A small number of tests with simulated
annealing at 15s indicated these values
may be better
And this is verified at long 60+0.05 TC
LLR: 2.95 (-2.94,2.94)
Total: 40658 W: 7821 L: 7501 D: 25336
bench: 4931544
This patch is the sum of:
- Grainsize of 4 instead of 8
- Removing "depth < DEPTH_ZERO"
- Change DEPTH_QS_RECAPTURES = -5 to -7
All the patches individually failed to pass SPRT but scored
around 50%.
Together they pass easily short TC:
LLR: 2.96 (-2.94,2.94)
Total: 4429 W: 964 L: 844 D: 2621
And with some difficult long TC of 60+0.05:
LLR: 2.95 (-2.94,2.94)
Total: 64133 W: 11968 L: 11532 D: 40633
bench: 4821467
Performed more or less well at short TC
LLR: 2.95 (-2.94,2.94)
Total: 50517 W: 9815 L: 9574 D: 31128
And a bit better at long TC
LLR: 2.96 (-2.94,2.94)
Total: 15564 W: 2805 L: 2624 D: 10135
bench: 4375253
It seems that do not limiting checking the
trapped rook only on rank 1 improves the
score.
At long TC
LLR: 2.97 (-2.94,2.94)
Total: 6581 W: 1346 L: 1204 D: 4031
bench: 4985012
A rook is trapped if on rank 1 as is the king.
Currently the condition aloows for the rook
to be also in front of the pawns as long
as king is on first rank.
Verified with short TC test:
LLR: -1.71 (-2.94,2.94)
Total: 23234 W: 4317 L: 4317 D: 14600
Here what it counts is that after 23K games
result is equal.
bench: 4696542
Allow to use EvalInfo struct, populated by
evaluation(), in search.
In particular we allocate Eval::Info on the stack
and pass a pointer to this to evaluate().
Also add to Search::Stack a pointer to Eval::Info,
this allows to reference eval info of previous/next
nodes.
WARNING: Eval::Info is NOT initialized and is populated
by evaluate(), only if the latter is called, and this
does not happen in all the code paths, so care should be
taken when accessing this struct.
No functional change.
Rationale:
- Current settings seem to make engine *significantly* weaker in analysis mode.
- In practice this setting only has effect when king safety scores are high.
- Even in analysis mode its far more important to know if one side is getting mated,
rather than get evaluation correct with 1cp accuracy.
No functional change
Give a bonus if a bishop can pin a piece or can
give a discovered check through an x-ray attack.
Seems good after 24000 games at 15"+0.05 (single thread):
ELO: 12.30 +- 99%: 5.79 95%: 4.40 LOS: 100.00%
Total: 24000 W: 4931 L: 4082 D: 14987
bench: 4917064
They are not self-describing and create a lot of user
requests about them.
Given that the values are already well tuned there
is no need to expose them as UCI options.
No functional change.
Since &-ing with SpaceMask restricts the set to the home
half of the board, it is possible to use just one popcount
instead of 2 by shifting "safe" to the other half of the
board. This gives a small speedup especially on systems
where hardware popcount is not available.
Patch kindly sent by Richard Vida.
No functional change.
And return on using TT as backing store for position
evaluations.
Tests (even on single thread) show eval cache was a regression.
In multi thread result should be even worst because eval cache
is a per-thread struct, while TT is shared.
After 4957 games at 15"+0.05 (single thread)
eval cache vs master 969 - 1093 - 2895 -9 ELO
So previous reported result of +18 ELO was probably due to an
issue in the testing framework (a bug in cutechess-cli) that
has been fixed in the meanwhile.
bench: 5386711
