Komodo 9.42

[mjlef]

Will the Komodo team be reiterating, the "4 updates" commitment with K10 subscribers? It looks like the only version, to clearly show an ELO increase across the rating lists, while remaining bug free, was 9.2

I saw increases in all versions (9.1-9.3) for CEGT, except for the bug in 9.3 that seems to only affect it when on 4 threads.

9.1 and 9.3 were skipped in some lists (IPON). 9.3 get below 9.2 on CCRL 40/40 and 9.1 gets below 9 on CCRL 40/4. That's why I said "across the rating lists", results are inconsistent except for version 9.2

Just a reminder that the error bars are pretty big in most rating lists. In Komodo 9.3 we did some changes that helped MP with many processors quite a lot, but it was not so great for 4 processors which is what is generally at the top of the lists. That was improved in Komodo 9.42. Both Komodo 9.2 and 9.3 came in ahead of Stockfish 7 on say the CCRL 40/40 list single thread results, but behind on the 4 thread runs. But with error margins + and - 19 elo, it is hard to really say who is better.

[Ozymandias]

Just a reminder that the error bars are pretty big in most rating lists. […] with error margins + and - 19 elo, it is hard to really say who is better.

That goes without saying. But when it's clearly better, like 9.2, it will shine through. Maybe 9.4 will also be noticeably stronger, but in this case, the problem has been bugs. All in all, it seems like 4 updates in less than a year, might be too much of a commitment.

[Nordlandia]

What happends in case "King Safety" is adjusted to maximum possible?

Will the King stay in the corner for the entire game or what

[lkaufman]

Hmm, I currently own 9.3, can I get a good deal on 9.42, then?

You can get 9.42 for 20% off; if you want to keep current it's cheaper to buy subscription.

[Gurcan Uckardes]

Larry, Mark,
Any plan for Android? The tablets didn't like 9.3 after 9.2 either...

[lkaufman]

What happends in case "King Safety" is adjusted to maximum possible?

Will the King stay in the corner for the entire game or what

"King Safety" is a bit of a misnomer; a more accurate but awkward name would be something like "king attack and defense". A maximum value will make Komodo strive very hard to attack the enemy king while keeping its own king very safe, at the expense of everything else (material, pawn structure, mobility, etc.).

[Frank Quisinsky]

Hi Larry,

I am not sure with the Komodo 9.42 Settings I should used for my test.

Remember also that I start a test vs. 55 engines x 50 games = 2.750 games with longer time controls and most of the opponents are clearly weaker. Correction ... all others are weaker ... Komodo 9.3 is on place 1 in my list. Not Stockfish because Stockfish have around 3% more draw quote with standard setting!

Conditions:
http://www.amateurschach.de/main/_conditions.htm

It would be a great help if you can give me a tipp ... means settings your prefer.

Should I test Komodo 9.42 x64 with standard settings or should I switched different of the newer settings?

Maybe you find the time for an answere and I can go to bed without headache. Looking in the messages here and found a lot .. but again ... I am not sure!

Best
Frank

[lkaufman]

Hi Larry,

I am not sure with the Komodo 9.42 Settings I should used for my test.

Remember also that I start a test vs. 55 engines x 50 games = 2.750 games with longer time controls and most of the opponents are clearly weaker. Correction ... all others are weaker ... Komodo 9.3 is on place 1 in my list. Not Stockfish because Stockfish have around 3% more draw quote with standard setting!

Conditions:
http://www.amateurschach.de/main/_conditions.htm

It would be a great help if you can give me a tipp ... means settings your prefer.

Should I test Komodo 9.42 x64 with standard settings or should I switched different of the newer settings?

Maybe you find the time for an answere and I can go to bed without headache. Looking in the messages here and found a lot .. but again ... I am not sure!

Best
Frank

In general, use default settings for testing (other than Hash table size which depends on the details of the test). Table Memory could be raised to a quarter of Hash size (probably shouldn't exceed 1 GB). It might be that Dynamism could be a bit higher than the default for non-blitz games, but I don't really know that yet so I'm not proposing it. Regarding Contempt, if you are only playing against Stockfish a value of zero (or very near zero) is probably best, but for a range of opponents like you normally use I think the default (10) is fine.

[Frank Quisinsky]

Thanks Larry!!

Best and enough chess for today ... 02:05 German time.
Frank

[mjlef]

Larry, Mark,
Any plan for Android? The tablets didn't like 9.3 after 9.2 either...

I do not know of any issues with running any versions of Komodo on Android tablets. A Komodo 9.4 is undergoing testing and will a a free upgrade for Komodo 9 Android owners once it is tested. It will basically be Komodo 9.42.