Cutting Edge - long time control tournament - please vote!

[Martin Thoresen]

Hello everyone,

I've recently purchased quite a number of new engines. In the following days I'll start a tournament consisting of the following engines:

Rybka 3 64-bit 4CPU
Deep Junior 2010 64-bit 4CPU
Stockfish 1.6.3 JA 64-bit 4CPU
Deep Sjeng WC2008 64-bit 4CPU
Thinker 5.4C Inert 64-bit 4CPU
Deep Shredder 12 64-bit 4CPU
Crafty 23.1 64-bit 4CPU
Zappa Mexico II 64-bit 4CPU
Naum 4.1 64-bit 4CPU
Spark 0.3A 64-bit 4CPU

Most notably missing is Deep Fritz 11 and Hiarcs 12.1, but I have decided not to buy these and instead by DF12 and Hiarcs 13 when they are released.

Except for those two, the current top-of-the-line engines should all be there.

Going to use Harry Schnapp's most recent Arena mainbook limited to 8 moves. 1GB Hash. Ponder off. All engines run with 4 threads.

Question is, what time control would you prefer?

Please vote in the poll.

[Dr.Wael Deeb]

Hi Martin,
Running these engines on a quad,I use 20 minutes + 20 seconds increment....I personalty think that it's a good tme control considering the speed of the nowadays hardware and you get a decent number of games per day....
Dr.D

[Rubinus]

90m+30s
Regards

[Martin Thoresen]

Hi Martin,
Running these engines on a quad,I use 20 minutes + 20 seconds increment....I personalty think that it's a good tme control considering the speed of the nowadays hardware and you get a decent number of games per day....
Dr.D

Hello Wael!

Well, it's not so much about getting a decent number of games per day. I want to have some high quality games in this tournament.

I can see that people have voted for several different options.

I personally favor the "rest in xx" because then the games can't go on forever in the end game.

[Albert Silver]

I choose other. Actually, there are two other factors to consider: ponder on or off (I saw you chose off), and the CPU speed.

Personally, I view ponder ON as the true test. It means the engines are not waiting and doing nothing while the other thinks, and can not only analyze, but show how well they use the opponent's time as well. As a player, the idea of not being able to analyze the position while my opponent is thinking is incomprehensible. This also means the engine is in effect getting double the thinking time, so higher quality chess throughout.

The issue of quantity is not negligible so while slowest seems best, it isn't practical to have only one game a day. In my personal slow TC testing, on a core2duo @3.6GHz, I had 'ponder on' with 40/40 + 40/40 +40KO. As a result, some positions were analyzed for as long as 10 mins, and the opponent would also have analyzed for that same length of time. Bear in mind, I was using the SilverSuite, which already comes with 5-15 opening moves, so the first TC was actually 40/35 to 40/25, which is fairly reasonable to determine the game.

[yanquis1972]

my perspective about ponder is that it's entirely a function of computer-computer games (computer-human if you're really, really, really good) & has zero to do with chess analysis, which imo is the primary function of an elite 21st century chess engine. i much prefer ponder off in terms of assessing engines. however, if the goal is a fun human-style tournament, ponder would be preferred i agree.

[Tomcass]

My view is:

It depends of the speed of you machine. I have two quads, one running three times faster than the other. So that the recommendation must be different for each of them.

I dare to suggest you to take Stockfish as a reference and calculate an average of no less than 30 billion nodes per game for this engine (rounding up or down to the next multiple of 2 minutes). It must be enough to get a set of good quality games. But this is obviously a totally subjective vision from my side.

Good luck for your Tournament and regards from Barcelona.

Tom.

[Albert Silver]

my perspective about ponder is that it's entirely a function of computer-computer games (computer-human if you're really, really, really good) & has zero to do with chess analysis, which imo is the primary function of an elite 21st century chess engine. i much prefer ponder off in terms of assessing engines. however, if the goal is a fun human-style tournament, ponder would be preferred i agree.

He is playing a tournament, not testing positions, therefore the purpose is to see which engine plays best.

[Martin Thoresen]

Thanks everyone for your input.

The decision has been made, and it's going to be option #4.

40 / 120 + 20 / 60 + 30 for the rest

Best Regards,
Martin

[ernest]

Since this is LTC, too bad you don't include FB...