New engine releases & news H1 2022

[Frank Quisinsky]

Hi Norman,

Sorry, a good idea!

Best
Frank

[Frank Quisinsky]

Hi there,

on ... Ipman
http://ipmanchess.yolasite.com/compiles.php

1. Koivisto 7.13
2. Berserk 9 dev.

Koivisto by Kim Kåhre & Finn Eggers (Finland, Germany)
Berserk by Jay Honnold (U.S.A.)

I don't know who is Ipman but he do a strong work with all the compiles.
Thank you!!

Best
Frank

[Frank Quisinsky]

The next one ...

Krasimir Topchiyski is programmer of Bagatur (Bulgaria).

Java Engine: https://github.com/bagaturchess/Bagatur/
Updated for 8 days ... (v3.2, current)

JRE: https://www.java.com/de/download/manual.jsp (v1.8.0.321, current)
JDK: https://www.oracle.com/java/technologies/downloads/ (v17.02, current)

Best
Frank

[Frank Quisinsky]

Hi there,

not sure new for others but new for me ...
I saw a newer Texel version in the Dropbox test directory by the programmer:

Dropbox:
https://www.dropbox.com/sh/v16dye7ww33x ... 1ihcsBCu3a

Programmer:
Peter Österlund, Sweden

Version number: 1.08 a18 (from 12th, Sept. 2021).

Best
Frank

Maybe the wrong thread?
Should search the thread for 2021, saw it to late.

[jonkr]

Slow 2.83 for Windows is up on my webpage. I plan on compiling a Linux version some time in next week or so.

It is newly trained nets, and some search changes, but I don't think rating lists need to test it.
In my own tests I got +6 Elo standard, +20 Elo FRC. So unless a lot of games are played luck is probably the biggest factor in whether it scores above 2.8.

There are also a few fixes I didn't want to just sit around. For instance I had broken reading en-passant square in FENs, now fixed. I also added left-right arrows for stepping through games. ( Also my interest in computer chess is low right now so I stopped running chess training/testing on my computer over night. I still fully expect to come back to it eventually and make a Slow 2.9 though.)

[jonkr]

At the time of first upload there was a search bug that left Slow 2.83 almost unable to find some mates. Probably not noticeable playing strength difference but I did just update the download with a fix.
(Slow should find Qxf4 instantly : 4r1k1/p1qr1p2/2pb1Bp1/1p5p/3P1n1R/1B3P2/PP3PK1/2Q4R w - - 0 1)

[Guenther]

Slow 2.83 for Windows is up on my webpage. I plan on compiling a Linux version some time in next week or so.

It is newly trained nets, and some search changes, but I don't think rating lists need to test it.
In my own tests I got +6 Elo standard, +20 Elo FRC. So unless a lot of games are played luck is probably the biggest factor in whether it scores above 2.8.

There are also a few fixes I didn't want to just sit around. For instance I had broken reading en-passant square in FENs, now fixed. I also added left-right arrows for stepping through games. ( Also my interest in computer chess is low right now so I stopped running chess training/testing on my computer over night. I still fully expect to come back to it eventually and make a Slow 2.9 though.)

Jonathan, I sent you an email and a PM four weeks, respectively two weeks ago, but it seems you didn't get/check them?
It was about crashes of SlowChess 2.8 (the sse version). I sent some games/logs and IIRC, I later also found the cause,
which was the Syzygy bases. Not using them made the crashes disappear.
Is it possible that you compiled your tablebase code with other flags/instructions?

[jonkr]

Jonathan, I sent you an email and a PM four weeks, respectively two weeks ago, but it seems you didn't get/check them?
It was about crashes of SlowChess 2.8 (the sse version). I sent some games/logs and IIRC, I later also found the cause,
which was the Syzygy bases. Not using them made the crashes disappear.
Is it possible that you compiled your tablebase code with other flags/instructions?

Turns out the syzygy table base code defaulting to hardware popcount for my compile, so I updated 2.83 sse in the download to not use hardware pop count.

I also fixed a multi-pv bug in positions where mate was possible.
Otherwise play should be exactly the same so didn't update version number.

[Guenther]

Kingfisher 1.2dev r59 GS
(I took the freedom to finally add a version number to the uci id - there never was one in previous releases)

By accident I realized that Kingfisher was worked on a lot again for a month in February 2021
and the author himself estimated a rating progression of around 100 points.
But he suddenly stopped working on it after 2021-03-01.

Code: Select all

original readme

Versions

    dev version (unreleased): >100 ELO stronger than v1.1.1 (estimated)
    v1.1.1: 2315±27 ELO (40/15), 2298±18 ELO (2'+1")
    v1.0: 2281±29 ELO (40/15), 2303±17 ELO (2'+1")

https://github.com/eric-ycw/kingfisher

Code: Select all

My little sanity test at 60s+1s in selfplay against the previous version gave this result:
(games included in download)

Score of Kingfisher_12devr59-64 vs Kingfisher_111-64: 42 - 21 - 37 [0.605]
...      Kingfisher_12devr59-64 playing White: 25 - 10 - 15  [0.650] 50
...      Kingfisher_12devr59-64 playing Black: 17 - 11 - 22  [0.560] 50
...      White vs Black: 36 - 27 - 37  [0.545] 100
Elo difference: 74.1 +/- 55.0, LOS: 99.6 %, DrawRatio: 37.0 %
100 of 100 games finished.

Player: Kingfisher_12devr59-64
   "Draw by 3-fold repetition": 30
   "Draw by adjudication": 1
   "Draw by fifty moves rule": 4
   "Draw by insufficient mating material": 2
   "Loss: Black mates": 10
   "Loss: White mates": 11
   "Win: Black mates": 17
   "Win: White mates": 25
Player: Kingfisher_111-64
   "Draw by 3-fold repetition": 30
   "Draw by adjudication": 1
   "Draw by fifty moves rule": 4
   "Draw by insufficient mating material": 2
   "Loss: Black mates": 17
   "Loss: White mates": 25
   "Win: Black mates": 10
   "Win: White mates": 11

[Frank Quisinsky]

Hi there,

1.
Little Goliath Evolution 3.15.3 NN
by Michael Borgstädt (Germany) ... it seems clearly stronger as 3.15 NN.
Version 3.15.3 used the same "net".

http://www.goliathchess.de/home.html

2.
Rebel 14.1 NN (newer version from Jan. 30th, 2022, first 14.1 NN version from Jan. 25th, 2022).
by Ed Schröder (The Netherland)

http://rebel13.nl/windows/rebel-14.html

Best
Frank