LMR - [for starters] - [Advance] and [Expert]

[Michael Sherwin]

For those engines that use history tables for LMR you might want to try something like this.

Code: Select all

              s32 m = histTblg[fig][fs][ts];
              s32 n = histTbln[fig][fs][ts];
	      if(n > 49 && m/n < 16) {
                reduce += (int)(1.0 + sqrt(sqrt(depth) * sqrt(count)) / (log(m/n+1)+1));
              }

So if depth = 4 and count = 16 and hist% is the max 15 then it would look like

R = 1 + sqrt(2 x 4) / 2.2 = 1

however if hist% = 0 it would look like this

r = 1 + sqrt(2 x 4) / 1 = 3

An extreme case might look like

r = 1 + sqrt(6 x 10) / 1 = 8;

[Michael Sherwin]

Even though everyone seems to be out to lunch on this topic I will report that it is giving me new best results. However I had to add the line.

reduce = min(reduce, 5);

[zenpawn]

What are the two history tables in your code snippet? (I only have one.)

Thanks,
-Erin

[Henk]

Yes please don't use unidentified symbols in your code snippets. Stupid nerds.

[Michael Sherwin]

What are the two history tables in your code snippet? (I only have one.)

Thanks,
-Erin

s32 m = histTblg[fig][fs][ts]; +100 when move qualifies
s32 n = histTbln[fig][fs][ts]; +1 every time

so m/n is a %

[zenpawn]

Ah, the so-called butterfly table. I have one of those. Thanks.