Posts

Showing posts from December, 2018

New site

There were delays posting here, occasionally quite long. Thus new posts are at https://tcecbayeselo.chessdom.org/ .

TCEC Season 14, Division 1, Game 13 speculation simulation

For ratings using all games in this season regardless of upgrades and also previous season(s) if same version see previous post Output from BayesElo program Rank Player name Points EPoints StdDev ERank 1 2 3 4 5 6 7 1 Laser 181205 2.0 16.62 4.03 3.03 33 21 13 9 7 6 5 2 Ginkgo 2.18b 2.0 16.06 4.06 3.31 28 21 13 9 8 7 6 3 LCZero v19-TP-11248/v19.1-RC2-11248/v19.1-11248 1.5 15.28 2.55 3.41 12 22 24 18 11 7 4 4 Chiron S14 2.0 14.93 4.23 3.92 21 17 13 10 9 8 9 5 KomodoMCTS 2180.00/2210.00/2217.00/2221.00 1.5 13.36 2.43 4.78 3 9 15 19 19 16 13 6 Fizbo 2 1.5 12.95 2.33 5.10 2 6 12 17 20 19 15 7 Jonny 8.1 1.0 11.97 2.28 5.83 1 3 7 12 17 21 23 8 Fritz 16.10 0.5 10.81 2.33 6.61 0 1 3 6 10 16 25 Situation before game, simulation about most common position 100k simulations, sorted by who has most 1., then who has most >=2., then most >=3., etc.. Columns 1.-8. are how often it ended in simulation at that position. Pos Name 1. 2. 1-2. 3. 4. 5. 6. 7-8. 7. 8. 1.

TCEC Season 14, Division 1, Game 12: Laser 181205-Fritz 16.10

Currently playing: 13. Fizbo 2-Fritz 16.10 Previous game 12. Laser 181205-Fritz 16.10 1-0 added to BayesElo lists, result table: Seed: 1544883383494712 Pos Engine Rating Points Games -Crash Direct encounter Wins Black wins SB Coin 1. Ginkgo 2.18b 3324±233 2.0 3 0 0.5 1 0 2.25 0.63 2. Laser 181205 3315±255 2.0 3 0 0.5 1 0 2.25 0.21 3. Chiron S14 3315±248 2.0 3 0 0.0 1 0 1.75 0.74 4. Fizbo 2 3262±36 1.5 3 0 1.0 0 0 2.50 0.84 5. KomodoMCTS 2221.00 3279±57 1.5 3 0 1.0 0 0 2.50 0.40 6. LCZero v19.1-11248 3336±73 1.5 3 0 1.0 0 0 2.00 0.19 7. Jonny 8.1 3250±30 1.0 3 0 0.0 0 0 1.75 0.04 8. Fritz 16.10 3232±52 0.5 3 0 0.0 0 0 1.00 0.33 Gull 3 is used as offset at 3217. Difference for some values in parenthesis if any. Created using Bayesian-Elo and some local scripts. Search for 'version' for individual rating for each version. Search for 'One rating' for one rating for each progam (not useful).