Berserk 11 released. Number 2 free engine!?

[Jouni]

Code: Select all

Score of berserk11 vs Koivisto9.2: 22 - 5 - 115 [0.560]
...      berserk11 playing White: 15 - 0 - 56  [0.606] 71
...      berserk11 playing Black: 7 - 5 - 59  [0.514] 71
...      White vs Black: 20 - 7 - 115  [0.546] 142
Elo difference: 41.8 +/- 24.3, LOS: 99.9 %, DrawRatio: 81.0 %
142 of 400 games finished.

60 + 0,6 and 8moves book. WOW!

[chessica]

good post

[peter]

Code: Select all

Score of berserk11 vs Koivisto9.2: 22 - 5 - 115 [0.560]
...      berserk11 playing White: 15 - 0 - 56  [0.606] 71
...      berserk11 playing Black: 7 - 5 - 59  [0.514] 71
...      White vs Black: 20 - 7 - 115  [0.546] 142
Elo difference: 41.8 +/- 24.3, LOS: 99.9 %, DrawRatio: 81.0 %
142 of 400 games finished.

60 + 0,6 and 8moves book. WOW!

With VSTC (15"+0.25) single threaded I had even a rather suspect result against Ethereal 14.00:

Score of berserk-11-x64-avx2-pext vs Ethereal-14.00-pext-avx2: 96 - 33 - 171 [0.605]
... berserk-11-x64-avx2-pext playing White: 82 - 1 - 67 [0.770] 150
... berserk-11-x64-avx2-pext playing Black: 14 - 32 - 104 [0.440] 150
... White vs Black: 114 - 15 - 171 [0.665] 300
Elo difference: 74.1 +/- 25.6, LOS: 100.0 %, DrawRatio: 57.0 %
300 of 300 games finished.

Not sure, if the 15 concurrent games at the 16x3.5GHz CPU were really of equal chances as for thread- binding, but I didn't see any problems in task manager as long as I was watching. Single games I had checked so far indicate Ethereal coming into time- shortness now and then sooner than Berserk, but no losses on time as far as I saw till now neither.
If your and my results really are to be seen as for height of difference to Koivisto and Etheral quantitatively (which of course they aren't with only two opponents and two different hardware- TCs, openings were UHO 6mvs of evaluation 110-119 again), distance to Dragon even would be only half of what it was, guess at least as for my VSTC- results I doubt that much.
I'd wait for results at big lists.

Edit: STS- MEA again too (like in the other thread about Berserk 11)

Code: Select all

    EPD  : LAN_vs5.epd
    Time : ms
                                                Max     Total    Time    Hash          
    Engine           Score   Found  Pos   ELO   Score   Rate     ms      Mb  Cpu      
 1  SF230203         144406   1193  1500  4342  149703  96.5%    200     8    8
 2  lc0221229-325M   144040   1189  1500  4329  149703  96.2%    500     2    2
 3  Dragon3.2        143310   1180  1500  4306  149703  95.7%    200     8    8
 4  Berserk11        139297   1094  1500  4185  149703  93.0%    200     8    8
 5  Kovisto9.2       139203   1089  1500  4185  149703  93.0%    200     8    8
 6  Berserk230130    138822   1081  1500  4171  149703  92.7%    200     8    8
 7  Berserk10        138643   1081  1500  4167  149703  92.6%    200     8    8
 8  Ethereal14       138604   1087  1500  4167  149703  92.6%    200     8    8

                                    Created with MEA
                                          by
                                       Ferdinand
                                         Mosca

[Jouni]

Final result:

Code: Select all

Score of berserk11 vs Koivisto9.2: 57 - 14 - 329 [0.554]
...      berserk11 playing White: 37 - 0 - 163  [0.593] 200
...      berserk11 playing Black: 20 - 14 - 166  [0.515] 200
...      White vs Black: 51 - 20 - 329  [0.539] 400
Elo difference: 37.5 +/- 14.0, LOS: 100.0 %, DrawRatio: 82.3 %
400 of 400 games finished.

[pohl4711]

My testruns (SPCC-Ratinglist and UHO-Top10-ratinglist) of Berserk 11 are running since some hours here and I can say, that Berserk 11 intermediate results look very promising...Results probably tuesday morning (CET).

Stay tuned!

[Dariusz]

Short thinking time 1m + 0.6s move on MacBook Pro M1, 1100 games.

A noticeable increase in the strength of Berserk 11.
Congratulations to Jay Honnold and jherrera80

It's great that the Berserk source code can be easily compiled for different systems. I respect that and thank you very much!

https://www.chessengeria.com/post/berserk-11-released

[Jouni]

In Mark Young match Berserk has 45% score vs SF dev .

[Jouni]

There is quite serious tablebase bug.
[fen]8/8/2pk4/8/8/8/3KPP2/8 w - - 0 0[/fen]
Berserk don't output anything here with syzygy access!
[fen]2K5/k3q3/6pR/6p1/6Pp/7P/8/3R4 w - - 0 1[/fen]

Analysis by Berserk 11:

1.Rh7 Qxh7 2.Rd7+ Ka8 3.Kc7 Qg8 4.Rd8+ Qxd8+ 5.Kxd8 Kb7 6.Ke7 Kb6 7.Kf6 Kb5 8.Kxg6 Kb4 9.Kxg5 Kb5 10.Kxh4 Ka4 11.g5 Kb4 12.g6 Ka5 13.g7 Kb4
-+ (-#1932) Depth: 41/40 00:00:02 26500kN, tb=472307

I played test games without tablebases.

[RubiChess]

Not really a bug.
When position is in tablebase, Berserk just plays the winning move without showing PV or eval. I must admit that this is confusing when using the GUI in analyse mode and "go infinite" is not allowed to display a move, so Berserk doesn't show anything.
When seraching and finding a TB win, the score in Berserk output is probably something like 32000 - DTZ. Which is misinterpreted as "mate in x moves" by the GUI cause there is no specification in UCI for scores showing a TB win.

Regards, Andreas

[abgursu]

About topic, you know, lc0 is free too.