Introduce futility margin array. #2270
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Nice. BTW, you can make the futility_margin[][] constexpr. |
|
Line 75 there are two extra spaces at the beginning. Line 800 there's an extra space between ] and [. |
|
I guess it should be FutilityMargin[] now instead of futility_margin(). |
anshulongithub
referenced
this pull request
in protonspring/Stockfish
Aug 19, 2019
snicolet
pushed a commit
that referenced
this pull request
Aug 21, 2019
@Vizvezdenec array suggested that alternate values may be better than current master (see pull request #2270 ). I tuned some linear equations to more closely represent his values and it passed. These futility values seem quite sensitive, so perhaps additional Elo improvements can be found here. STC LLR: 2.95 (-2.94,2.94) [0.50,4.50] Total: 12257 W: 2820 L: 2595 D: 6842 http://tests.stockfishchess.org/tests/view/5d5b2f360ebc5925cf1111ac LTC LLR: 2.96 (-2.94,2.94) [0.00,3.50] Total: 20273 W: 3497 L: 3264 D: 13512 http://tests.stockfishchess.org/tests/view/5d5c0d250ebc5925cf111ac3 Closes #2272 ------------------------------------------ How to continue from there ? a) we can try a simpler version for the futility margin, this would be a simplification : margin = 188 * (depth - improving) b) on the other direction, we can try a complexification by trying again to gain Elo with an complete array of futility values. ------------------------------------------ Bench: 4330402
|
Since equasion is merged I'm closing this one. |
|
@Vizvezdenec we shud retest the implementation here with a normal [0..4] procedure against the new master now that the equation has been committed to see how this fares against it. |
MichaelB7
pushed a commit
to MichaelB7/Stockfish
that referenced
this pull request
Aug 23, 2019
@Vizvezdenec array suggested that alternate values may be better than current master (see pull request official-stockfish#2270 ). I tuned some linear equations to more closely represent his values and it passed. These futility values seem quite sensitive, so perhaps additional Elo improvements can be found here. STC LLR: 2.95 (-2.94,2.94) [0.50,4.50] Total: 12257 W: 2820 L: 2595 D: 6842 http://tests.stockfishchess.org/tests/view/5d5b2f360ebc5925cf1111ac LTC LLR: 2.96 (-2.94,2.94) [0.00,3.50] Total: 20273 W: 3497 L: 3264 D: 13512 http://tests.stockfishchess.org/tests/view/5d5c0d250ebc5925cf111ac3 Closes official-stockfish#2272 ------------------------------------------ How to continue from there ? a) we can try a simpler version for the futility margin, this would be a simplification : margin = 188 * (depth - improving) b) on the other direction, we can try a complexification by trying again to gain Elo with an complete array of futility values. ------------------------------------------ Bench: 4330402
mstembera
pushed a commit
to mstembera/Stockfish
that referenced
this pull request
Aug 29, 2019
@Vizvezdenec array suggested that alternate values may be better than current master (see pull request official-stockfish#2270 ). I tuned some linear equations to more closely represent his values and it passed. These futility values seem quite sensitive, so perhaps additional Elo improvements can be found here. STC LLR: 2.95 (-2.94,2.94) [0.50,4.50] Total: 12257 W: 2820 L: 2595 D: 6842 http://tests.stockfishchess.org/tests/view/5d5b2f360ebc5925cf1111ac LTC LLR: 2.96 (-2.94,2.94) [0.00,3.50] Total: 20273 W: 3497 L: 3264 D: 13512 http://tests.stockfishchess.org/tests/view/5d5c0d250ebc5925cf111ac3 Closes official-stockfish#2272 ------------------------------------------ How to continue from there ? a) we can try a simpler version for the futility margin, this would be a simplification : margin = 188 * (depth - improving) b) on the other direction, we can try a complexification by trying again to gain Elo with an complete array of futility values. ------------------------------------------ Bench: 4330402
pb00068
pushed a commit
to pb00068/Stockfish
that referenced
this pull request
Sep 10, 2019
@Vizvezdenec array suggested that alternate values may be better than current master (see pull request official-stockfish#2270 ). I tuned some linear equations to more closely represent his values and it passed. These futility values seem quite sensitive, so perhaps additional Elo improvements can be found here. STC LLR: 2.95 (-2.94,2.94) [0.50,4.50] Total: 12257 W: 2820 L: 2595 D: 6842 http://tests.stockfishchess.org/tests/view/5d5b2f360ebc5925cf1111ac LTC LLR: 2.96 (-2.94,2.94) [0.00,3.50] Total: 20273 W: 3497 L: 3264 D: 13512 http://tests.stockfishchess.org/tests/view/5d5c0d250ebc5925cf111ac3 Closes official-stockfish#2272 ------------------------------------------ How to continue from there ? a) we can try a simpler version for the futility margin, this would be a simplification : margin = 188 * (depth - improving) b) on the other direction, we can try a complexification by trying again to gain Elo with an complete array of futility values. ------------------------------------------ Bench: 4330402
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
passed LTC №1 with [0.5, 4.5] bounds
http://tests.stockfishchess.org/tests/view/5d5418090ebc5925cf109c26
LLR: 2.95 (-2.94,2.94) [0.50,4.50]
Total: 85676 W: 14530 L: 14034 D: 57112
passed LTC №2 with [0, 3.5] bounds
http://tests.stockfishchess.org/tests/view/5d55cde40ebc5925cf10b79d
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 33913 W: 5800 L: 5529 D: 22584
bench 4087744
This is a continuation of a lot of people efforts (namely @jerrydonaldwatson) in improving child node futility pruning.
I counted at least 6 different patches that lowered maximum depth of it and made it more aggressive but failed to deliever at LTC. With recent @xoto10 patch (more like it proven non-linear scaling) I thought that this part of search also scales non - linearly.
So I decided to make array instead of a simple heuristic that spreads to higher depths and tune it at LTC. All tuning work done by @Alayan-stk-2 so big props to him because I don't really know how to use SF tuner :)
This patch had an STC run http://tests.stockfishchess.org/tests/view/5d53e9a40ebc5925cf109750 which was decent, but since it's an LTC tuning AND this part of code has non-linear scaling (according to my hypothesis) it should be tested at LTC and it passed 2 SPRTs there, both [0.5, 4.5] and [0, 3.5].
What can be done from this?
This all probably will require a lot of LTC tuning with not that clear results but this tests do bring confidence that there is something more and they produce something that seem to scale well.