Increase reduction at root when the best move does not change frequently #3225
Conversation
|
Congratulations! |
|
@FauziAkram Thanks. NNUE = true tests have high draw rate around 90% unlike NNUE = false tests, see #3148 |
src/search.cpp
Outdated
| r -= 2; | ||
|
|
||
| // Increase reduction at root and non-PV nodes when the best move does not change frequently | ||
| if (!(PvNode && !rootNode) && depth > 10 && thisThread->bestMoveChanges <= 2) |
There was a problem hiding this comment.
To me
if ((rootNode || !PvNode) && depth > 10 && thisThread->bestMoveChanges <= 2)
looks much more easy to read/understand.
There was a problem hiding this comment.
it also corresponds 1:1 to what the comment says
|
@SFisGOD I lost the proper author in git in squashing the commit, and it slipped through the cracks.. sorry for that.. |
|
@vondele @snicolet Is it possible to amend the author of last commit? It is like my only source of dopamine these days. Sad to know that I can't see it listed here https://github.com/official-stockfish/Stockfish/commits?author=SFisGOD |
|
Unfortunately not on master (i.e. no force pushes on master). I had the rebase/squash the branch on the previous commit as the pull request had the wrong base. In resolving the conflicts I lost the proper author... sorry for that. |
|
@vondele I hope it will not happen again. It's rare to have an Elo gainer to pass so should be more careful in my opinion. |
|
yes, I'll try of course. But in that case please help by having your PRs and tests against master only, not against other pending PRs, and condense them into a single commit. The alternative is that I ask everybody to rebase their PRs on master, and only accept those that are a single commit on top of latest master. I'm trying to make it easier for contributors by doing some manual work on their PRs. Sometimes mistakes happen. |
|
@vondele Yes, maybe you should have told me to do so and I would have gladly complied. Anyway, lesson learned so I will not hurry in making a PR anymore. I will just wait until all pending PRs are merged and then I'll open my pull request. That way, it will be easier for you. |
|
yes, same here, but let's not worry too much about it. After all, you made the contribution, it progressed the project, which is what really counts. |
when the best move does not change frequently
STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 51320 W: 5159 L: 4956 D: 41205
Ptnml(0-2): 215, 3897, 17242, 4082, 224
https://tests.stockfishchess.org/tests/view/5faa072367cbf42301d6a767
LTC:
LLR: 2.98 (-2.94,2.94) {0.25,1.25}
Total: 15952 W: 762 L: 642 D: 14548
Ptnml(0-2): 8, 561, 6725, 667, 15
https://tests.stockfishchess.org/tests/view/5faa4c3567cbf42301d6a794
closes #3225
Bench: 3954692
|
@snicolet Thank you so much snicolet! |
|
unless this breaks abrok, fishtest and other workflows out there... |
|
at least abrok doesn't seem to update to this commit... |
|
fishtest will presumably drain, as all workers reference master with a commit that has now been erased. |
|
If I have broken anything, I'll take the blame, so at least both SFisGOD and yourself can have a third person to direct the bitterness to 🙂. Maybe there is a chance that fishtest does not drain, however, as fishtest workers pull the sources to compile from individual user commits rather than from the official-stockfish master branch? For instance, it seems that the commit in question was with a SHA beginning with But if I have broken something, could I undo the force-push to restore the previous commit in the master branch? |
|
We'll have to see... basically I don't know how this will impact all things. Abrok still seems stuck, fishtest might survive on the cached binaries, or by finding back somehow the commit that was erased. Reverting with another force push... I have no idea However, I hope I'm not sounding too bitter, I'm just a little, because of the time lost for such a non-issue. |
|
great find, 2.5 Elo gain does not around come every day ... |
Passed STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 171152 W: 34715 L: 34202 D: 102235
Ptnml(0-2): 3278, 20155, 38228, 20606, 3309
https://tests.stockfishchess.org/tests/view/5fa861f467cbf42301d6a68e
Passed LTC:
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 149616 W: 20471 L: 19882 D: 109263
Ptnml(0-2): 1172, 14434, 43102, 14833, 1267
https://tests.stockfishchess.org/tests/view/5fa9c8ff67cbf42301d6a74f
closes #3226
Bench: 3597730
when the best move does not change frequently
STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 51320 W: 5159 L: 4956 D: 41205
Ptnml(0-2): 215, 3897, 17242, 4082, 224
https://tests.stockfishchess.org/tests/view/5faa072367cbf42301d6a767
LTC:
LLR: 2.98 (-2.94,2.94) {0.25,1.25}
Total: 15952 W: 762 L: 642 D: 14548
Ptnml(0-2): 8, 561, 6725, 667, 15
https://tests.stockfishchess.org/tests/view/5faa4c3567cbf42301d6a794
closes official-stockfish/Stockfish#3225
Bench: 3954692
10 participants
STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 51320 W: 5159 L: 4956 D: 41205
Ptnml(0-2): 215, 3897, 17242, 4082, 224
https://tests.stockfishchess.org/tests/view/5faa072367cbf42301d6a767
LTC:
LLR: 2.98 (-2.94,2.94) {0.25,1.25}
Total: 15952 W: 762 L: 642 D: 14548
Ptnml(0-2): 8, 561, 6725, 667, 15
https://tests.stockfishchess.org/tests/view/5faa4c3567cbf42301d6a794
Bench: 3954692