Skip to content

SPRT numerics issue #879

@vondele

Description

@vondele

@87flowers maybe you can have a look at this one as well, it is the continuation of #878

The following match

./fastchess  -rounds 20000 -games 2 -repeat -srand $RANDOM \
       -concurrency 16 -use-affinity 0-19 -openings file=UHO_Lichess_4852_v1.epd format=epd order=random\
       -engine name=sf0 cmd=./stockfish option.Threads=1 option."Skill Level=10" tc=3+0.03 \
       -engine name=sf1 cmd=./stockfish option.Threads=1 tc=3+0.03 \
       -each proto=uci \
       -ratinginterval 10 -scoreinterval 0 -report penta=true \
       -sprt elo0=-1500 elo1=-500 alpha=0.05 beta=0.05 model=normalized

doesn't terminate

--------------------------------------------------
Results of sf0 vs sf1 (3+0.03, 1t, 16MB, UHO_Lichess_4852_v1.epd):
Elo: -827.87 +/- nan, nElo: -2248.42 +/- 39.58
LOS: 0.00 %, DrawRatio: 0.68 %, PairsRatio: 0.00
Games: 296, Wins: 1, Losses: 292, Draws: 3, Points: 2.5 (0.84 %)
Ptnml(0-2): [144, 3, 1, 0, 0], WL/DD Ratio: inf
LLR: -nan (nan%) (-2.94, 2.94) [-1500.00, -500.00]
--------------------------------------------------

probably because the LLR is NaN.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions