Unrealistic play | Forums

[Netgearz22]

Like you guys, I'm really excited about CM10 coming out soon. I was playing a mock tournament in CM9k with a number of random computer opponents and was astonished at the following line: (Keep in mind that I haven't played CM9k in a LONG time, so I'd forgotten how absurd some of these sacrifices can be)

Me-Alicia(about 1335 I think, the lowest seed)
1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. g3 Nb4 5. Bg2 Bg4 6. a3 O-O?? 7. axb4 Bxb4, and so on. I traded down, won a few pawns in the process, which is to be expected given her rating and won easily. My point is that a 1335 player would NEVER play 6...O-O. I mean, even a beginner would respond to the a2-a3 push, there's no discretion involved. I would only expect that sort of sacrifice(which isn't even a good one at all)from a 2000+ player down at my club in an effort to cramp me.
I have no idea what my rating is, it was 1475 a while back. I play on US Chess Live and maintain about an 1800 rating std there, but that means it could be anywhere from 1300-1600 USCF, it's a real grey area.
Anyways, my hope is that in CM10 we'll see a more realistic style of play from these lower-rated computer opponents. I'm teaching my friend how to play from scratch and have encouraged him to pre-order the game, so I hope that he can practice against the comps and gain good practical experience.

[Netgearz22]

Like you guys, I'm really excited about CM10 coming out soon. I was playing a mock tournament in CM9k with a number of random computer opponents and was astonished at the following line: (Keep in mind that I haven't played CM9k in a LONG time, so I'd forgotten how absurd some of these sacrifices can be)

Me-Alicia(about 1335 I think, the lowest seed)
1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. g3 Nb4 5. Bg2 Bg4 6. a3 O-O?? 7. axb4 Bxb4, and so on. I traded down, won a few pawns in the process, which is to be expected given her rating and won easily. My point is that a 1335 player would NEVER play 6...O-O. I mean, even a beginner would respond to the a2-a3 push, there's no discretion involved. I would only expect that sort of sacrifice(which isn't even a good one at all)from a 2000+ player down at my club in an effort to cramp me.
I have no idea what my rating is, it was 1475 a while back. I play on US Chess Live and maintain about an 1800 rating std there, but that means it could be anywhere from 1300-1600 USCF, it's a real grey area.
Anyways, my hope is that in CM10 we'll see a more realistic style of play from these lower-rated computer opponents. I'm teaching my friend how to play from scratch and have encouraged him to pre-order the game, so I hope that he can practice against the comps and gain good practical experience.

[Grim_O]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> 1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. g3 Nb4 5. Bg2 Bg4 6. a3 O-O?? 7. axb4 Bxb4 <HR></BLOCKQUOTE>
Your line is miss spelled. Black cannot play Bg4 nor O-O. Perhaps it's Be7 instead of Bg4? Anyway the sacrifice is a typical computer one, yes. I too hope they balance the personalities a little more for the 10th Edition (in the demo they play ok to me) and since the engine has been improved so should the personalities.

[Netgearz22]

What's interesting is that if you watch the mentor lines in the Game Room and set it to like the 10 best moves and play a computer, you can actually see it choose moves that aren't so great. I swear sometimes it waits and examines deep down nodes just so it can decide what move is worst in a given situation. But yeah, bad players are hard to simulate, but a 1350 comp player(which I guess is supposed to be comprable to USCF) shouldn't be making those particular types of blunders. Something like a conditional statement that would force a response to the previously played move if it's obviously bad to ignore should be in place. Psuedocode: "If previousmove.threatensPiece() == 1, then forceDefendPiece(). I know this is very basic, and would need much more complication to work properly, but you get my idea.

[Netgearz22]

Well, here's an idea I've had for a long time about computer chess engine programming. Along with the mentor line idea of having the computer score the game and positions following the given move, assign a strength of move number to each candidate move. For example: a pawn push at a knight. Mentor lines say that anything other than moving the knight results in a 2.8 point loss, and the SoM(Strength of Move) needed to see the knight move is ~800, whereas if you're in a tense tactical situation where you could stand to lose 2.8 points, but the defensive move is very difficult, assign a SoM of 2000. Now, how to program the SoM function accurately is another question altogether.

[Spinne_3.-JG51]

Is it possible to realistically make a computer chess program both very strong and very weak without having the very weak part appear fake? To program the strong part in, you'd take a really good player and put his logic into the game, right? But then the program can't make realistic weak moves by using simple tricks like trunctuating the analysis at a given point, or searching for lines of play that involve material loss. To get a truly weak opponent, you'd have to put another engine together, imo.

[c1arinetboy]

Good article . It's nice to have an explanation for CMX's random-looking sacrifices, although I still think Ubi did a pathetic job of covering the "odds" up.

[Xpr3sso]

Agreed that was a great article that hit home about how I feel when I win or lose against either the computer or a human opponent.
I also agree with clarinetboy. I'm playing a lot of games on Shredder at the 1400 level and the program beats me more than I beat it. At 1400 Shredder will not give away very many pieces; you have work some to earn a victory. It will allow for an occasional fork with a protected pawn or miscalculate a complicated capture sequence just like a human at this level. On the other hand in two recent games with Cal at 1507 on my machine both times it literally gave away both knights in each game.
This forum might be a good place to start tracking the good and bad personalities.