Guys,

See forum entry on time management below form Kurt's site.

Any help concerning this item is greatly apreciated. Hope everybody can read this.. It is not translated in German.
Specifically i am looking for a general time management system. Within the engines i will use you can tell the engine how much time you will give per move. Resurrection (Ruffian) also can be told how much moves there are left before time limit will be changed or will be reached. I have noticed that when i tell it all remaining moves and the remaining time it will use up too much time per move. I cannot reach Perola about what is going on here, so i have to adjust for myself ....

So for a '30 minutes / all' i program Ruffian before every move with the info that it has 5 moves left and calculate the remaining time for it.

For example on a 30 minutes / all game and being in the 20th move (based on 60 moves per game) while used 12 minutes i tell Resurrection it must do the next 5 moves in (30-12)/40*5 = 135 seconds. Now Resurrection will use some 36 seconds a move.

Resurrection wil have now 27 seconds per move because it has used more time for the first 20 moves than the normal 10 minutes.

Especially i am interested in : The algorithm for a x minute match. For example should i give Resurrection more or less time based upon the current score value. Or is it dependent on which movenumber it is.. I am looking off course for a general algorithm, and not a detailed one based on the position. This should be in the program istelf... And i am not the programmer....

Forum entry :

'Close call, a tie is a good thing. Leaves everything open for the future. I am curious though how the Res and the R40 will do on a 30 (avg) second per move match. That is how i won Kaufbeuren (by the way, i took a draw against a R30 as well).

How long were the average thinking times of the Resurrection and the R40 in a 30 minute match ? Did Resurrection lose games on strange time behavoour during game play ?

Furthermore, Ruffian itself has a strange time mechanism. I have programmed a mechanism around it which explains the time taken per move.

If somebody could help me with a algorithm for a x minute match. For example should i give Resurrection more or less time based upon the current score value. Or is it dependent on which movenumber it is..

Questions which are difficult to answer. Maybe you guys can answer this.
Another question is : you can enter how much moves on average there will be in a match. This is set at 60 (but changeable). At this time i evenly arrange time (30 minutes per game on 60 moves) means 30 seconds average per move. But what to do if the 60 moves are in sight ? Should i lower to 20 or 10 seconds a move ?
Any ideas are greatly welcomed.'

Ruud Martin

__________________
Grusse,
Ruud Martin

hi Ruud

'Close call, a tie is a good thing.
a very good thing
it seems the Res and Tasc's are very very close
the Tasc however will never get any better and the Res will ONLY get better

So for a '30 minutes / all' i program Ruffian before every move with the info that it has 5 moves left and calculate the remaining time for it.

to be honest i think you have already enabled the Res to play a correct "ALL' game
this is to say the user can set the "ALL" to 98 moves(it will switch back to 20 after 98 and does not allow 99)

i think its correct to set this once and not manipulate it after every move


there are several computers mainly Saiteks which consider an "ALL" game to be ALL /99
some computers even offer choices like this when switching levels:
5/99
10/99
30/99 etcetc

however i used the 60 moves for all games ..i had no time over steps

How long were the average thinking times of the Resurrection and the R40 in a 30 minute match ? Did Resurrection lose games on strange time behavoour during game play ?

R40 moved quicker ..avg was about 30-45 seconds per move
it never took more then 90 seconds for a move

Res avg move was about the same.however it would sometimes go into 'DEEP " thinks taking as much as 3 minutes a move
this happenned about 3-4 times in the 5 games
but it had nothing to do with the time control..this "DEEP" thinking could occur right out of the opening,or even in the endgame

in game 3 which had an early exchange of queens it did not once go into a "DEEP " think

"Deep" Sjeng Regards
Steve

How DEEP is your chess module ? Or should i state when ?

Greetings

Ruud

__________________
Grusse,
Ruud Martin

Hi Steve, Hi Ruud,

It seems to me, that setting "ALL"to 99 moves for Action Chess is the best solution at the moment. At move number 40-50 Res. has about the same time-consumption as London 68030!
sure, but does "99" by the Saiteks really mean "99" or does it simply mean "unlimited" number of moves"? What happens with the Saitek's time, when move 99 is reached? I don't think, that they overstep, do they?

In my database of about 1100 (self-played) Computer-Computer-matches there are about 25, that go over 100 moves. Ok, we can say, these long matches are so rare, that we could ignore it and let's take Res the risc to lose on time.

But then we had to play every game (at least these where Res. has advantage!) until his opponent is checkmated; because if Res' opponent could make it to 100 moves, it would win on time...
I did it in the first 4 matches against London too; but one of these Res. lost on time...

Setting "ALL" to 99 (or 98?) seems to be the best (in my opinion), because I often have games with more than 60 moves. Now we could argue, if 99 or 90 or 80 is best...

But I already suggested an "ALL"-time-management here and in Kurt's forum (newer version is in Kurt's forum! ) I would be interested in your and of course Ruud's opinion about it!


curious regards,
Robert

But how abpout the following suggestion:
In general, at xxmin/game, Res should have about xx/60 as average thinking-time per move. And when the last 10-15% of the xx mins are reached, it should be reduced to 30% of xx/60. And when the last minute is reached, it could be reduced further! (1s/move?)

Of course, we can discuss about it... But it should have nothing to do with number of moves played!


its a fine suggestion of course but i think game/98 is perfectly acceptable


Best Regards
Steveb

Attention Robert

VERY IMPORTANT CRUCIAL NEW INFORMATION!!

the number of moves can be set to 99
if you first go to 98 then it switches to 20 then use the left arrow and it will cycle down to 19..18...1..then 99

i know this is highly important information

)


News Breaking Regards
Steve

Hi guys,

somewhere else
- actually here: https://www.schachcomputer.info/foru...61&postcount=2
I suggested the simple algorithm always to take the remaining time in min, say x, and to tell Resurrection to use it as x sec/move. This would lead in Ruud's example (30 min game, 12 used) to (30-12) sec/move = 18 sec/move at that moment.

Of course another simple solution would be indeed to use 30 min/99 moves and maybe (inside the algorithm) to switch after (30-5) minutes to 5 min/99 (additional) moves or something (Or after 30-6 minutes just to be sure or whatever).

Or one could use a mixture of those rules...

Best,
Martin

Giving it another thought, maybe just a very simple

80% of the total time for the first 60 moves,
10% for the next 40 moves,
10% for the next 60 moves

would be sufficient (or 70%, 20%, 10% or something)...

Thinking Regards,
Martin

yes, i found this a few days before by accident!
because of this, i was a little confused about your "98"; I always went too far and had to go back!

What i don't know until now: does "00" maybe mean "100"?
I'm sure, this makes an enormous difference!


Robert

Hi Robert

What i don't know until now: does "00" maybe mean "100"?

i can test this very easily

if i play a game against Res and i can manage to even draw then we know it means all moves in 0 seconds


Best
Steveb