I'm not sure if I have to keep adjusting this depending on which color I'm analyzing with.
> does that mean that if I'm doing IA as black the contempt value for Black is 0, and 15 for White?
It'd mean -15 for Black, 15 for White.
I'm trying to play the same contempt whether I'm Black or White. If it's always from White's perspective then I'm afraid that keeping track of which side I last analyzed as would be difficult to maintain, especially if I have many correspondence games going on. The only option would be to do IA with contempt 0..which I no longer want to do.
The reason I'm asking is because I'm trying to determine if raising contempt to 15 would make my R3 stronger if I do IA longer than my opponent, and my opponent has a contempt value of 0). Basically, I'm trying to avoid draws against weaker opponents and can't seem to find a configuration for R3 that accomplishes this. I seem to often win as white and often draw as black...almost for sure against equally rated players (which I'd expect to be normal) but against weaker players (100 or 200 less) I often draw. I'm afraid to raise the contempt to 15 or higher because if they're also using R3 but with a lower contempt then I may lose more often, but perhaps win more often too.
> asically, I'm trying to avoid draws against weaker opponents and can't seem to find a configuration for R3 that accomplishes this
The opening is the single most important factor when it comes to draw rates. If you're drawing weaker opponents with the white pieces, your opening repertoire is to blame. Contempt will do nothing to reduce draws if you're not reaching winnable middlegames. It's up to you to outplay your opponent, Rybka is not infallible - find improvements and drive the advantage home.
Ok how about this. I want black to analyse with a contempt of 15...what do I set the contempt analyze value to? 0, -15? I wish it was based on what color you analyze with, not from white's perspective. Too confusing.
> Actually, that would be a difference of 30 instead of 15...wouldn't black's contempt be 0 and 15 for white?
No. You're not getting the point at all. 15 for white is the same as -15 for black.
> I wish it was based on what color you analyze with
No. You wish it was based on what color Rybka analyzes with. You're not doing any analysis, it's all Rybka. The way contempt works is intended for interactive infinite analysis.
Just set a looooong time control (but not fixed time per move) and "play" against Rybka.
If R3 is analyzing as black (IA)...what contempt do I set if I want it to play with a contempt of 15?
> That's what I'm doing. Interactive IA following the suggested moves by R3.
Seriously? So you think it'd be easier to always make moves in pairs (because that's what side-to-move-dependent contempt would force you to do) than to change a parameter between analysis sessions?
> If R3 is analyzing as black (IA)...what contempt do I set if I want it to play with a contempt of 15?
You've got 1 correspondence game where you are playing using R3. Say one of the opponent's name is Vempele (W) against me using R3 (B). Assuming you're also using R3. I set my R3 to a contempt value of -15 whenever analyzing as black, to avoid a draw because I'm rated, for example, 100pts higher.
You've got another correspondence game where I'm playing R3 as white against Yojimbo. He's a stronger opponent (+50) so I want to use a lower contempt, say 0, and so that's what I set it to whenever I do IA as white against Yojimbo.
The purpose of all this is to experiment and see if using a contempt value that's dependent on your opponent's rating helps you win more games, rather than draw. If it means I may now lose a game for every 5 I normally would've drawn then I think it'd be worth it. The trick is, how to keep track of all this? Only way I can think of is writing down the contempt value for each of your opponents and remembering to change the contempt for each.
> Only way I can think of is writing down the contempt value for each of your opponents and remembering to change the contempt for each.
Or you could create multiple personalities.
In my opinion, Rybka Dynamic is the engine that benefits the most from adjusting the contempt. One of my favorite personalities is Dynamic with contempt at 35. Distributed/interactive analysis with this personality often leads to interesting discoveries - it's also very fun!
Or are you suggesting I use R3D only when I'm in the middle games or end games to try and squeeze out a win? I tried that anyways...it never changed anything.
I think using a contempt higher than 15 against top correspondence players will smack you with more losses where you'd normally draw. I tried playing top 5 correspondence players a while back and although I've never lost, I also never won. So nothing but draws. That's with a contempt of zero mind you. I think if I raised it and they had it at zero I'd probably lose more than I'd win, and still draw a lot. My guess is, out of 10 games with me setting contempt to 15 or higher, I'd have a record of: 0 W, 2 L, 8 D. If that was to be the case, I'd rather just draw them.
Do you play correspondence chess regularly?
> Do you play correspondence chess regularly?
Yes. Assuming I outlive my opponents, I'm 2.0 points away from my first norm with 3 games still to conclude. I played casual postal chess for years but since it's slowly going extinct, I decided that it was time to play correspondence chess with the big boys. So far, so good...my rating is almost 300 points higher than it was when I started (17.5-3.5 with a record of 14-7-0) and I'm further along than I expected to be at this stage.
From reading your post, it doesn't sound like you play CC in either the ICCF or IECG because the top players in both of those organizations rarely ever play anyone except other top players. You stand a better chance of playing a game with Zeus or Superman than you do with one of the Top 5 ICCF players. I hope you're not playing some place where engines aren't allowed. If you are, it's not "real" correspondence chess.
Contempt is a tool for measuring risk and generating new ideas. It's not a switch you can flip or a button you press that turns a drawn game into a win. It's one of many tools in the toolbox. I use it to see what I can get away with (e.g. test a speculative sacrifice etc) or I use it to add diversity to an IDeA tree or sometimes I'll use it for something as simple as helping me decide between two moves that look equal in all other respects. What I don't do is configure contempt before I go to bed hoping to wake up to a won game.
You have to make the engines work for you instead of the other way around. You're the boss, the engines report to you - they make recommendations, some of them are good and some of them are bad, some are risky and some are safe. It's the bosses responsibility to weigh all of the options and decide on the best course of action. Sometimes the best course of action is to start digging for new ideas. As I've said many times on this forum, engines are great at finding good moves but they do a very poor job of finding the best moves (in non-forcing situations). It takes time and a lot of biological CPU cycles to find the best moves....
One last thing, there are significant differences in the Rybka engines and if you never noticed them, it's because you weren't paying attention. Spending more time on interactive analysis will undoubtedly improve your results.
No, I don't play any GMs out there. It's simple engine vs engine games.
I have tried several times trying my own ideas in interactive mode using Rybka 3 but I've never yet found a move that scores better than what R3 found. Sure, I could make guesses, which I've tried but again...they're usually worse, and rarely I'll find a move that scores about the same at the same depth. So I'm no GM, and this makes it almost impossible for me to find that "best" move interactively. And even if I found the best move, I wouldn't really know if it was.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill