Is anything new here?
https://www.newinchess.com/fritz-21
Fritz 21
Moderator: Ras
-
peter
- Posts: 3595
- Joined: Sat Feb 16, 2008 7:38 am
- Full name: Peter Martan
Re: Fritz 21
New engine (Frank Schneider's Ginkgo- Fritz again), cb- advertising says 40 Elo stronger, which I already can confirm from a single little head to head match with 3'+1" single thread (4.3GHz), UHO 85-94cp 6mvs:
Score of Fritz21 vs Fritz20: 138 - 87 - 275 [0.551]
Elo difference: 35.6 +/- 20.4, LOS: 100.0 %, DrawRatio: 55.0 %
500 of 500 games finished.
Additional to update- bonus for F18, F19, F20- customers 25% discount today (only)
Score of Fritz21 vs Fritz20: 138 - 87 - 275 [0.551]
Elo difference: 35.6 +/- 20.4, LOS: 100.0 %, DrawRatio: 55.0 %
500 of 500 games finished.
Additional to update- bonus for F18, F19, F20- customers 25% discount today (only)
Peter.
-
reflectionofpower
- Posts: 1665
- Joined: Fri Mar 01, 2013 5:28 pm
- Location: USA
Re: Fritz 21
I bought it today as an upgrade. Only $38. Good deal.
"Without change, something sleeps inside us, and seldom awakens. The sleeper must awaken." (Dune - 1984)
Lonnie
Lonnie
-
cc2150dx
- Posts: 466
- Joined: Sat Nov 30, 2013 9:51 am
- Full name: Jason Coombs
Re: Fritz 21
I really don't like the 'ChessBase Product Installer'. I sent them an email to send me the offline installers (download links) so I can copy them to USB etc. We'll see how that goes.
-
Murat
- Posts: 174
- Joined: Thu Mar 09, 2006 5:57 am
- Location: Canada
Re: Fritz 21
You can select to keep the installation files after the download during installation.
-
Frank Quisinsky
- Posts: 7411
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: Fritz 21
Peter,
in self-play +33,4 is not a lot!
500 games is nothing.
I think you are using the time control 1+1 (GUI need also time).
Unclear is against which version of Fritz 20 (two are available).
But if +33,4 Elo in selfplay, gain can be 30%
So 10 Elo is a lot, but nobody in the World have the feeling for 10 Elo more.
Elo is increasingly becoming useless information that nobody knows what to do with anymore. At most, maybe a few computer geeks out there, whatever they do with it, that is the secret every brain build with other prioritys.
With other words ...
You could have saved yourself the trouble writing this post; as always, I can't make head or tail of these results.
Elo is really bad.
At the skill levels we've reached, it just drives people crazy.
Best
Frank
in self-play +33,4 is not a lot!
500 games is nothing.
I think you are using the time control 1+1 (GUI need also time).
Unclear is against which version of Fritz 20 (two are available).
But if +33,4 Elo in selfplay, gain can be 30%
So 10 Elo is a lot, but nobody in the World have the feeling for 10 Elo more.
Elo is increasingly becoming useless information that nobody knows what to do with anymore. At most, maybe a few computer geeks out there, whatever they do with it, that is the secret every brain build with other prioritys.
With other words ...
You could have saved yourself the trouble writing this post; as always, I can't make head or tail of these results.
Elo is really bad.
At the skill levels we've reached, it just drives people crazy.
Best
Frank
-
cc2150dx
- Posts: 466
- Joined: Sat Nov 30, 2013 9:51 am
- Full name: Jason Coombs
Re: Fritz 21
-
nmcrazyim5
- Posts: 29
- Joined: Sat Jan 28, 2023 12:03 am
- Full name: P
Re: Fritz 21
Have they added any GPU bound Raytracing to the 3D boards?
-
peter
- Posts: 3595
- Joined: Sat Feb 16, 2008 7:38 am
- Full name: Peter Martan
Re: Fritz 21
Frank, as you could have read on your own, I wrote in the posting you answered to, that TC was 3'+1" single thread, and I even gave the clock rate. 500 games aren't much, but as you can see in result too, LOS of 100% was achieved, so for the single match you already can see the statistical relevance of its own, can't you? So 500 games of that kind aren't just "nothing", unless Elo doesn't mean anything to you anymore at all, as you write,Frank Quisinsky wrote: ↑Thu May 21, 2026 5:47 am in self-play +33,4 is not a lot!
500 games is nothing.
I think you are using the time control 1+1 (GUI need also time).
but in that case 500.000 games don't mean a thing to you neither, if you measure the results in Elo one more time, and yes I could have saved myself and others the posting then, but so you could have done as well, couldn't you?Frank Quisinsky wrote: ↑Thu May 21, 2026 5:47 am You could have saved yourself the trouble writing this post; as always, I can't make head or tail of these results.
Elo is really bad.
Of course this single one little match is only more or less comparable to bigger sample size with more different TCs against selective pools of opponents as well as with different testsets or books (wrote about the test- set I used too, you see?), but for that aim we have the big lists anyhow, Fritz 21 for sure will get much testing from many testers and I don't want to compete thereby, I just wanted to have a first one look of my own at the new engine. More testing of very special ways of mine you probably already saw at CSS too, didn't you?
Peter.
-
Frank Quisinsky
- Posts: 7411
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: Fritz 21
Yes, Elo is actually less than nothing.
Since many years ... engines are to strong. For humans not really important ... engine have 2800 or 3500 Elo. The style is interesting.
What are your thoughts after watching the match?
If we compare the style of Gingko with other stronger engines I think problem are different holes in the late mid-games, tactical holes. The endgame is around the same strength stronger engines have. Quantity of short wins is OK, move average is good. Maybe a bit to many to fast draw games. Thats my impressions after the first games. Double-Pawn ... I saw to many games the double-pawn is a problem.
But indeed:
Gingko is one of the engines with really an interesting style. Gingko forced open-lines, that's is strong!! Pawns are aggressive, good too. Only the double-pawns. But what I saw in the games, can be pure imagination.
To Elo:
If Chessbase say +40 does that mean it would be good if it were half?
Thats a problem, for a group of people TheKing is playing with 3000 Elo, for an other group of people TheKing is playing with 2500 Elo and that's the same with Elo gain. There are more test combinations than there are atoms in the universe. Again, Elo is less than nothing! More nothing then anything ... NOTHING!
ETOC-G test-run is still running.
So, ETOC-G produced Elo is also nothing!
Best
Frank
3+1 is OK, not saw that ... sorry!
Since many years ... engines are to strong. For humans not really important ... engine have 2800 or 3500 Elo. The style is interesting.
What are your thoughts after watching the match?
If we compare the style of Gingko with other stronger engines I think problem are different holes in the late mid-games, tactical holes. The endgame is around the same strength stronger engines have. Quantity of short wins is OK, move average is good. Maybe a bit to many to fast draw games. Thats my impressions after the first games. Double-Pawn ... I saw to many games the double-pawn is a problem.
But indeed:
Gingko is one of the engines with really an interesting style. Gingko forced open-lines, that's is strong!! Pawns are aggressive, good too. Only the double-pawns. But what I saw in the games, can be pure imagination.
To Elo:
If Chessbase say +40 does that mean it would be good if it were half?
Thats a problem, for a group of people TheKing is playing with 3000 Elo, for an other group of people TheKing is playing with 2500 Elo and that's the same with Elo gain. There are more test combinations than there are atoms in the universe. Again, Elo is less than nothing! More nothing then anything ... NOTHING!
ETOC-G test-run is still running.
So, ETOC-G produced Elo is also nothing!
Best
Frank
3+1 is OK, not saw that ... sorry!