Houdini 3 running for the IPON

[MM]

I think it's time for CEGT and CCRL to seriously consider switching to increment play. Repeating time control play is obsolete for both humans and computers.

Ahhh ... !!! Repeating time controls are played because analog clocks could not add increments. Nowadays it a a crusted tradition which is played because people in chess clubs play it because when they entered the club it was played like this ... it as a bad habit which is hard to wipe out!

Thx for your statement
Ingo

Nonsense, there is nothing crusted.
40/4 repeated is a timecontrol among others, no more, no less !

Best wishes,
G.S.

IMO repeating TC is the best way to play chess and i think it is the best way to test engines too. Everyone may have of course a different opinion. Grahams is right, here nobody can claim to be right.

Sometimes it happens that in games with increment between humans and between computers, one goes in zeitnot and it plays a lot of moves till the end with a few seconds on the clock, often ruining the 1st half of the game.

This happens frequently in blitz, for example.

It seems to me better to ''restart'' the clock after a time control.

Best Regards

[ThatsIt]

IMO repeating TC is the best way to play chess and i think it is the best way to test engines too. Everyone may have of course a different opinion. Grahams is right, here nobody can claim to be right.

Sometimes it happens that in games with increment between humans and between computers, one goes in zeitnot and it plays a lot of moves till the end with a few seconds on the clock, often ruining the 1st half of the game.

This happens frequently in blitz, for example.

It seems to me better to ''restart'' the clock after a time control.

Best Regards

Thats one of the reasons why we use this kind of timecontrol for our CEGT lists.

Best wishes,
G.S.

[Modern Times]

Thats one of the reasons why we use this kind of timecontrol for our CEGT lists.

Best wishes,
G.S.

Yes, same for CCRL. Everyone is free to choose what they want to do, but there are strong arguments like that in favour of a repeating time control.

[IWB]

But the same goes for a repeating time control.

Engine start to think longer after the time control and realizes tha the last moves played in a hurry were crab ... The argument is true - for all time controls.

BYe
Ingo

[Uri Blass]

But the same goes for a repeating time control.

Engine start to think longer after the time control and realizes tha the last moves played in a hurry were crab ... The argument is true - for all time controls.

BYe
Ingo

In most cases in repeating time control there is no serious time trouble for engines and basically engines can use nearly the same time for every move.

It is not the case in fischer time control.

Note that I do not say that repeating time control is better than fischer time control but only that I understand the reason that people explained.

I think that time is part of the game in both cases(fischer time control and repeating time control) and assuming the games take the same time
I see no reason to prefer having more time in the endgame in testing.

I would like to see some rating list that have both time control(for example both 40/40 and 60+20 because there is no reason that 2 opponents have to play the same time control in the same game and a good interface should support different time control for different engines)

It may be interesting to find fischer time control that is equivalent to 40/40 that means that the average rating for engines with 40/40 is the same as the average rating for the same engines with the fischer time control.

I am not sure if 60+20 is good to have the same rating as 40/40 or if you need faster time control or slower time control.

I suspect that fischer time control may be better in the meaning that engines can spend less time to get the same rating but I am not sure about it and only testing can find if it is better.

[IWB]

In most cases in repeating time control there is no serious time trouble for engines and basically engines can use nearly the same time for every move.

That is true, but theoretical (At least for the top engines). Just watch ONE game and you will see that the engines use more in the beginning and less towards the end of a time period.

I think that time is part of the game in both cases(fischer time control and repeating time control) and assuming the games take the same time
I see no reason to prefer having more time in the endgame in testing.

Actually I think the best time control would be 'sudden death'. With every other arbitrary choosen time control the tester is taking away the need of programming a propper time algo. And as time is part of the game the tester decides about that, not the program. The minimum decision a human can do is to tell the programm that game has to be finished in X minutes (for practical reasons). Everything else influences with the outcome.

I would like to see some rating list that have both time control(for example both 40/40 and 60+20 because there is no reason that 2 opponents have to play the same time control in the same game and a good interface should support different time control for different engines)

It may be interesting to find fischer time control that is equivalent to 40/40 that means that the average rating for engines with 40/40 is the same as the average rating for the same engines with the fischer time control.

I am not sure if 60+20 is good to have the same rating as 40/40 or if you need faster time control or slower time control.

I suspect that fischer time control may be better in the meaning that engines can spend less time to get the same rating but I am not sure about it and only testing can find if it is better.

To find out what increment represents best a repeating time control (or vice versa) would be a nice experiment, unfortunately I fear that it might be different for every engine ... the best one can find is a approximation ...

Bye
Ingo

EDIT: Thinking about it a time control like 1 move in X minutes is the logical conclusion out of the argumentation for a repeating time. Boring in drawish endgames but somehow logical.

[lkaufman]

Hi Larry,

OK, so CCRL blitz is quite a bit faster than CEGT blitz now, since they use 40/3 even if that is equivalent to 40/6 or more on your "base" hardware.

40/3' is a bottom line resp. a value we will not fall below even if we could, and we could, believe me... . On my AMD-Quads I could play at least 40/2':30'' due to our benchmark, Gerhards i5 are even faster. But Gerhard and me agreed not to fall below 40/3. We don't like time controls like 40/1'30" or less which would be played on an actual i7. And hardware is getting faster and faster, so in the near future we could play 40/0':30". Never ever!

But we would play more than 40/3 if a slower PC would be used, but IIRC there is no slower one available any more.

.....

I guess for CEGT the argument for switching to increment controls is stronger than for CCRL since CEGT already made one major change here.

I thought about creating a new (additional) list with Fischer time (3'+2" was my favourite) but I dropped the project (for now!) because we decided to build up our 40/20 with Permanent Brain = ON which I considered to be more important.

Best
Wolfgang

I agree that 3' + 2" is a very good time control for testing. If I had my own list that would be my time limit I think. But I don't think you would need a separate list for it, you could just allow it as a time control for your blitz list, just as you already allowed 40/3 even though it was longer than your older list used. You would just use the 40/3 control for old engines that do not support increment (and their opponents of course!).

Best,
Larry

[Modern Times]

I agree that 3' + 2" is a very good time control for testing. If I had my own list that would be my time limit I think.

Best,
Larry

Yes, coincidentally I've been considering some private testing at this time control.

[ThatsIt]

[...snip...]
You do a great job Ingo, but you need to be careful that your list doesn't become the Ipontificate list.
There is no right or wrong way. It is entirely a matter of preference.

Best wishes,
G.S.

[ThatsIt]

[...snip...]
EDIT: Thinking about it a time control like 1 move in X minutes is the logical conclusion out of the argumentation for a repeating time. Boring in drawish endgames but somehow logical.

The best timecontrol is "x moves in y time repeated", because the
program knows in every stage of the game exactly how much time is left.

Best wishes,
G.S.