Is Fire now number 3. CPU engine?

[Jouni]

I think it is. But not playing in CCC or TCEC. Why?

[Chessqueen]

I think it is. But not playing in CCC or TCEC. Why?

Is it behind Ice

[Damir]

I think it is. But not playing in CCC or TCEC. Why?

Most probably is because it is not using own network but Stockfish one instead..

[Madeleine Birchfield]

I think it is. But not playing in CCC or TCEC. Why?

Fire is like John Woe's Mayhem, the author took the evaluation straight from Stockfish 12 dev. Thus neither is on the rating lists (CCRL, FGRL, GRL) or in any of the tournaments, like TCEC or CCC, as it runs afoul of their uniqueness rules.

Fire also took its search from Stockfish 9 dev, so for all points and purposes, Fire is Stockfish 9 dev NNUE, and is a derivative of Stockfish like asmfish and Brainlearn.

[Vernon Crawford]

I think it is. But not playing in CCC or TCEC. Why?

Fire is like John Woe's Mayhem, the author took the evaluation straight from Stockfish 12 dev. Thus neither is on the rating lists (CCRL, FGRL, GRL) or in any of the tournaments, like TCEC or CCC, as it runs afoul of their uniqueness rules.

Fire also took its search from Stockfish 9 dev, so for all points and purposes, Fire is Stockfish 9 dev NNUE, and is a derivative of Stockfish like asmfish and Brainlearn.

My understanding is that Fire uses a NN from trained by Sergio Vieri...it was tested in Fishtest but rejected for use in any release.
(ie although it was created with SF games, it's certainly related to SF, clearly not the same evaluation as SF12).

Looking at the source code, it's search is significantly different than SF9. When running it reaches much lower depths that SF 9.
Of course there are similarities, but similarities to SF would qualify many many (or most) engines as derivatives. SF derivatives like Brainlearn and others often offer new features (experience etc.) but on the whole are barely changed or rewritten...this is clearly not the case with Fire.

[connor_mcmonigle]

I think it is. But not playing in CCC or TCEC. Why?

Here's TCEC's stance on Fire:

On behalf of TCEC I want to formally acknowledge that the later versions of Houdini and Fire that played at TCEC have unfortunately been found to covertly contain code copied and taken from other engines, and these two engines have therefore been removed from further participation as well as from those past final standings in which they took a prize place." -kan aka kanchess 20 Dec 2020

Furthermore, as Fire currently uses an old Stockfish network (trained by SV on Stockfish data using tools developed for and by the Stockfish project) + Stockfish inference code, it is considered far too similar to Stockfish for participation on those grounds alone.

For similar reasons, Fire does not participate at CCC.

[dkappe]

I think it is. But not playing in CCC or TCEC. Why?

Here's TCEC's stance on Fire:

On behalf of TCEC I want to formally acknowledge that the later versions of Houdini and Fire that played at TCEC have unfortunately been found to covertly contain code copied and taken from other engines, and these two engines have therefore been removed from further participation as well as from those past final standings in which they took a prize place." -kan aka kanchess 20 Dec 2020

Furthermore, as Fire currently uses an old Stockfish network (trained by SV on Stockfish data using tools developed for and by the Stockfish project) + Stockfish inference code, it is considered far too similar to Stockfish for participation on those grounds alone.

For similar reasons, Fire does not participate at CCC.

Birchfield writes too many things that are outright or partially wrong, though I don’t think it’s out of malice. Connor’s posts are generally accurate and well researched. So I think we can wipe away the “Fire uses SF9 search” business, as I don’t think that’s a version that’s been cited before.