Komodo has i little bit gain i ELO with Selective=150

[consen]

After 10580 games the result is this.
Komodo is not in standard setting. I have Selective 150 on it.
All other option is standard.
I have not use opening book.
Have set 4 CPUs on both engine on a i7 3770 cpu.
Tested in Fritz 15 GUI with windows 10.

1m+1s is playing time.

1 Stockfish 7 64 POPCNT +22 +2348/=6539/-1693 53.10% 5617.5/10580

2 Komodo 9.3 64-bit -22 +1693/=6539/-2348 46.90% 4962.5/10580

Can mention that is nearly +21 to Sf. If 53,08%, it is +21 ELO.


I took the same test with standard option on Komodo 9.3.
Selective is then on 131.

This is the result.

1 Stockfish 7 64 POPCNT +24 +1262/=3514/-874 53.43% 3019.0/5650

2 Komodo 9.3 64-bit -24 +874/=3514/-1262 46.57% 2631.0/5650

I am sorry for to little games in the last result.
But windows 10 go in automatic restart after a update.
When i testing now, I take out the internett-plug so this not happend more times.

Conklusion:
Maybe Komodo 9.3 is a little bit stronger with Selective 150 than standard 131.
---------

I notes the % with Selective 150 for Stockfish.

6024 games: 53,11%
6144 games: 53,08%
6388 games: 53,18%
6690 games: 53,14%
6978 games: 53,08%
7000 games: 53,06%
7220 games: 53,14%
7340 games: 53,16%
7542 games: 53,12%
8034 games: 53,19%
8278 games: 53,15%
8386 games: 53,16%
8730 games: 53,17%
9002 games: 53,13%
9336 games: 530,9%
9754 games: 53,12%
9896 games: 53,08%
10066 games: 53,10%
10560 games: 53,09%
10580 games: 53,10%

As we can see, Sf is between +21 and +22 Elo stronger than Stockfish with Selective 150. And Sf is +24 ELO stronger than Komodo with Selective 131 (standard option). We gain 2-3 ELO with Selective 150.

53,08% is +21 ELO for Stockfish.
Allover 53, 08 is +22 ELO for Stockfish.

[consen]

Komodo 9.3 - Sf7

Time control 1m+1s
Without sysygy
Without opening book

--------------------------------------------------------------------------------
With selective=131

sf7 +1262/=3514/-874 53.43% 3019/5650 +24
K9.3 +874/=3514/-1262 46.57% 2631/5650 - 24

--------------------------------------------------------------------------------

With selective=150

Sf7 +2348/=6539/-1693 53.10% 5617.5/10580 +22
K9.3 +1693/=6539/-2348 46.90% 4962.5/10580 - 22

--------------------------------------------------------------------------------

With selective=170

Sf7 +924/=2537/-661 53.19% 2192.5/4122 +22
K9.3 +661/=2537/-924 46.81% 1929.5/4122 - 22

--------------------------------------------------------------------------------

It surprises me that there is so little difference between the various selectivtiy.

There are some few games in the last test (selectivity=150), but I see that it goes against the same as the other settings.

Since Komodo 9:42 came, I'd rather test it. Therefore I interrupted last test.

It may look like Komodo 9.3 is best with selectivity = 150.

Too few played games to prove anything.

[yurikvelo]

Default
80 for Komodo 9.01 and 9.02,
68 for Komodo 9.1
74 for Komodo 9.2
130 for Komodo 9.3
131 for Komodo 9.4

[consen]

Yuri wrote:
Default
80 for Komodo 9.01 and 9.02,
68 for Komodo 9.1
74 for Komodo 9.2
130 for Komodo 9.3
131 for Komodo 9.4
-------------------

When I push on Defaults on my Komodo 9.3, the Selective is 131, not 130 as you wrote. But maybe there are different versions of it.

[consen]

I have done same Selective test on Komodo 9.42 as 9.3.

I have tested it with Selective 131 (standard) and Selective 150.
All others option is standard.
Without opening book and without sysygy.
Time: 1m+1s

This is the result:


-------------------------------------------------------------------------
With Selective 131 (standard)

Stockfish 7: +552/=1268/-364 54,30% 1186,0/2184 +30
Komodo 9.42: +364/=1268/-552 45,70% 998,0/2184 -30
-------------------------------------------------------------------------

Stockfish 7: +549/=1323/-366 54,09% 1210,5/2238 +28
Komodo 9.42: +366/=1323/-549 45,91% 1027,5/2238 -28
-------------------------------------------------------------------------

Komodo still gain +2 ELO against Stockfish 7 with selective 150.

This is not a prove. But I can see the pattern.
In short game, selective 150 gain a little bit against stockfish 7.

In long game I don't know anything. It can be opposite.

And! Too few game to conclude of anytning.

I can altso see a pattern that Komodo 9.42 is weaker than 9.3 in game with 1m+1s. 6 ELO actual.

Of course it can be opposite with 2m+1s.

[consen]

A new test is done.

Stockfish 7 - Komodo 9.42 with Selective = 100. All others is standard.

1m+1s

1 Stockfish 7 64 POPCNT +23 +2231/=5325/-1624 53.31% 4893.5/9180
2 Komodo 9.42 64-bit -23 +1624/=5325/-2231 46.69% 4286.5/9180

This is the best result so far for Komodo.

The test is done without opening book and without sysygy.
4 cpu is used on Fritz 15 gui, (8 cpu is possible), on a i7 3770 cpu.
Standard Hash is used in Fritz 15 gui for engine match.