Berserk 14 / 4.7.0 HCE

[Frank Quisinsky]

Hi there,

I was surprised this morning, and I'm really happy about it.

Two new Berserk's are available!

v. 14
v. 4.7.0 ... interesting for the fight for the HCE crone!

In the past I criticized the high move average from Berserk.
But fact is that the version before this new release ... Berserk 250622 CE dev ... is on place 3 for my new ETOC system.
Why, with the fighting games for a draw vs. clearly weaker engines Berserk is really successfull. Also the number of quick wins is clearly better as in older versions. My criticism was therefore excessive at the time! Fact is, it is a very nice engine and more as only "nice to have".

Often, you regret having written something, and there’s no way to take it back. That’s my curse when I write with too much passion and end up doing others a disservice.

Thank you to the programmer from USA, Jay Honnold
https://github.com/jhonnold/berserk/

Today is May 25th, 2026 ... a good day for computer chess.
A new Berserk after a longer time, you know ...

Best
Frank

[Modern Times]

Yes I did not expect to see a new Berserk !

I've put a post in the JA thread to see if he may compile it for us, but in the meantime I'm also taking the opportunity to follow the instructions and see if I can learn something myself.

[Frank Quisinsky]

Hallo Moderne Zeit,

In the past, I once compiled the source code for a “non-chess project,” many years ago. That isn't my world. Nor is reading specifications. That's something for the younger generation or the experts to do. With all this material on engines, we often forget to play chess ourselves. It’s difficult to keep track of all the engines in the first place. So I tell myself, no more than an hour a day for organizational tasks. That’s how I automate processes to simply save time. I’m glad we have very good experts in many areas and I simply trust their work.

It's a good day!
I thought Berserk would be lost because of the Torch project. We would have lost one of the best chess programmers on the planet, even though we would have had a super-strong Berserk. We’ve lost far too many talented chess programmers. As far as I know, the reasons why so many have lost interest vary widely. But maybe the day will come when suddenly a new Seer, a new rofChade, or a new TheBaron, Xiphos, Laser... the list is very long will be available. That, too, is like a curse, I think, even for chess programmers... once you get into computer chess, you're hooked for life.

Best
Frank

[Modern Times]

Yes - I failed, I did get a Linux pgo compile easily, but doing a Windows one seems to be a job for the experts. At some point I'll do some learning, but in any case I wouldn't trust my own, will wait for JA.

[Modern Times]

The Linux pgo compile I made running under Ubuntu WSL is a lot faster than the supplied windows avx2-pext.exe. I look forward to the JA windows compile if he has time.

[Frank Quisinsky]

I don't dare ask Jim anymore. He must think we're all nuts, or wonder what on earth we're doing with all those chess programs. It's a good thing he doesn't know what we actually do with them.

Somehow, every community ... no matter what the topic ... is completely crazy. There are guys like Quisinsky who take it a bit too far. I hope he doesn't read this, or he'll start lecturing us again.

[Frank Quisinsky]

Hi Modern-Times,

what I like is the fight for the HCE crone.
I don't know why, but for me it's much more interesting as NN.
NN is boring me a bit. Too much perfection.

I don't want a world where everything is perfect.
To much AI, yesterday or for two / three days on my S24 comes also all the new AI stuff, GUI 8.5, horrible.
The reverse is true, it drives interests apart.
Computer chess has always provided good examples of this in all the years.

The HCE version of Berserk was important, really very very nice to have.

Best
Frank

[Krzysztof Grzelak]

Hi there,

I was surprised this morning, and I'm really happy about it.

Two new Berserk's are available!

v. 14
v. 4.7.0 ... interesting for the fight for the HCE crone!

In the past I criticized the high move average from Berserk.
But fact is that the version before this new release ... Berserk 250622 CE dev ... is on place 3 for my new ETOC system.
Why, with the fighting games for a draw vs. clearly weaker engines Berserk is really successfull. Also the number of quick wins is clearly better as in older versions. My criticism was therefore excessive at the time! Fact is, it is a very nice engine and more as only "nice to have".

Often, you regret having written something, and there’s no way to take it back. That’s my curse when I write with too much passion and end up doing others a disservice.

Thank you to the programmer from USA, Jay Honnold
https://github.com/jhonnold/berserk/

Today is May 25th, 2026 ... a good day for computer chess.
A new Berserk after a longer time, you know ...

Best
Frank

There is nothing to regret. That’s how it should be done, so that they understand what’s going on and don’t make such mistakes in future.

[Frank Quisinsky]

Hi there,

looking in my backup (I don't think the backup always contains everything from the engine history).

I found v.4.2.0 (210612), v4.3.0, v4.4.0, v4.5.1, v4.6.0 (210918).
I am thinking this are the HCE versions.

So the new update is v4.7.0 for HCE (version number is logical)!

v4.6.0 = 0.998.565 bytes = avx2-pext
v4.7.0 = 1.446.332 bytes = avx2-pext

It seems that Jay has maintained its HCE version, all the time!

Found v.6, v.7, v.8, v.8.5 and so one.
This seems to be the first NN versions.

For ETOC-G I will test of course the HCE version too.
Will do that in July 2026 with many other updates from tested engines.

Best
Frank

[Frank Quisinsky]

Hi Krzysztof,

to write about playing-styles is in most of cases a bit danger.
I know that and and I've already tried a lot of different ways to find out more about it through various statistics. Looking on quick wins and move average is very easy, looking in late mid-game, pawn structures, transposition into endgames is not easy. Incomplete statistics mingle with impressions, some of which are critical once again. That's enough to annoy any proud programmer. Ever since chess computers came along, we’ve been trying to personalize things, that’s just human nature. Our brains trick us into thinking we can categorize everything. I enjoy this. Not, of course, to annoy a chess programmer.

I often notice something and then find myself wondering why I thought that. What I had noticed was completely wrong. But that happens to us so often in life. The alternative is to avoid writing about controversial topics. Fact is "Playing-Styles" of engines is the most wonderful topic in computer chess, my opinion since the beginning of the good and old chess-computers.

But if I am wrong, I should give the information ... think so!

Best
Frank

In the past I am playing a tourney. I added in the tourney a lot of engines with a higher move-average. Many games over 300 moves, many engines with a move average from over 105 moves. And to watching such games again and again and lost interest in watching (all the boring endgames) ... so I wrote nonsens about this tournament, a not working contempt in case of Berserk and so on. Berserk is indeed a very hard fighter but with the different, too often succeeds in the battle for the half-point. That, too, is an art of programming I not understand in the past. Quick wins isn't all but a topic humans like a lot.