5rkf6FcOF2rEM4I7I/edit?usp=sharing\n\nNNUE nodes vs classical nodes\n-----------------------------\n\n```\nRank Name                                       Elo     +/-   Games   Score    Draw\n   1 stockfish-12-dev-nodes-4000000             101      53      60   64.2%   61.7%\n   2 stockfish-12-dev-nodes-3000000              47      56      60   56.7%   60.0%\n   3 stockfish-12-dev-nodes-3500000              17      38      60   52.5%   81.7%\n   4 stockfish-fishnet-20200613-nodes-4000000  -176      61      60   26.7%   50.0%\n120 of 120 games finished.\n```\n\n```\nScore of stockfish-fishnet-20200613-nodes-4000000 vs stockfish-12-dev-nodes-2500000: 4 - 23 - 23 [0.310]\n...      stockfish-fishnet-20200613-nodes-4000000 playing White: 4 - 7 - 14  [0.440] 25\n...      stockfish-fishnet-20200613-nodes-4000000 playing Black: 0 - 16 - 9  [0.180] 25\n...      White vs Black: 20 - 7 - 23  [0.630] 50\nElo difference: -139.0 +/- 72.2, LOS: 0.0 %, DrawRatio: 46.0 %\n50 of 50 games finished.\n```\n\n```\nScore of stockfish-fishnet-20200613-nodes-4000000 vs stockfish-12-plus-nodes-2000000: 1 - 25 - 24 [0.260]\n...      stockfish-fishnet-20200613-nodes-4000000 playing White: 1 - 7 - 17  [0.380] 25\n...      stockfish-fishnet-20200613-nodes-4000000 playing Black: 0 - 18 - 7  [0.140] 25\n...      White vs Black: 19 - 7 - 24  [0.620] 50\nElo difference: -181.7 +/- 68.7, LOS: 0.0 %, DrawRatio: 48.0 %\n50 of 50 games finished.\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fniklasf%2Ffishnet-tuning","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fniklasf%2Ffishnet-tuning","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fniklasf%2Ffishnet-tuning/lists"}