The current skill levels (1-20) allow for adjusting playing strengths, but
do so in big steps (e.g. level 10 vs level 11 is a ~143 Elo jump at STC).
Since the 'Skill Level' input can already be a floating point number, this
patch uses the fractional part of the input to provide the user with
fine control, allowing for varying the playing strength essentially
continuously.
The implementation internally still uses integer skill levels (needed since they pick Depths),
but non-deterministically rounds up or down the used skill level such that the average integer
skill corresponds to the input floating point one. As expected, intermediate
(fractional) skill levels yield intermediate playing strenghts.
Tested at STC, playing level 10 against levels between 10 and 11 for 10000 games
level 10.25 ELO: 24.26 +-6.2
level 10.5 ELO: 67.51 +-6.3
level 10.75 ELO: 98.52 +-6.4
level 11 ELO: 143.65 +-6.7
http://tests.stockfishchess.org/tests/view/5cd9c6b40ebc5925cf056791http://tests.stockfishchess.org/tests/view/5cd9d22b0ebc5925cf056989http://tests.stockfishchess.org/tests/view/5cd9cf610ebc5925cf056906http://tests.stockfishchess.org/tests/view/5cd9d2490ebc5925cf05698e
No functional change.
Use comparison of eval with beta to predict potential cutNodes. This
allows multi-cut pruning to also prune possibly mislabeled Pv and NonPv
nodes.
STC:
LLR: 2.95 (-2.94,2.94) [0.50,4.50]
Total: 54305 W: 12302 L: 11867 D: 30136
http://tests.stockfishchess.org/tests/view/5d048ba50ebc5925cf0a15e8
LTC:
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 189512 W: 32620 L: 31904 D: 124988
http://tests.stockfishchess.org/tests/view/5d04bf740ebc5925cf0a17f0
Normally I would think such changes are risky, specially for PvNodes,
but after trying a few other versions, it seems this version is more
sound than I initially thought.
Aside from this, a small funtional change is made to return
singularBeta instead of beta to be more consistent with the fail-soft
logic used in other parts of search.
=============================
How to continue from there ?
We could try to audit other parts of the search where the "cutNode"
variable is used, and try to use dynamic info based on heuristic
eval rather than on this variable, to check if the idea behind this
patch could also be applied successfuly.
Bench: 3503788
Fixes issues #2126 and #2189 where no progress in rootDepth is made for particular fens:
8/8/3P3k/8/1p6/8/1P6/1K3n2 b - - 0 1
8/1r1rp1k1/1b1pPp2/2pP1Pp1/1pP3Pp/pP5P/P5K1/8 w - - 79 46
the cause are the shuffle extensions. Upon closer analysis, it appears that in these cases a shuffle extension is made for every node searched, and progess can not be made. This patch implements a fix, namely to limit the number of extensions relative to the number of nodes searched. The ratio employed is 1/4, which fixes the issues seen so far, but it is a heuristic, and I expect that certain positions might require an even smaller fraction.
The patch was tested as a bug fix and passed:
STC:
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 56601 W: 12633 L: 12581 D: 31387
http://tests.stockfishchess.org/tests/view/5d02b37a0ebc5925cf09f6da
LTC:
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 52042 W: 8907 L: 8837 D: 34298
http://tests.stockfishchess.org/tests/view/5d0319420ebc5925cf09fe57
Furthermore, to confirm that the shuffle extension in this form indeed still brings Elo, one more test at VLTC was performed:
VLTC:
LLR: 2.96 (-2.94,2.94) [0.00,3.50]
Total: 142022 W: 20963 L: 20435 D: 100624
http://tests.stockfishchess.org/tests/view/5d03630d0ebc5925cf0a011a
Bench: 3961247
Since root_probe() and root_probe_wdl() do not reset all tbRank values if they fail,
it is necessary to do this in rank_root_move(). This fixes issue #2196.
Alternatively, the loop could be moved into both root_probe() and root_probe_wdl().
No functional change
This is a non-functional simplification that combines vote counting and thread selecting in the same loop.
It is possible that the best thread would be updated more frequently than master, but I'm not sure it matters here. Perhaps "mostVotes" is a better name than "bestVote?"
STC (stopped early).
LLR: 0.70 (-2.94,2.94) [-3.00,1.00]
Total: 10714 W: 2329 L: 2311 D: 6074
http://tests.stockfishchess.org/tests/view/5ccc71470ebc5925cf03d244
No functional change.
Treat all threads the same as main thread and increment
failedHighCnt on fail highs. This makes the search try
again at lower depth.
@vondele suggested also changing the reset of failedHighCnt
when there is a fail low. Tests including this passed so the
branch has been updated to include both changes. failedHighCnt
is now handled exactly the same in helper threads and the main
thread. Thanks vondele :-)
STC @ 5+0.05 th 4 :
LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 7769 W: 1704 L: 1557 D: 4508
http://tests.stockfishchess.org/tests/view/5c9f19520ebc5925cfffd2a1
LTC @ 20+0.2 th 8 :
LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 37888 W: 5983 L: 5889 D: 26016
http://tests.stockfishchess.org/tests/view/5c9f57d10ebc5925cfffd696
Bench 3824325
This functional simplification removes the PvNode reduction and adjusts
the ttPv lmr condition accordingly. Their definitions only differ by the
inclusions of ttPv. Aside from this, shallow move pruning definition
will be the only other functional difference, but this does not seem to
matter too much.
STC:
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 58908 W: 12980 L: 12932 D: 32996
http://tests.stockfishchess.org/tests/view/5cd1aaaa0ebc5925cf046c6a
LTC:
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 20351 W: 3521 L: 3399 D: 13431
http://tests.stockfishchess.org/tests/view/5cd23fa70ebc5925cf047cd2
Bench: 3687854
only the extension part of the shuffle patch is sufficient to
pass [0,3.5] bounds at VLTC as shown by two more tests.
http://tests.stockfishchess.org/tests/view/5cc168bc0ebc5925cf02bda8
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 120684 W: 17875 L: 17400 D: 85409
http://tests.stockfishchess.org/tests/view/5cc14d510ebc5925cf02bcb5
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 68415 W: 10250 L: 9905 D: 48260
this patch proposes to simplify back to this basic and easier to
understand version. In case there is a need to run a [-3, 1] VLTC on
this one, it can be done, but it is resource intensive, and not needed
IMO.
Bench: 3388643
It causes a serious regression hanging a simple fixed
depth search. Reproducible with:
position fen q1B5/1P1q4/8/8/8/6R1/8/1K1k4 w - - 0 1
go depth 13
The reason is a search tree explosion due to:
if (... && depth < 3 * ONE_PLY)
extension = ONE_PLY;
This is very dangerous code by itself because triggers **at the leafs**
and in the above position keeps extending endlessly. In normal games
time deadline makes the search to stop sooner or later, but in fixed
seacrch we just hang possibly for a very long time. This is not acceptable
because 'go depth 13' shall not be a surprise for any position.
This patch reverts commit 76f1807baa.
and fixes the issue https://github.com/official-stockfish/Stockfish/issues/2091
Bench: 3243738
Introduce a new search extension when pushing an advanced passed pawn is
also suggested by the first killer move. There have been previous tests
which have similar ideas, mostly about pawn pushes, but it seems to be
overkill to extend too many moves. My idea is to limit the extension to
when a move happens to be noteworthy in some other way as well, such as
in this case, when it is also a killer move.
STC:
LLR: 2.96 (-2.94,2.94) [0.50,4.50]
Total: 19027 W: 4326 L: 4067 D: 10634
http://tests.stockfishchess.org/tests/view/5cac2cde0ebc5925cf00c36d
LTC:
LLR: 2.94 (-2.94,2.94) [0.00,3.50]
Total: 93390 W: 15995 L: 15555 D: 61840
http://tests.stockfishchess.org/tests/view/5cac42270ebc5925cf00c4b9
For future tests, it looks like this will interact heavily with passed
pawn evaluation. It may be good to try more variants of some of the more
promising evaluations tests/tweaks.
Bench: 3666092
The current update only by main thread depends on the luck of
whether main thread sees any/many changes to the best move or not.
It then makes large, lumpy changes to the time to be
used (1x, 2x, 3x, etc) depending on that sample of 1.
Use the average across all threads to get a more reliable
number with a smoother distribution.
STC @ 5+0.05 th 4 :
LLR: 2.95 (-2.94,2.94) [0.50,4.50]
Total: 51899 W: 11446 L: 11029 D: 29424
http://tests.stockfishchess.org/tests/view/5ca32ff20ebc5925cf0016fb
STC @ 5+0.05 th 8 :
LLR: 2.96 (-2.94,2.94) [0.50,4.50]
Total: 13851 W: 2843 L: 2620 D: 8388
http://tests.stockfishchess.org/tests/view/5ca35ae00ebc5925cf001adb
LTC @ 20+0.2 th 8 :
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 48527 W: 7941 L: 7635 D: 32951
http://tests.stockfishchess.org/tests/view/5ca37cb70ebc5925cf001cec
Further work:
Similar changes might be possible for the fallingEval and timeReduction calculations (and elsewhere?), using either the min, average or max values across all threads.
Bench 3506898
Shuffle detection procedure :
Shuffling positions are detected if
the last 36 moves are reversible (rule50_count() > 36),
the position have been already in the TT,
there is a still a pawn on the board (to avoid special endings like KBN vs K).
The position is then judged as a draw.
An extension is realized if we already made 14 successive reversible moves in PV to accelerate the detection of the eventual draw.
To go further : we can still improve the idea. The length of the tests need a lot of ressources.
the limit of 36 is logic but must be checked again for special zugzwang positions,
this limit can be decreased in special positions,
the limit of 14 moves for extension has not been tuned.
STC
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 32595 W: 7273 L: 7275 D: 18047 Elo +0.43
http://tests.stockfishchess.org/tests/view/5c90aa330ebc5925cfff1768
LTC
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 51249 W: 8807 L: 8486 D: 33956 Elo +1.85
http://tests.stockfishchess.org/tests/view/5c90b2450ebc5925cfff1800
VLTC
LLR: 2.96 (-2.94,2.94) [0.00,3.50]
Total: 137974 W: 20503 L: 19983 D: 97488 Elo +1.05
http://tests.stockfishchess.org/tests/view/5c9243a90ebc5925cfff2a93
Bench: 3548313
Adding a clamp function makes some of these range limitations a bit prettier and removes some #include's.
STC
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 28117 W: 6300 L: 6191 D: 15626
http://tests.stockfishchess.org/tests/view/5c9aa1df0ebc5925cfff8fcc
Non functional change.
always use the implementation of gives_check in position, no need to
hand-inline part of the implementation in search.
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 57895 W: 12632 L: 12582 D: 32681
http://tests.stockfishchess.org/tests/view/5c9eaa4b0ebc5925cfffc9e3
No functional change.
While looking at pruning using see_ge() (which is very valuable)
it became apparent that the !extension test is not adding any
value - simplify it away.
STC:
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 56843 W: 12621 L: 12569 D: 31653
http://tests.stockfishchess.org/tests/view/5c8588cb0ebc5925cffe77f4
LTC:
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 78622 W: 13223 L: 13195 D: 52204
http://tests.stockfishchess.org/tests/view/5c8611cc0ebc5925cffe7f86
Further work could be to optimize the remaining see_ge() test. The idea of less pruning at higher depths is valuable, but perhaps the test (-PawnValueEg * depth) can be improved.
Bench: 3188688
This is a functional simplification that removes the FutilityMoveCounts array with a simple equation using only ints.
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14175 W: 3123 L: 2987 D: 8065
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 9900 W: 1735 L: 1597 D: 6568
Bench: 3380343
This is a non-functional patch which shrinks the reductions array.
This saves about 8Kb of memory.
The only slow part of master's reductions array is the calculation
of the log values, so using a separate array for those values and
calculating the rest real-time appears to be just as fast as master.
STC
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 63245 W: 13906 L: 13866 D: 35473
http://tests.stockfishchess.org/tests/view/5c7b571f0ebc5925cffdc104
No funcional change.
xoto10