Sawfish 2TC

[OrgZ]

I will try it in my free time.

Thanks D

I'd like to "hear" your input

First, I quickly checked the changes in the source. They seem reasonable so as not to significantly change the style of the engine. I checked the engine positions and unfortunately I'm not satisfied. According to my tests, the engine is only slightly more tactical than Stockfish. I'd say it's simply tactically faster than Stockfish (to be precise, it's 22% faster per move with 100 seconds on one core), the same result I had with Sawfish 1.0, but it scored on more solved positions, almost like Blue Marlin 15.7. I got 98.9% similarity to Stockfish with Sawfish 2TC, and only 94.6% with Sawfish 1.0 on selected positions.

The changes you make may be good. They probably are, but I often choose weaker settings that change the way the engine plays.

Recently, it has become increasingly difficult to improve Stockfish derivatives. The last Stockfish I tested deeply from 7th of November 2023 plays very well tactically. Tactically, it performs as well as my Swordfish 15.4 and 15.5 and Blue Marlin 14.9 and 14.9a, but has a much higher Elo. Taking into account time and difficult positions solved, the probability that Swordfish 15.1 would show the correct move earlier than Stockfish on which it was based was about 96%. The last best version of the unreleased Swordfish 15.6, by comparison, had only 60% chance of solving the position faster than the current Stockfish. With a constant loss of about 10-15 Elo, it was unprofitable, because theoretically, with about 1/3 more time per move, Stockfish is already catching up and is clearly stronger. This is why Swordfish 15.6 was not released, and instead the clearly more aggressive Sawfish 1.0. However, these differences are no longer as large as they used to be for example a year ago. This is because the current Stockfish is already very polished.

Thank you for your detailed and open review D.
Well if you like 'em (changes) you can polish them and get it to where you will be satisfied......

I agree with you stockfish 071123 is quiet a jewel and well when it comes to puzzle solving am not really focused on that am interested in game play. Because the engine can solve puzzles but in game play not so much . Thats why we have people like me and you...those who provide solving engines and game play....its really difficult to balance both sides puzzle solving and game play one must suffer hey....

Anyway i was not trying to replace you engine but to share this little experiment. And Thanks for not overreacting like some would and thanks for your honest review

[dorsz]

when it comes to puzzle solving am not really focused on that am interested in game play.

That is true. I measure the strength in the last phase, when I am satisfied with the previous results. I once wanted to make an engine better than Stockfish, but despite about 20 changes that I thought were good, instead of being +10 Elo better, the engine was 2 Elo weaker. Then I switched to something else, not to compete with Stockfish strength, but with originality and perhaps a different perspective during the analysis.

Next time I try to release this engine I will look at some of your changes again. Are the time management changes your own or taken from another version of Stockfish?

[OrgZ]

Latest stockfish has better time management so i merged that with the one of sawfish trying to accommodate the tactical feature and the big net.

I like your engines very much, if you recall my first public derivative had some of swordfish patches yes....

So am just here to learn am not here to take someone's spot, but others think am taking their spot, am not interested in being number 1 noo am only interested in learning new things and i once said this on outskirts forum, and a kind person MZ guided me in some things and grateful for that....My only competition is my self, am competing with previous self (if previous me was able to make Raid vx better then current me must make Raid v2x better than the other one....thats my focus....

So if i look back i see progress in what i can see or do currently. We also have people like Mr Sedat, Homayoun, Anton etc alot who are very encouraging + yourself as well.

This community needs that....

[dorsz]

This community needs that....

Unfortunately, there are also threads full of arguments on this forum.

[OrgZ]

This community needs that....

Unfortunately, there are also threads full of arguments on this forum.

indeed. some of those arguments are unnecessary.
anyway Stay well

[Homayoun]

This community needs that....

Unfortunately, there are also threads full of arguments on this forum.

indeed. some of those arguments are unnecessary.
anyway Stay well

But they provide us fun corner too. Some of these arguments are really funny.

[OrgZ]

This community needs that....

Unfortunately, there are also threads full of arguments on this forum.

indeed. some of those arguments are unnecessary.
anyway Stay well

But they provide us fun corner too. Some of these arguments are really funny.

Ohhh so you enjoy some of these. I have to admit some are hilarious

[Homayoun]

This community needs that....

Unfortunately, there are also threads full of arguments on this forum.

indeed. some of those arguments are unnecessary.
anyway Stay well

But they provide us fun corner too. Some of these arguments are really funny.

Ohhh so you enjoy some of these. I have to admit some are hilarious

If you stay far and just watch the story, it would be funny, really funny but if you participate in their arguments then you will find it terrible.
Greetings

[OrgZ]

Test by Adrian

SF PB 021223 AVX2 vs. Sawfish 2TC_avx2

Processor AMD A10 – 9700 R7,10 Compute Cores 4C + 6G - 3.50 GHz - RAM 8.00 GB - 64-bit operating system - x64 processor - Windows 10 Pro - 1024 mb hash Fritz 16 GUI – Permanent Brain: OFF – Both Engines Play with 4 Threads and Large Pages - Opening Book: OliBiyu 10.4 Opening Suite ctg - Final Tablebases: Syzygy 3-4-5 pieces – NNUE: nn-0000000000a0.nnue for Both Engines

Book Settings:

1.Tournament ON
2.Min games 0-10
3.Variety of play 300
4.Influence of learn 100
5.Learning strength 500

DESKTOP-TORM2PL, Blitz 4min+1seg 0


1 Sawfish 2TC_avx2 +6/=288/-6 50.00% 150.0/300 6116.00
2 SF PB 021223 AVX2 +6/=288/-6 50.00% 150.0/300 6116.00

Games in PGN: https://www.mediafire.com/file/9w6w1h74 ... 9.pgn/file

[dorsz]

Congratulations OrgZ!