Fat Titz 1.0 released

[dkappe]

I wonder if mainstream chess publications and sites would permit “Fat Titz by Stockfish” advertisements.

[Eduard]

I'm running a match between Fat Titz 1.0 and Stockfish dev 270821:

GUI Fritz 16
Both engines with Ponder ON and 8 Threads Ryzen 3900X
Timecontrol 1+1
Hash 1024 MB
Slow Mover 100
Move Overhead 30
Syzygy all 5 men.

Start:



My test set can be downloaded here (100 variants in CBH and PGN):

Solista-Test:

https://filehorst.de/d/dGirEegj

My Test live after 82 games:



Fat Titz impressive!

Only one variant was won by both sides. Stockfish Dev is having a hard time. After 100 games I take a break and continue the match tomorrow. It will be exciting!

[Eduard]

My test set database had some errors. I was only able to correct it after interrupting the match after 100 games. Here is the corrected database:

Edited Testset:
https://filehorst.de/d/dpxyufuq

[Eduard]

Here is a live match on Youtube against Stockfish dev:
https://www.youtube.com/channel/UCVkew3 ... F-zZ9rfsw

[Eduard]

My match is over. First of all, a few brief comments:
I chose TC 1m+1s to get a first impression. This is my first ENG match on my Ryzen 3900X. 80% of the opening variants are those that are played on the 3 servers (PlayChess, InfinityChess, LiChess). I didn't always choose the best defenses or there would be even fewer wins. Perhaps I'll swap some variants with sharper ones. I don't like bullet, I won't test any more bullet matches.

Because Ponder ON is always played on the servers, only this type of match is important for me!

When I saw from another tester that Fat Titz had only lost -1 in the bullet against Stockfish 260821, and even won at 10m with +1, I couldn't believe it, I thought it was a bad joke. Until then, I thought the Fat Titz engine was also a joke. To convince myself, I started my own match. Coincidentally, a new green Stockfish network and a new Stockfish engine from 270821 appeared at this time.

I was counting on a clear win for Stockfish dev. Just before going to bed yesterday, I was very surprised by the interim result.

Here is the end result:

Code: Select all

Ryzen 3900X, Blitz 1min+1sec  0
                                        
1 Fat Titz 260821 64 AVX2 NUMA  +7/=186/-7 50.00%  100.0/200
2 Stockfish 270821              +7/=186/-7 50.00%  100.0/200

GUI=Fritz 16
Hash=1024 MB
Ponder=ON
SlowMover=100
Move Overhead=30
Testset=My own new testsuite with 100 variants
Threads= Each engine 8 Threads Ryzen 3900X

Kns in Startposition:




Basically, I can say that Stockfish dev. didn't play better. Fat Titz seemed to be reminiscent of Lc0 in a few games, I feel. If Fat Titz was supposed to be a joke engine, then it turned out incredibly well.

In the winning games, Fat Titz won 2x with white pieces against French, while Stockfish only got a draw here. Against Caro-Kann and Alekhine both engines were able to shine with white. All other wins were different.

Here are all 14 winning games in PGN for download:

https://filehorst.de/d/dvEFDCvs

Eduard.

[Robert Flesher]

Hello, is there an .exe download available?

[Robert Flesher]

Hello, is there an .exe download available?

found it, lol!

[Eduard]

PGN only,

[Eduard]

Fat Titz 1.0 wins Blitz Tour on PlayChess!

I'm (E N) tied for 3rd place, but Fat Titz was the only engine that won a game. Places 1, 2, 4 and 5 received a point from Bye. Bye was not given to me . Fat Titz (on 22 Threads) had to fight hard all the games (hahaha... ), Polomo played on 126 threads.

Code: Select all

Saturday 28 August Blitz 5+0, 15 Rounds 2021
(TD: Pooper Scooper#, Time: 5 min, 15 rounds)

1 polomo,Eman 7.50 64-bit AVX             2733 ♦ ½/5 ♦ ½/2 ◊ ½/6 ◊ ½/4 ♦ ½/7 ◊ ½/3 ♦ ½/8 ♦ ½/5 ◊ ½/2 ◊ ½/4 ♦ ½/6 ◊ ½/3 ◊ 1/9 ♦ ½/5 ◊ ½/2 8.0 / 15 61.50   
2 AHGST,CorChess 2 270821                 2674 ♦ ½/8 ◊ ½/1 ◊ ½/7 ♦ ½/3 ◊ ½/6 ♦ ½/4 ◊ ½/5 ♦ ½/8 ♦ ½/1 ◊ ½/6 ♦ ½/3 ◊ ½/7 ◊ ½/4 ◊ 1/9 ♦ ½/1 8.0 / 15 61.00   
3 E N,Fat Titz 260821 64 A                2748 ♦ ½/6 ♦ ½/4 ◊ ½/5 ◊ ½/2 ♦ ½/8 ♦ ½/1 ◊ ½/7 ◊ ½/6 ♦ ½/4 ◊ ½/5 ◊ ½/2 ♦ ½/1 ♦ ½/7 ◊ ½/6 ♦ 1/5 8.0 / 15 60.50 96.00 
4 Victorious,Fat Titz 260821 64 P         2699 ♦ ½/7 ◊ ½/3 ◊ ½/8 ♦ ½/1 ◊ ½/5 ◊ ½/2 ♦ ½/6 ◊ ½/7 ◊ ½/3 ♦ ½/1 ♦ ½/8 ◊ ½/6 ♦ ½/2 ♦ ½/7 ◊ 1/9 8.0 / 15 60.50 88.00 
5 Pooper Scooper#,HekaStockfishNNUE 28    2696 ◊ ½/1 ♦ ½/8 ♦ ½/3 ◊ ½/7 ♦ ½/4 ◊ ½/6 ♦ ½/2 ◊ ½/1 ◊ ½/8 ♦ ½/3 ◊ ½/7 ◊ 1/9 ♦ ½/6 ◊ ½/1 ◊ 0/3 7.5 / 15 61.50   
6 TicTacto,Eman 7.50 64-bit BMI           2714 ◊ ½/3 ♦ ½/7 ♦ ½/1 ◊ ½/8 ♦ ½/2 ♦ ½/5 ◊ ½/4 ♦ ½/3 ◊ ½/7 ♦ ½/2 ◊ ½/1 ♦ ½/4 ◊ ½/5 ♦ ½/3 ◊ ½/7 7.5 / 15 60.00 97.50 
7 Comchess,Stockfish 170821               2739 ◊ ½/4 ◊ ½/6 ♦ ½/2 ♦ ½/5 ◊ ½/1 ◊ ½/8 ♦ ½/3 ♦ ½/4 ♦ ½/6 ◊ ½/8 ♦ ½/5 ♦ ½/2 ◊ ½/3 ◊ ½/4 ♦ ½/6 7.5 / 15 60.00 95.00 
8 Cubail,SF PB 270821                     2714 ◊ ½/2 ◊ ½/5 ♦ ½/4 ♦ ½/6 ◊ ½/3 ♦ ½/7 ◊ ½/1 ◊ ½/2 ♦ ½/5 ♦ ½/7 ◊ ½/4 5.5 / 11     
9 (Bye)                                        ♦ 0/5 ♦ 0/1 ♦ 0/2 ♦ 0/4 0.0 / 4     

[Eduard]

There are so many new nn networks. A new stockfish boy is born almost every day. He knocks everything down. Then a wizard by the name of Sopel comes along, takes an old man Cfish, mixes him a new magic drink, and a new star is born.

Youtube match by --> VendableBee 420:
Fat Titz 26/08/2021 + fat_titz.nnue
VERSUS
Stockfish 15 Dev 26/08/2021 + nn-517c4f68b5df.nnue

100 games each Round Robin

Book Details:
'TCEC 21_superfinal'
= 50 x openings.
Sequential Openings with colors reversed.

Time control Blitz 10m + 5s

Fat _Titz versus SF15 dev with SuFi 21 Book, last game is running.