New engine releases & news H1 2024

[Frank Quisinsky]

Obsidian 10.0 NN results are now in tournament selection from TalkChess.
Also the link to the games in *.pgn ...

Best
Frank

[peter]

Please note:
A bug produced in a game must not be comprehensible with edit a position, you do!

A bug in the engine that shows up in 56 games of yours should or at least could be reproducible by any other one tablebase- postion of same kind.
A bug you have in your installation only won't be reproducible by anybody but you anyhow, but if you'd at least give a single one of your buggy games or at least a single one position out of them, then anybody could try to reproduce it. Sorry, but I really don't want to look for that in your big database on my own, as much as I appreciate it as such, regards thankfully

[Frank Quisinsky]

Hi Peter,

the other engines plays with the same syzygy directory.
Can be see in the database.

You have been involved with computer chess for so long that you should know that an error in the game can have various causes.

Code: Select all

[ENGINE]
Name=Obsidian 10.0 NN
Author=Gabriele Lombardo
Filename=C:\Chess\Shredder_12\Engines\Obsidian\Obsidian_100-NN.exe
[OPTIONS]
Move Overhead=100
SyzygyPath=c:\Chess\_syzygybases-5

The directory for the bases are allways the same.
Can be see for all the engines in my 40 in 20 tournament, no engine produced here a mistake.
And again, also not in the current tourney, can be also see in the databases I offer.

Maybe I should not write about mistakes I found in engines?
But if so, I hear that first time in all the years.

Thanks Frank, that you search in over 800 games the examples:
Peter, no problem I always like to do that!

You know, I like it more friendly!
And sorry, that you can't reproduce the mistake with the possibilites you have tried.

Best
Frank

[peter]

And sorry, that you can't reproduce the mistake with the possibilites you have tried.

Maybe I could, if you'd give one of your many examples

Nevermind, Frank, if everything's fine for you, then it's just the same for me of course anyhow, regards

[Frank Quisinsky]

And please, if you like to reproduce something ...
Try at first to do the same ...

I use for that 4+2 tournament 5-man syzygy and not 6-man syzygy.
Maybe it's the reason but I think that have other reasons.

Code: Select all

### 2.: to Obsidian 10.0 NN (0): quit

### 3.: to Obsidian 10.0 NN (0): uci

### 4.: to Obsidian 10.0 NN (0): debug on

### from Obsidian 10.0 NN (0): Obsidian 10.0 by gabe

##### Unknown token: Obsidian
### from Obsidian 10.0 NN (0): id name Obsidian 10.0

### from Obsidian 10.0 NN (0): id author gabe

### from Obsidian 10.0 NN (0): option name Hash type spin default 64 min 1 max 33554432

### from Obsidian 10.0 NN (0): option name Clear Hash type button

### from Obsidian 10.0 NN (0): option name Threads type spin default 1 min 1 max 1024

### from Obsidian 10.0 NN (0): option name Move Overhead type spin default 10 min 0 max 1000

### from Obsidian 10.0 NN (0): option name SyzygyPath type string default 

### from Obsidian 10.0 NN (0): uciok

### from Obsidian 10.0 NN (0): Unknown command: 'debug on'.

##### Unknown token: Unknown
### 5.: to Obsidian 10.0 NN (0): setoption name Hash value 128

### 6.: to Obsidian 10.0 NN (0): isready

### from Obsidian 10.0 NN (0): readyok

setoption name Move Overhead value 100

### 7.: to Obsidian 10.0 NN (0): setoption name Move Overhead value 100

setoption name SyzygyPath value c:\Chess\_syzygybases-5

### 8.: to Obsidian 10.0 NN (0): setoption name SyzygyPath value c:\Chess\_syzygybases-5

### 9.: to Obsidian 10.0 NN (0): isready

### from Obsidian 10.0 NN (0): info string Syzygy tablebases loaded. Pieces: 5

Syzygy tablebases loaded. Pieces: 5
### from Obsidian 10.0 NN (0): readyok

Engine: Obsidian 10.0 NN (128 MB)
by Gabriele Lombardo

Again, 56 games added in the thread (*.cbh download file) but can be also found in the database to the tourney as (*.pgn), link can be found in TalkChess tournament selection.

Best
Frank

[peter]

And please, if you like to reproduce something ...
Try at first to do the same ...

I use for that 4+2 tournament 5-man syzygy and not 6-man syzygy.

Sorry, Frank, I don't have 3,4,5 men as subset, just together with 6men. Maybe somebody else can reproduce your issues, as told with my 6men- installation everything's fine with Obsidian 10.0 and Shredder (13)- GUI.

[Ciekce]

Again and again ... I hope the programmers use for test-games "resign=off".

again and again, no one routinely does this, no one will routinely do this, it is not worth the (much) slower testing

[Frank Quisinsky]

And you will never be able to see such bugs ...
A good example is the bug in Peacekeeper!

Did not say that all of the test games should be play with resign = off.

You can't say no one ...
You can speak for yourself only and maybe different of others.

I am looking this evening in Stromphrax games, still running here.
I saw a lot of very nice late-midgames.

Congrutatlion, quantity of wins in late-midgame from equal earlier mid-games is very high, give the total strength a booster (if I compare with stats from many of other engines). Different of nice stats are possible with quantity of pieces on board, some of these stats are more interesting as stats with quantity of moves. People like to look a bit in weaknesses and strengths are working very tricky with the combination of quantity of pieces and quantity of moves at first. Here Chessbase database offer nice features, but I have a tool for it. I wrote that because with resign=off games you have better material for such things. So Elo is not important for me, more the style of the engines, so I am looking for it. For your understandings ...

Best
Frank

[mvanthoor]

Hello everyone.

I announced the release of v1.0 and v1.1 of my engine, Tcheran, on the CCRL forums while waiting for my TalkChess account to be approved.

https://github.com/jgilchrist/tcheran

v1.0 was rated 1981 on CCRL Blitz (now 1879 after the rating recalibration) and although v1.1 hasn't been tested yet, it was a modest ~14-18 Elo improvement in self-play (due to a large set of movegen optimisations).

Hi

Congratulations with the release. I've taken a look at your engine. It looks good; I like the code. In places, it looks somewhat like Rustic's code. You either saw my engine, or we have similar programming styles I like the fact that you have already crate-ified the chess part. That's on my planning as well. I also like the fact that I can just start the engine and type "go" and it starts analyzing. I've tried to crash it on the command-line, but didn't succeed (yet). It even gives error messages in the style of Rustic

On GitHub you state this feature set for Tcheran 1.1:

Code: Select all

    Board
        Bitboard board representation
        Redundant mailbox representation for square lookups
        Zobrist hashing

    Move generation
        Fully legal move generation
        Fancy Magic bitboards

    Search
        Iterative deepening
        Negamax
        Quiescence search
        Principal variation search (PVS)
        Check extensions
        Move ordering
        Transposition table

    Move ordering
        Previous best move
        Most Valuable Victim - Least Valuable Aggressor (MVV-LVA)

    Evaluation
        Material difference
        Midgame and endgame piece square tables
        Tapered midgame vs. endgame evaluation
        Incremental updates

This is _EXACTLY_ the feature-set that Rustic 4-beta also has. (Rustic 4 is basically going to be a massively refactored Alpha 3, with a tapered and tuned evaluation added.) I ran some tests because Rustic 4-beta and Tcheran 1.1 have the exact same feature-set, with these results:

Tcheran 1.1:

Code: Select all

marcel@workstation23:~/Chess/Engines/Uci/Tcheran$ ./tcheran_1.1 
uci
id name Tcheran (v1.1)
id author Jonathan Gilchrist
option name Hash type spin default 256 min 0 max 1024
option name Strategy type string default Main
option name Log type check default false
uciok
setoption name Hash value 256
go
info depth 1 seldepth 1 score cp 36 time 0 nodes 48 nps 3404979 hashfull 0 pv g1f3
info depth 2 seldepth 4 score cp 0 time 0 nodes 250 nps 4877477 hashfull 0 pv g1f3 g8f6
info depth 3 seldepth 7 score cp 35 time 0 nodes 2223 nps 17132419 hashfull 0 pv g1f3 g8f6 d2d4
info depth 4 seldepth 10 score cp 0 time 0 nodes 14054 nps 24309243 hashfull 0 pv g1f3 g8f6 d2d4 d7d5
info depth 5 seldepth 12 score cp 33 time 4 nodes 119982 nps 28421637 hashfull 0 pv g1f3 g8f6 d2d4 d7d5 b1c3
info depth 6 seldepth 14 score cp 0 time 28 nodes 761939 nps 26453161 hashfull 1 pv g1f3 g8f6 d2d4 d7d5 b1c3 b8c6
info depth 7 seldepth 19 score cp 28 time 232 nodes 6237501 nps 26777951 hashfull 11 pv g1f3 g8f6 d2d4 d7d5 b1c3 b8c6 c1e3
info depth 8 seldepth 23 score cp 7 time 2405 nodes 56949052 nps 23675911 hashfull 82 pv e2e4 d7d5 e4d5 g8f6 c2c4 c7c6 d1a4 e7e5
info depth 9 seldepth 24 score cp 36 time 11169 nodes 274384716 nps 24565186 hashfull 290 pv e2e4 e7e5 b1c3 g8f6 g1f3 f8d6 d2d4 b8c6 c1e3
info depth 10 seldepth 29 score cp 6 time 64794 nodes 1450117834 nps 22380345 hashfull 808 pv e2e4 e7e5 c2c3 g8f6 g1f3 f8e7 f3e5 f6e4 d1g4 e4g5
quit

Rustic 4-beta:

Code: Select all

marcel@workstation23:~/Chess/Engines/RusticTest$ ./rustic-alpha-xb-3.59.100-linux-64-bit-bmi2 -q -m256

d888888b                      dP   oo        
88     88                     88             
88oooo88  88    88  d8888b  d8888P dP d88888b
88    88  88    88  8ooooo    88   88 88     
88     88 88    88       88   88   88 88     
88     88  88888P  888888P    dP   dP 888888P
ooooooooooooooooooooooooooooooooooooooooooooo

Engine:    Rustic Alpha 3.59.100
Author:    Marcel Vanthoor
EMail:     mail@marcelvanthoor.nl
Website:   https://rustic-chess.org/
Type:      64-bit
TT size:   256 MB
Protocol:  uci
Threads:   1

go
info score cp 29 depth 1 seldepth 1 time 0 nodes 24 nps 0 pv d2d4
info score cp 0 depth 2 seldepth 2 time 0 nodes 90 nps 0 pv d2d4 d7d5
info score cp 27 depth 3 seldepth 5 time 0 nodes 672 nps 0 pv d2d4 d7d5 g1f3
info score cp 0 depth 4 seldepth 8 time 0 nodes 2292 nps 0 pv d2d4 d7d5 g1f3 g8f6
info score cp 27 depth 5 seldepth 12 time 1 nodes 13957 nps 13957000 pv d2d4 d7d5 g1f3 b8c6 b1c3
info score cp 0 depth 6 seldepth 18 time 6 nodes 66545 nps 11090833 pv d2d4 d7d5 g1f3 g8f6 f3e5 f6e4
info score cp 19 depth 7 seldepth 18 time 23 nodes 301070 nps 13090000 hashfull 1 pv d2d4 d7d5 g1f3 g8f6 b1c3 b8c6 f3g5
info score cp 8 depth 8 seldepth 23 time 181 nodes 2069953 nps 11436204 hashfull 14 pv e2e4 e7e5 b1c3 b8c6 g1f3 g8f6 d2d4 e5d4 f3d4 c6d4 d1d4
info score cp 27 depth 9 seldepth 23 time 806 nodes 9439693 nps 11711778 hashfull 60 pv e2e4 b8c6 g1f3 g8f6 f1d3 d7d5 e4d5 f6d5 e1g1
info score cp 6 depth 10 seldepth 24 time 3641 nodes 42194307 nps 11588659 hashfull 249 pv g1f3 g8f6 b1c3 b8c6 e2e4 e7e5 f1c4 f8d6 e1g1 e8g8
info score cp 19 depth 11 seldepth 30 time 11438 nodes 136168509 nps 11904923 hashfull 634 pv g1f3 g8f6 e2e3 b8c6 d2d4 e7e6 b1d2 d7d5 f1e2 f6e4 e1g1
info score cp 11 depth 12 seldepth 31 time 48437 nodes 526861780 nps 10877259 hashfull 997 pv g1f3 g8f6 b1c3 b8c6 e2e4 e7e5 f1c4 f6e4 c3e4 d7d5 c4d3 d5e4 d3e4
quit

As you can see in the time to depth, it turns out that Rustic 4 is a (much) faster engine. (For some reason your NPS is much higher, but that can be due to the way you count it. I only count nodes that do actual work and don't count discarded / early-exit nodes. I also take care not to count q-search nodes twice.)

This bears out in the one-to-one test:

Code: Select all

Score of Rustic Alpha 3.59 vs Tcheran 1.1: 2409 - 727 - 864 [0.710]
...      Rustic Alpha 3.59 playing White: 1253 - 335 - 412  [0.730] 2000
...      Rustic Alpha 3.59 playing Black: 1156 - 392 - 452  [0.691] 2000
...      White vs Black: 1645 - 1491 - 864  [0.519] 4000
Elo difference: 155.8 +/- 10.2, LOS: 100.0 %, DrawRatio: 21.6 %
SPRT: llr 0 (0.0%), lbound -inf, ubound inf
4000 of 4000 games finished.

Player: Rustic Alpha 3.59
   "Draw by 3-fold repetition": 546
   "Draw by fifty moves rule": 91
   "Draw by insufficient mating material": 213
   "Draw by stalemate": 14
   "Loss: Black mates": 335
   "Loss: White mates": 392
   "Win: Black loses on time": 1
   "Win: Black mates": 1155
   "Win: White loses on time": 1
   "Win: White mates": 1252
Player: Tcheran 1.1
   "Draw by 3-fold repetition": 546
   "Draw by fifty moves rule": 91
   "Draw by insufficient mating material": 213
   "Draw by stalemate": 14
   "Loss: Black loses on time": 1
   "Loss: Black mates": 1155
   "Loss: White loses on time": 1
   "Loss: White mates": 1252
   "Win: Black mates": 335
   "Win: White mates": 392
 

Rustic 4-beta is stronger by 155 Elo. Going by your calculation of a 14 Elo improvement in version 1.1, that would put Rustic 4-beta at:

1879 + 14 + 155 = 2048

In my early SPRT testing, Rustic 4-beta (then still called 3.1) gained 248 Elo over version Alpha 3. Alpha 3 is rated at 1801, which would put version 4 at 2049... within 1 Elo-point as predicted by extrapolating Rustic 4-beta's strength from its performance against Tcheran.

This means that, if you optimize version 1.1 (and later versions obviously) by profiling it and removing bottlenecks, you can gain about 150 Elo points just by increasing the speed of your engine, without even adding any features.

I hope these insights are helpful to you. Good luck on further implementation

[Frank Quisinsky]

Starzix 4.0 NN
Ricardo Pinto, POR
https://github.com/zzzzz151/Starzix

A battle has began ...
Between Starzix and Pawn (both are Portugal).
End of the battle ... both are stronger as Stockfish.
Go Starzix / Pawn ... go!
Loud thinking!

Thank you!