Minor threats b

[ajithcj]

Add bonuses for Minors attacking enemy pieces(except pawns) even when they are protected by enemy pawns.

Patch has passed both STC
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 8206 W: 1426 L: 1304 D: 5476

and LTC
LLR: 2.97 (-2.94,2.94) [0.00,6.00]
Total: 19534 W: 2821 L: 2640 D: 14073

Note: LTC was run at SPRT [0,6] as it is not a complicated change.

[ajithcj]

There were several previous tests that tried to pick the highest valued piece(for WeakEnemies). But all those patches failed.
I picked lsb as weakEnemies was also computed in the same way and somehow using lsb was the most successful approach so far.

[ajithcj]

Morover the value of two threats is not always the sum of two threats. For example I would think that when there are two threats the resultant bonus should never exceed ThreatenedByPawn[] bonus.
That way a piece doesn't get threatened by a pawn to create threats on two minor pieces.

[ajithcj]

Lucas, refer to below tests by Joona which tried out both ideas(both failed fairly quickly):
https://groups.google.com/forum/?fromgroups=#!searchin/fishcooking_results/threat$20/fishcooking_results/OqKtbAXOL_w/CTaSTSH2fDUJ

https://groups.google.com/forum/?fromgroups=#!searchin/fishcooking_results/threat$20/fishcooking_results/18M9bmOaZeU/cG04qbbBBk4J

[zamar]

I need to think about this patch. It adds some complexity, so it might be a good idea to measure its value in ELO before committing.

[glinscott]

Interesting idea! I think this would be clearer without adding ei.attackedByMinors. Most of the code paths where it's used are behind if conditions, so it might not be a speed up as well. Can you push an update without that? Then it should be quite a small diff. Thanks!

[ajithcj]

Gary, I had tested it without adding ei.attackedByMinors
The test was minor_threats and it failed yellow at STC:
http://tests.stockfishchess.org/tests/view/53ae767b0ebc5926dec5ce10

ei.attackedByMinors is an important part of this patch as it avoids replicating Minor attacks by Bishop and Knight at 5 different places in the source code.

[glinscott]

That could have just been bad luck though. Have you measured ei.attackedByMinors to be a speed-up?

[ajithcj]

No I haven't. I have to read Marco's instructions on testing speedup. Will test it later when I have time. Have to leave for work.

[lucasart]

@ajithcj: ok. I will try to remove these lsb once this patch is commited. Problem might be that it requires retuning...

[lucasart]

@zamar: the added complexity is small. once you remove the attackedByMinors stuff, which is non functional code refactoring, the diff is clean and there is really one idea in this patch (not a mix). it already passed sprt(0,6) at ltc after an sprt(-1.5,4.5) preselection at stc. i don't think it's a good idea to add some ad hoc third testing step after that.

[ajithcj]

Joona/Gary, feel free to launch an elo gain test if you feel the need.

[zamar]

OK, attackedByMinors stuff lead me astray. If the added complexity is small, it's fine for me to merge this. But whoever merges this, please remove the attackedByMinors stuff while merging, I don't think that it provides any real value... If it is a speed up it can be tested separately. Thanks.

[ajithcj]

Joona, as I said in my comment to Gary above, attackedByMinors is a simplification which works very well with this patch. Without it You are effectively committing an inefficient patch as shown by below test(without attackedByMinors);
http://tests.stockfishchess.org/tests/view/53ae767b0ebc5926dec5ce10
I understand the need to separate performance gains from functional changes. But these are not two separate ideas and making a patch inefficient to separate out the performance changes doesn't seem to make sense.
I'm not sure the minor_threats_B change or attackedByMinors changes will separately pass SPRT(0,6) simply because they complement each other well.

[ajithcj]

Collating all four minor threats for reference:
http://tests.stockfishchess.org/tests/view/53ae767b0ebc5926dec5ce10
http://tests.stockfishchess.org/tests/view/53ae79250ebc5926dec5ce12
http://tests.stockfishchess.org/tests/view/53b00c450ebc5943266b4bf7
http://tests.stockfishchess.org/tests/view/53b02dcc0ebc5943266b4bfd

[glinscott]

Merged in 6b35430.

@ajithcj, I tested with and without the attackedByMinors change. It was slightly faster without the array, so I haven't committed it.

With attackedByMinors
Total time (ms) : 4585
Nodes searched : 9942172
Nodes/second : 2168412

Without attackedByMinors
Total time (ms) : 4561
Nodes searched : 9942172
Nodes/second : 2179822

[ajithcj]

OK. Thanks.