TalkChess.com

When you get that many Sprt(0,5) greens ( passed) in three days - good things are happening...plus having contributer Voyager One back in the fold is awesome! He's one of the most productive SF contributors with very creative and ELO gaining patches.

The most interesting patch for me in the above snapshot is by our Joe Ellis - that one is a game changer at LTC from my testing at 90/9 (90 sec/.9 sec) :
The only difference between SF-McBv2.5 and SF-McB2.5se is that the se version is missing the Joe Ellis' patch.

SF-McB 2.5 will be released tomorrow ( source and all flavors of exe's) , immediately after HGM's Saturday tourney ( note the current source on Github is not the latest source - although it was updated since the last release and it does say v2.5 - this is not the v2.5 will be released tomorrow. Github is updated nearly simultaneously when I make a release announcement.

SF-McBrain-v2.5 will have most of the patches above , including the Joe Ellis patch and all of the usual McB features ( cerebellum book, Matefinder option etc.). I have also gone scavenging into the SF archives for some failed patches which improve analysis where the current SF fails on certain positions ( we know which ones they are - they are posted here on a regular basis) - and updated them for today's SF code - (although some of these do hurt ELO a tad) .

https://groups.google.com/forum/?fromgr ... ng_results

The SF base for tomorrow's release is the current dev- SF June 6th branch:
https://github.com/official-stockfish/S ... 4c60ba2705

Look for the annoucnement around 6 PM tomorrow EDST.

The final results on that little run - and it was only ran once - there was no cherry picking as is sometimes alleged. Also keep in mind, if there was any ELO gain here the result of the changes, most ( maybe all) of those Elo gaining changes made are from the SF team and our still going the review process. The SF teams takes their work very seriously and not do commit patches unless that are 99.9% sure the patch is good - either as an Elo gaining or a simplification. I'm not tied to those constraints. Likewise I also have the liberty to sacrifice a tiny amount of ELO to help with analysis in those hard to find the correct move positions.
The games :

https://www.dropbox.com/s/25ypfsdxy08yj ... 7.pgn?dl=1
Openings were TCEC #8 Super Final

Link to download engine, please.

shrapnel wrote:Link to download engine, please.

Tomorrow my friend - by 6PM - as soon as HGM's Monthly Saturday Tourney is over.

I see nothing great

These patches failed at LTC except the last patch that will probably also fail and has negative LLR at this moment.

12-06-17
sn
futility_relax3
diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 18145 W: 2255 L: 2315 D: 13575
sprt @ 60+0.6 th 1
LTC: Relax futility condition, take 4

12-06-17
pb
bishop_knight_ex
diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20244 W: 2616 L: 2668 D: 14960
sprt @ 60+0.6 th 1
LTC: Take 2: Handle Bishop Knight trade as good-capture in dynamic situations (on recapture)

13-06-17
El
nmp_mvc
diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 23614 W: 3072 L: 3113 D: 17429
sprt @ 60+0.6 th 1
Take 2 (LTC)

13-06-17
Sa
FutilityMargin
diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 74082 W: 9623 L: 9578 D: 54881
sprt @ 60+0.6 th 1
Tweaking parameters for futility tuning.

14-06-17
sg
king_safety_tuned
diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 51770 W: 6806 L: 6755 D: 38209
sprt @ 60+0.6 th 1
LTC: Final tuned values after nearly 120K games

15-06-17
Sa
SingularNonPVDepth
diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 40285 W: 5160 L: 5148 D: 29977
sprt @ 60+0.6 th 1
Tweak depth of singular search for non-pv

great fine, everybody is only helping SF to become near 4000 elo, bad that nobody is helping other engines to improve is engine.... thats pity....

I don't think the lack of ideas is an issue. You can browse this forum and find many novelties. Everybody can now use Texel tuning and get a reasonable good eval. At this point most of the Stockfish changes are not far from random changes.

Stockfish's advantage is the Fishtest platform that decouples resources from developers. All other engines require developers with enough testing resources. Just adapt fishtest to your engine and see whether your statement still stands.

Hmm,

I see only between Jun 6th and Jun 16th some non-functional patches,

and after the 16th of June no other patch appeared.

Paloma wrote:Hmm,

I see only between Jun 6th and Jun 16th some non-functional patches,

and after the 16th of June no other patch appeared.

I jumped the gun - I saw a bunch of greens, but most failed LTC. My bad.