WCCC: schedule, pairings, live broadcast

[mclane]

in the open tournament you measure the hardware strength !

[Gian-Carlo Pascutto]

In general engines get stronger if you give them more time. And they get stronger if you give them more CPUs.

Not all engines benifit equally on in terms of Elo when doublingthe search time or doubling the number of CPUs. How well they can profit depends on the quality of the engine.

Giving an engine more CPUs has much more potential to make the engine weaker if the programmer is not careful.

Conversely, there aren't many engines that get weaker if they get more time.

[Gian-Carlo Pascutto]

in the open tournament you measure the hardware strength !

Particularly the programmers ability to make use of it.

The "handicap" tournament also measures hardware strength. Some engines are on quads, others are on (much faster) dual Nehalems. You're measuring hardware strength within a completely arbitrary limitation.

The open tournament opens up more possibilities for true innovation.

[Gian-Carlo Pascutto]

In other words, the Open Tournament is one in which the final standings least reflect the strengths of the engines.

What is strength? If a great engine on single CPU can't even use a dual, does it deserve to be called strong?

[hgm]

Giving an engine more CPUs has much more potential to make the engine weaker if the programmer is not careful.

Conversely, there aren't many engines that get weaker if they get more time.

Buggy engines can easily get weaker when given more time. E.g. if they have a fixed probability for a fatal crash per unit of time. Once the engine gets beyond a certain quality.

For 1 vs 2 CPUs you might have a point. But by the time an engine can ue 8 CPUs, it is beyond the quality level where it would benefit from more. So there is little point in giving it more, just like there is little point in increasing the TC from 2 hours to 2 days.

[Gian-Carlo Pascutto]

. But by the time an engine can ue 8 CPUs, it is beyond the quality level where it would benefit from more.

I have absolutely no idea what that is supposed to mean.

[pedrox]

Asked for someone who is in Pamplona:

Can people enter in the room for see the games?

I will be in Pamplona on Thursday to participate in the Rapid Tournament

Best,

Pedro

[hgm]

I have absolutely no idea what that is supposed to mean.

It is supposed to mean that to beneficially use 8 CPUs, you will need an advanced algorithm, and simple tricks like shared hash with unsynchronized searches won't cut it anymore.

[bob]

Giving an engine more CPUs has much more potential to make the engine weaker if the programmer is not careful.

Conversely, there aren't many engines that get weaker if they get more time.

Buggy engines can easily get weaker when given more time. E.g. if they have a fixed probability for a fatal crash per unit of time. Once the engine gets beyond a certain quality.

For 1 vs 2 CPUs you might have a point. But by the time an engine can ue 8 CPUs, it is beyond the quality level where it would benefit from more. So there is little point in giving it more, just like there is little point in increasing the TC from 2 hours to 2 days.

This is simply wrong. Using 8 is not twice as hard as using 4. It is worse. Using 16 is not twice as hard as using 8. It is worse. Using a cluster with two nodes is not twice as hard as using a single node. It is worse.

The problem is called "scalability" and it is an extremely difficult one to deal with. Past WCCC events have rewarded those that take the time to make this stuff work. Now we have a crippled/irrelevant tournament, and a small "open" tournament that correlates to what has been done in previous events...

[Gian-Carlo Pascutto]

I have absolutely no idea what that is supposed to mean.

It is supposed to mean that to beneficially use 8 CPUs, you will need an advanced algorithm, and simple tricks like shared hash with unsynchronized searches won't cut it anymore.

I think that's a great argument to have much more cores rather than less, because the effect is greater on 16 or on clusters.

I always thought computer chess was about advanced algorithms, rather than making the fastest Fruit clone. It's whatever you want to make of it, I guess, I know what my choice is...