New engine releases 2020

[Guenther]

Sorry to hear this, I guess no compile works for you then, neither mine, nor Daniels release?
It seems strange though, as my Q8200 and your T6600 seem both Core2 and therefore SSE4.1?
I have also Win7 here (Ultimate).

I am using Windows 7 Home Premium. I did get it working now, by using Windows 7 compatibility mode. Thanks to sarona for the tip. 621000 nps at depth 15. The same compatibility setting also allowed me to run Daniel's compile.

Great. I tried the previously released compile by Daniel too that way and it also runs here in compatibility mode.
It's strange to be forced to enable Win7 compatibility mode on Win7. (even more, as this was the day Win7 support officially ended) :)

I will remove my compilations.

[abulmo2]

I just release Amoeba 3.1. It fixes a few bugs and probably adds a bunch of others. It should be about 30 Elo stronger than the previous version
https://github.com/abulmo/amoeba/releases/tag/v3.1
Have fun with it.

[tmokonen]

Great. I tried the previously released compile by Daniel too that way and it also runs here in compatibility mode.
It's strange to be forced to enable Win7 compatibility mode on Win7. (even more, as this was the day Win7 support officially ended)

I will remove my compilations.

Because Win7 support has ended, I have spent the entire day backing up and reluctantly "upgrading" this machine to Windows 10, and will not be able to test for Win7 compatibility any more.

[tmokonen]

Bagatur 1.9a and some sort of DEV version are available:
https://github.com/bagaturchess/Bagatur ... ads/Engine

Kouri 1.15
https://github.com/hamzainaan/Kouri/releases/

Sugar-NN 150120
https://github.com/Zerbinati/SugaR-NN/releases

FoxSEE 0.2.0
https://github.com/redsalmon91/FoxSEE/releases

[Guenther]

Kouri 1.15
https://github.com/hamzainaan/Kouri/releases/

This is Vice plus enhancement from an open source Lichess bot, also derived from Vice,
so nothing original in it. (forking SF, DefenChess, RubiChess, Ethereal, Laser will make
things not better in the future)

FoxSEE 0.2.0
https://github.com/redsalmon91/FoxSEE/releases

Needs checking - looks familiar, but my guess is that it is 'modelled' after another already available rust program,
w/o mentioning this.

[Guenther]

FoxSEE 0.2.0
https://github.com/redsalmon91/FoxSEE/releases

Needs checking - looks familiar, but my guess is that it is 'modelled' after another already available rust program,
w/o mentioning this.

It seems what looked familiar was the usual CPW influence - I will add it to my chronology for now

OTH the rust chess engines 'market' exploded...
(there was already TSCP + Vice + SF in rust, but now there are dozens of new ones to check)
https://github.com/search?o=desc&q=rust ... positories

I hope for this one in near future, uci is planned for 1.0 milestone
https://github.com/ColonelPhantom/fe_chess

[jonkr]

Slow Chess Blitz Classic 1.8 is now available : https://www.3dkingdoms.com/chess/slow.htm

- Thoroughly reworked a lot of code, and tweaked eval, but not sure what specific to mention.
- Further increased value of king safety/checks, pawn promo threats, and some other threats. (Realized with modern search this helps to follow threats even if values seem too high to me.)
King safety is getting pretty aggressive for near term danger, but accuracy isn't great. And still needs work building to an attacking position (and also to be better at building positional advantages to press, which sometimes creates weaknesses leading to attack.)
- Optimized code to be about 20% faster + changed Threat Moves to detph > 3 to minimize slow-down and be more about not missing something. Seems the hard-coded threat move detection is only a small elo increase at this point, though hard for me to measure.
- Did a code cleanup pass to prepare for multi-threading and code portability. Maybe version 1.9 will support multi-threaded search.

I realized I still don't do nearly enough testing to estimate strength accurately, I was trying lots of changes for no consistent improvement for last couple weeks. (Lot of variation in 500 game matches, then even much more when I added variety of opponents), but it was always stronger than 1.7, with a few scattered extra good results. (Realistically hoping for around +35 elo, though did have some better runs.)

My current plan is trying to add parallel search for a version 1.9 that probably will otherwise be similar to 1.8. That seems like a good way to close out the modernization of Slow Chess for the Classic for the 1.x series.

[Alayan]

I strongly suggest you to move towards the statistical techniques that have proven successful for so many engines to evaluate patches. SPRT is the most reliable way of assessing improvement without excessive resource usage.

Methods that stick to low sample sizes get more and more ineffective as the engine's strength increase, and testing against other engines dilutes your already limited hardware resources.

[jonkr]

Assuming I continue to a Classic 2.x series, I probably will try to be more scientific about testing (and also get a 16 or 32 core machine to at least have some resources for it.) So far I have mainly been trying ideas that seem interesting to code and re-writing old code (the Blitz wv2.1 code was a horrible mess in many aspects.) and watching games and sometimes inspecting individual positions, evaluation values, search trees and some stats.

For smaller individual patches I figured I have no hope of measuring, so have been only doing full version testing to verify some improvement. But playing a match with a small number of games (eg. 500) and getting results near the edge of expected range like +50 elo, then +15 elo, then +35 elo was the source of some frustration when I was hoping for +50. And of course bringing in other similar rated programs to try to verify only added to the confusion.

[Roland Chastain]

Slow Chess Blitz Classic 1.8 is now available : https://www.3dkingdoms.com/chess/slow.htm

Good news! I hadn't noticed the release of v1.7.

- Did a code cleanup pass to prepare for multi-threading and code portability. Maybe version 1.9 will support multi-threaded search.

For me, I would like to have a Linux version of the engine.

See you soon! I go to test the two latest versions.