online chess

Online Chess

Participate and you will discover!
Crazy name, real passions
[ Sign up | Log in | Guest ] (beta)
loreta 80 ( +1 | -1 )
Some thoughts about computer playing Eventually, I've gone through chesscafe and 'hanged' on an one article about computers against human {at first, I liked a note that computer program's operators take all things more enthusiasticly and are more excited during these 'matches' than human players... They (operators) looks for the blood}

At while reading one thought hit me and I want to share it.

I want to know if somebody who has chess programs noted IF that exists?

Let's take same program with the identical settings and run on different computers. That SAME program could make DIFFERENT moves on different computers...

Does anybody noticed that? Logcally thinking it could be!

As result - the same program with the same options could with against itself, or not?




olympio 41 ( +1 | -1 )
ALSO it would also make different moves because one would be playing white and one would be playing black. also because opening moves are chosen somewhat randomly with a general tree in mind. but out of the opening book on the same computers the same program playing the same color in a game with the same time settings would make the same moves
loreta 49 ( +1 | -1 )
Explanation... I have in mind - playing the same color and when moves are 'out of book' - that is when a prorem calculates a move itself.
.......................
Of course, it's not always but culd happen {especially in very complex positions} and more often when characteristics of computers differ much - for example - 2 times faster CPU, faster access to memory {newer chipset and bigger cashe memory} and so and so...
.......................
It would nice if somebody having these programs could check that...
adrianallen 80 ( +1 | -1 )
There can be a big difference when playing on different spec machines. I analyzed a game that a friend played and on one particular move it took my computer over 2 hours to discover that it was superior. On my slower machine it didnt discover it at all.

In ChessMaster 9000, I rated 200-300 points higher if I install it on Windows 2000, as compared with windows 98.

If I put GNU chess on my faster machine and Chessmaster on my slower machine GNU chess wins, if I swap them round ChessMaster wins.

There use to be a chess computer/program that stored previous games played and checked against these to see if a sequence of moves had been played before, it would then vary the moves if it lost and keep them the same if it won. Do any programs still do this? also does anyone remember it?


olympio 11 ( +1 | -1 )
that's difficult that's hard to do cause how can the program know which move it made that was the losing move?
coyotefan 38 ( +1 | -1 )
Fact is Computer programs, be them Fritz, Chessmaster, or others only have a rating of around 1800 when played on a typical PC with a 5 minute max time per move. Play them on a slower machine at a 5 minute max and they are only around 1500-1600. The programs that Kasparov played was the same one that we buy over the counter, but it was on an unbelievable machine.
bogg 144 ( +1 | -1 )
coyotefan The machine that Kasparov played against was a parallel processing machine. That incurs a lot of overhead. As far as chess playing goes the machine used against Kasparov was only about 3 times faster than a computer you can buy today for about 700 dollars! An average desktop computer plays at between 2500 and 2600 Elo at 40/2 time controls today.

I can not explain why Kasparov and Kramnik did so poorly against the bots. Maybe it is a matter of style. I would have expected Petrosian to have scored about 100% against them if they existed back in his prime. There are whole classes of positions that the bots can not begin to play and unless the algorithms can be changed to handle them brute force will never be enough. The bots have no concept of permanence. The fact that a piece is permanently locked out of play is only a minor inconvience as far as they can tell. 'Sure I haven't been able to move that rook in my entire search but that was only the first 10 moves', seems to be their thought process if I can be allowed to anthropomorphize a bit.

As far as slower machines go, every 7 fold increase in computing power and 10 fold increase in memory gains about 200 rating points in strength. You would have to go back about 20 years for the average desktop computer to be about 1800 strength running a program of similar quality to those available today.
lexherman 22 ( +1 | -1 )
Nemeth gambit Try this one against a program..
!.d4 h5 2. e4 e5....ok an example against an early Fritz (Fritz=white) it continued..3.dxe5 Nc6 4.Nf3 Be7 5.Nc3 a6 6.Bc4 Nh6 7.0-0 d6 8.exd6 Bxd6 9.Bd5 Bg4 10.Bc2 Na7 11.a4 f5 12.h3 f4 13.Bxa7 Rxa7 14.hxg4??hxg415.Nfd2 Nh6 16.axb3 Bf8 17.g3 Qg5 18.Ba4 Qh5 19.Nf3 Nf5 20.Nh4 Nxh4 21.bxa6+ Ke7 22.Re1 Nf5 23.Kf1 f3 mate
gta3master2987 6 ( +1 | -1 )
uhh... just one problem... 10. Bc2 cannot be playeed because of the pawn on c2.
lexherman 49 ( +1 | -1 )
correction Nemeth Sorry i mixed up two games. The game above is correct untill move 10. It is not the gambit but a game with a sacrifice of a bishop.

10.Bxc6+ bxc6 11.h3 Qe7 !2.hxg4?!..hxg4 13.Bg5 f6 14.e5 Qf7 15. exd6 0-0-0 16.Ne5 fxe5 17.Bxd8 Qh5 18.f3 g3 19.Re1 Ng4 20.Kf1 Qh1+ 21.Ke2 Qxg2+ 22.Kd3 Nf2+ 23. Kc4 Nxd1 24.Raxd1 Rxd8 Black won. (this game was played at blitz level)

Now about this absurd anti-computer gambit.(invented by mr Nemeth)
1.e4 c5 2. Na3 Nc6 3.d4 cxd4 4.Bc4 Nf6 5.h3 Nxe4 6.Bxf7+ Kxf7 7.Qh5+ Ke6 8.Qg4+ Kd5 9.c4+dxc3 10.Be3...
Try this one against your program if you have one. Have fun...
olympio 17 ( +1 | -1 )
that that second one looks nice and good i just doubt the computer would actually make some of those moves..

why wouldn't it play Kg8 instead of Ke6
finduriel 40 ( +1 | -1 )
coyotefan I don't think your "fact is" correct. Fritz (according to the manual) has a rating of Elo 2721 for a five min. blitz game, Elo 2661 for a 25 min game, Elo 2581 for a tournament game (2h for 40 moves, 1h for next twenty moves) on a 1GHz Pentium machine. Even on a more "average" machine, the rating won't be as low as 1800, unless you give the program only 10MB RAM to work with or something like that.
finduriel 15 ( +1 | -1 )
I see that bogg has already spoken on the issue. Sorry for reading superficially. :)
olympio 26 ( +1 | -1 )
coyotefan I'd also like to point out that putting them on a 5 minute max would increase their rating not decrease it. Programs play much better as the time control gets shorter.

I'm merely pointing this out to be educational not as an attack on you, coyotefan.
finduriel 48 ( +1 | -1 )
olympio That's why humans do quite well against machines in correspondence or "turn-based" chess, I guess. Of course, no offence to coyotefan!

"Programs play much better as the time control gets shorter."
-> Only if it gets shorter for their human opponent too, naturally. :) I assume programs play better chess the longer they are allowed to calculate - it's just that human players play much worse at 1 sec per move than machines do...

Fin
olympio 24 ( +1 | -1 )
venture i'd venture that fritz 8 on a machine similar to the one finduriel suggested (1.0 GHz which is actually about half the average now i'd say) in a 2 minute game would play near world champion level. and in a 1 minute game would be invincible
isiah 8 ( +1 | -1 )
Different My program played a different move completely on the first Nemeth Gambit--9. e5
soikins 52 ( +1 | -1 )
Same moves? I don't think a computer would repeat the same moves every game. Too many factors are implied here. If, let's say we record two engines game of blitz. And then try to repeat it again, starting from the position right after the opening, I don't believe we would get the same game. I have tried this on my computer, but unsuccessfully. Computer must repeat all the same calculations that it did in the first game and in the exact same time. It is very hard to do that even on the same exact computer.
bravesfan4ever 40 ( +1 | -1 )
another point did you realize that when you playin against a program, espeacially ones with tons of data base, its like playing against all the grandmasters of the past and present, and all of the experts around the world. Of course the program would take very long to find the best move possible, but its still like playing against a bunch of GMs and experts... that was just a thought
caldazar 224 ( +1 | -1 )
The same engine with the same settings and running on identical hardware will generate the same moves. There are a few caveats, though. If the engine is using an opening book, different moves could be played. Also, the same engine could potentially generate different moves depending on where in the analysis any main line switching occurs. For instance, suppose if an engine analyzes a position for 30 minutes and throughout those 30 minutes, it thinks Move A is best. Then, right at the 30 minute mark, it switches to Move B as its primary move. If the test is then run again, because of random occurrences, it might take the second engine 30 minutes and 1 second to find Move B. So if both engines are required to output a move right at the 30-minute mark, you might get two different results.

One engine on a piece of hardware can win/lose to an identical configuration due to horizon effects. As an example here, suppose in a certain position Computer A is playing an identical Computer B and both are analyzing the same position with Computer A to move. Both engines see that Move A is best after a 16-ply search. However, the refutation to Move A happens on the 17th ply. Computer A plays Move A and now both Computer A and Computer B analyze the new position. After a 16 ply search, both machines determine that Computer A is busted. This is vastly oversimplified since engines search to various depths depending on the characteristics of a particular line, but the general idea is there. It's also why one has to be careful in using computer chess engines as analysis tools; you constantly have to double-check the engine's work to make sure its not feeding you garbage lines, either due to horizon effects or because it is severely misevaluating strategic factors.

I was analyzing a position once and the position I was interested in arose after:

1. e4 c6 2. d4 d5 3. exd5 cxd5 4. c4 Nf6 5. Nc3 e6 6. Nf3 Bb4 7. cxd5 Nxd5 8. Bd2 Nc6 9. Bd3 Be7 10. O-O O-O 11. Qe2 Nf6 12. Ne4 Nxd4 13. Nxd4 Qxd4 14. Bc3 Qd5 15. Rad1.

The computer kept insisting on looking at 15... Nxe4 16. Bxe4 Qxa2, which just looks crazy to me. I was curious, so I left my machine running all night (admittedly, it's a pretty slow machine) and when I woke up, the computer had finally seen that this was not necessarily a good line to play into.
felixmiszta 54 ( +1 | -1 )
You did not realize... ... the most important reason, that the top chess software may generete diffrent moves even on the same computer: Engine of Fritz and some of chessprograms is based on Artificial Neural Networks. The mechanism of those networks could give two diffrent results on the same machine in the same time; it is hard to explain with my weak English, but pepole interested in this topis could read i.ex.
hem.hj.se/~de96klda/NeuralNetworks.htm

Sincerly,
Felix
soikins 61 ( +1 | -1 )
Hmmm... As far as I know chess engines don't run on bio-computers... At least I don't have a biocomputer at home... Thought, I'm not an expert on computers...

"The same engine with the same settings and running on identical hardware will generate the same moves." What about positions where several lines, say 10 lines, fall within evaluation of 0,06? What makes one think that a computer will always choose exactly the same move next time it analyses the same position. If it is a game, then just one second of extra thinking would make computer play a different move.
ccmcacollister 160 ( +1 | -1 )
Soikins, a Question... Your last 2 remarks posted, turned around, suggest that In the situation you gave, if that computer does NOT take just one second of extra thinking time(or whatever fraction of a second is needed to make a change, I assume) then it WILL PLAY an identical move.(If no deliberately built-in randomizing effects) So my Question is: "What makes one think that this computer will ever Use/Require just one second of extra thinking?"....to thereby Cause NOT exactly the same move?
.....I'm not debating you on the point. Just asking out of my interested-ignorance on it. Do some programs,for instance, consider how much time I"VE USED against it, & start moving quicker also, to keep pace if I move faster? Or if it sees ME as being short on time, will it move faster and try to "Flag" me out, on time?
....Or perhaps if it's been turned on longer or for any reason has more internal Heat built up, after all the same moves, can that add-up enough at some point to lower processing efficiency enough to shorten its analysis time/length by one move(or more)? If not these as such factors, what might they be? I can't think of others to make it vary the time it needs & uses to proceed thru what seems to me to require only an identical linear path of analysis. (?!)
..Conversely, I'd think it would consume different Time to get the same position by different moveorders. As if I'm WT & play 1.d4 e6 2.e4 d5 instead of 1.e4 e6 2.d4 d5., etc. A simplistic example for brevity; & with the proviso the machine has had to analyze it thru to generate the moves, instead of its Opening Book turned on.).
.
ccmcacollister 59 ( +1 | -1 )
To Caldazar, Oops another Q:... I had a Question to ask you also, similar to what I was asking of Soikins. What kind of event(s)/effect(s) do you mean, that could effect the computers analysis Time & move choice, when you refer to "random occurrances" in your posting above? Like I said, I'm not arguing against the concept. Just hoping to understand it. {I've done only a very little amount of programming; BASIC only. And only starting to learn my Windows 98. Aside from those, I have cumpoter-cunfosion! Appreciate
any enlightenment from either of you, on this here.Thanks}

loreta 106 ( +1 | -1 )
To ccmcacollister What could mean "random occurrances"?!
1) A processor never run at exactly the same speed. If it's written 2.4 GHz, that means one second it could run at 2.399999999 and other second at 2.39999998 :-)
2) Common computer performance depends on many factors:
a) Are there another programs running (They take additional memory and time)?
b) Memory structure. To run, programs allocates some area of memory. Structure of memory never is the same and area allocation of memory take a slightly different time.
c) A busyness of bus. To get data from memory/disk and so or put them to, programs use a bus. It it busy at the moment, for example, Windows performs at the moment some its task, that could take longe period.
d) and so, and so, and so .... :-)
------
So there a lot of "random occurrances"... Only question, how can they influence a selection of move (at what level - they INFLUENCE, of course!)
------
To felixmiszta: only these "random occurrances" could make different result for algorith based on "Neural Networks", too (or, otherwise, a possibility of different result is provided in algorithm itself).
ccmcacollister 111 ( +1 | -1 )
to Loreta; speed/multitasking Thanks Loreta. I see now. I'm sofar behind in the computer chess & computers...Still tend to think in terms of dedicated boards & single task PC's. Last year I was filled with wonder to find I could download & read forums @ the same time. Unlike late 1980's on the XT!
...For anyone. Like to ask another Q.
My Novag Constellation board used to have "arrogance" built in, it advertised, & truly it would not take a draw with even material. My CM2000 or 2100(I forget) was beyond arrogant to "stupid" & would never reallytake one even more than a piece down(on my slow machine then, anyway). Is there a Chess Program out there now that will 1)Know theoretical EGm draws & accept(aor Offer, even?) then?
2) One that will actually make a fairly accurate positional assessment even in a MGm, and accept or offer? If so, only if it stands Worse in reality? Or perhaps can (If set to , perhaps?) assess well, as Even or Equal, & would then take a draw?
......The latest I've seen is KnightStalker, which will not do any of this.Thanks.