Author: Robert Hyatt
Date: 19:09:29 04/06/00
Go up one level in this thread
On April 06, 2000 at 17:48:47, KarinsDad wrote: >Here are your issues as I read them: > >1) Rules should not be changed such that someone shows up at a tournament and a >last minute change results in the program not being coded correctly. > >I agree with you 100% on this. > >2) It is not desirable to have a separate way for a program to play the game for >computers as opposed to humans. > >Fine. Indicating the type of opponent at game start does not seem like a biggy >to me and in fact, you could have conditional compiled code for the 50 move >stuff so that you run a slightly different version for each, but ok. > >3) Programs should not have the ability to announce a >50 move mate and extend >the 50 move rule since other programs may not have that ability. > >I could easily disagree with this one. A GM cannot announce a >50 move mate >(usually) due to lack of ability. GMs just do not know the endgames to that nth >degree. But, a computer can. So, because a computer is "smarter" (i.e. can >search deeper into the endgame due to having EGTBs), why should a position that >is a forced win be a draw due to a human limitations rule? But, ok, I will drop >it as it doesn't occur that often. I think you simply have to ignore this at the present. A solution is known, and has been known for many years (store both DTC and DTM in the database.) At present it seems to be more interesting to ignore this and get into 6 piece files (my ftp site now has 32 gigs of databases for public consumption)... but before long, the interest will switch to solving the mates for real, once we start getting table mates in 300+. At present the number of mates that violate the 50 move rules are very infrequent. In the 6 piece files they will become more common. In the 7 piece files more so... So one day, this problem will totally disappear as the databases will let us go for the shortest mate that lets us do a 'conversion' prior to the 50-move rule. I can see how DTC will solve the case of a mate in 80, as you only need one conversion to mate... but for mates in 200, I am not yet sure how to solve this, as you might have two moves at the root, one that says mate in 300, conversion in 48, the other says mate in 320, conversion in 47. But after the conversion in 47, you discover that the next conversion is 60 moves away and you are dead... so it will take a good database, plus a search, to make sure that the final mate is reachable without violating 50 moves. But it seems doable, if it isn't too slow... > > >But, I think you were missing my main point in the earlier post. The reason for >the differences I listed was not that I think we should re-write the rules of >chess (and that I was trying to come up with obscure b******t examples). I think >we should brainstorm some ideas on how to make computer chess rules better. > >For example, since you are using human "blind" rules, why not upgrade computer >chess rules to human "normal" rules? This is impossible. Because a computer can't 'touch' a piece.. If the operator makes the wrong move, no one wants to make the program have to live with that bad move... the CC events are all about computer vs computer, not computer vs computer with human mistakes influencing the outcome... > >How do you currently keep track of a scoresheet? I assume that most operators >keep a scoresheet next to them during the game. That way, if there is a building >wide power failure, there is still a record of each game up to that point. > >Why not have a printer next to each computer where each program prints out the >current move and time? This would more closely emulate keeping a scoresheet >which is a rule for normal human chess. And, if a program does not yet have that >ability, fine. It would need an operator to do that for it as per the "blind" >chess rules. This isn't needed. Machines don't lose hard drives on a power failure. So the history could be (is in my case) written to disk. Or you could use a floppy... > >Another example is in the actual play. Why have a board at the site if both >computers could be hooked up to each other and playing via one of the >interfaces? An additional "official clock" interface could be created and the >sponsors of the match could maintain a computer with all official times and even >moves for that matter. The official computer could then post out this >information out to the Internet or something. This type of thing could also >minimize human operator inefficiently in sudden death time controls. > we do this on ICC. And it would be desirable at CC events as well. Ken Thompson and I stirred this issue up several times. It didn't work in the 70's/80's because of the 'dedicated computers' that didn't have a standard serial I/O port... >The point I am trying to make is to not limit our thinking to traditional >human-centric chess and human-centric chess rules and to come up with ideas on >how to take advantage of the computer's strength for running computer/computer >and computer/human tournaments. And maybe we could extend the current rules at >these types of tournaments to take advantage of technological improvements. > >That's all. > >I'm not trying to re-write the rules of the game. > >KarinsDad :)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.