Fire 021819 is surprisingly strong

Discussion of computer chess matches and engine tournaments.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Jouni
Posts: 2121
Joined: Wed Mar 08, 2006 7:15 pm

Fire 021819 is surprisingly strong

Post by Jouni » Wed Feb 26, 2020 7:24 am

It's currently at 2. place at TCEC and CCC :o ! I have not installed it as being too weak engine.
Jouni

Gary Internet
Posts: 45
Joined: Thu Jan 04, 2018 6:09 pm

Re: Fire 021819 is surprisingly strong

Post by Gary Internet » Wed Feb 26, 2020 12:17 pm

The current CCC filler games are played at short time control (1+1) which suits Fire and basically cripples Komodo. We all know Komodo is far stronger than Fire. Plus the version of Ethereal being used is 11.80 which is far from the current 11.99 (soon to be formally released when one more patch takes it to version 12.00).

The current TCEC League One games shouldn’t be taken too seriously as the sample size is even smaller than CCC and most of the field is significantly weaker than Fire. Its performance is not at all surprising if you look at where Fire 7.1 ranks on ratings lists compared with the likes of Winter, Nemorino, Vajolet and Arasan.

In the next round of TCEC and the upcoming CCC13, Fire will play against a much smaller selection of opponents in both tournaments where the weakest engine it will face will be Xiphos 0.6.1. It will be a much better indication of how strong it is.

I wasn’t aware that version 021819 was publicly available. The only places I have seen it are TCEC and CCC. I’ve not actually seen that version appear on rating lists so I have no idea how much stronger it is compared with Fire 7.1. I did hear at one time that it might only be 5 Elo stronger.

Jouni
Posts: 2121
Joined: Wed Mar 08, 2006 7:15 pm

Re: Fire 021819 is surprisingly strong

Post by Jouni » Wed Feb 26, 2020 1:20 pm

May be Fire needs minimum 90 cores to be competitive? In TCEC it uses 128 cores and now has rating +331 [3944]. Can't be totally sss.
Jouni

Gary Internet
Posts: 45
Joined: Thu Jan 04, 2018 6:09 pm

Re: Fire 021819 is surprisingly strong

Post by Gary Internet » Wed Feb 26, 2020 4:57 pm

I quite often wonder about the effects of small sample size and if they’re ever exaggerated.

In TCEC Cup 2 (the old 43 core setup)

Fire 011819 (which is only 1 month younger than the version you’re talking about)

Played 8 games against Jonny 8.1. The score was +1=7-0 in favour of Fire. Hardly an astounding show of strength against a much older and weaker engine. The rules state that all 8 games will be played, so if Fire was actually strong it surely would have won something like +5=3-0

In the next round Fire 011819 played against Houdini 6.03. The score was +1=9-0 in favour of Houdini. Although Fire lost it was an impressive performance holding out for so long against a much stronger engine. Again, the score could have been much worse. If you look at FastGM’s 10+6 rating list with a sample of 300 games, Houdini 6.03 beats Fire 7.1 over 300 games by a score of +110=186-4

In the next round Houdini 6.03 beat Stockfish 141019 (a version about 6 weeks out from the release of Stockfish 10) by a score of +1=11-0. If we go back to FastGMs 10+6 rating list, Stockfish 051019 (near enough the same version) beats Houdini 6.03 by a score of +90=204-6.

In TCEC Cup3 Fire 021819 (the version you’re talking about) beat Xiphos 0.5.3 by a score of +1=7-0. Pretty close.

Then is TCEC Cup 4 Xiphos 0.5.6 (only 3 more patches over 0.5.3) beat Fire 021819 by a score of +1=7-0.

The score between Fire 7.1 and Xiphos 0.6 on FastGMs 10+6 list is +40=221-39 in favour of Xiphos.

If you look at the small sample sizes from the TCEC Cup, you could make the leap that both Fire 011819 and Houdini 6.03 were pretty much the same strength as Stockfish from October 2019.

I think Fire 021819 is going to have a fight on its hands at both TCEC and CCC in the near future.

RubiChess
Posts: 168
Joined: Fri Mar 30, 2018 5:20 am

Re: Fire 021819 is surprisingly strong

Post by RubiChess » Wed Feb 26, 2020 6:42 pm

Jouni wrote:
Wed Feb 26, 2020 1:20 pm
May be Fire needs minimum 90 cores to be competitive? In TCEC it uses 128 cores and now has rating +331 [3944]. Can't be totally sss.
The +331 Elo at TCEC is compared to FireNN which played (badly) last season and is a completely different engine.

Alayan
Posts: 254
Joined: Tue Nov 19, 2019 7:48 pm
Full name: Alayan Feh

Re: Fire 021819 is surprisingly strong

Post by Alayan » Wed Feb 26, 2020 8:36 pm

The TCEC live elo is broken and useless.

After a game, it uses the ratings of both engines at the start of the league to compute an elo gain/loss for each. This means that underrated engines end up with an inflated live elo, and overrated engines with a too low live elo.

Fuddur
Posts: 49
Joined: Sun Mar 18, 2018 5:35 am

Re: Fire 021819 is surprisingly strong

Post by Fuddur » Fri Mar 06, 2020 3:06 am

Fire must be available for Android ARM64.
Android chess lover waiting for it.
Thanks

Jouni
Posts: 2121
Joined: Wed Mar 08, 2006 7:15 pm

Re: Fire 021819 is surprisingly strong

Post by Jouni » Tue Mar 10, 2020 9:37 am

Still going strong https://tcec-chess.com/!
Jouni

Jouni
Posts: 2121
Joined: Wed Mar 08, 2006 7:15 pm

Re: Fire 021819 is surprisingly strong

Post by Jouni » Wed Mar 11, 2020 6:20 pm



In the move 76. Fire has M69 and and Komodo 1,35 :!: :?:
Jouni

Post Reply