USA Basketball Returns to the Top.
6 years ago
Frequently Read Blogs
-
-
4 weeks ago
-
10 years ago
-
4 days ago
-
16 hours ago
-
14 years ago
-
14 years ago
-
10 years ago
-
13 years ago
-
-
14 years ago
-
13 years ago
-
-
12 years ago
-
8 years ago
-
7 years ago
-
9 years ago
-
6 years ago
-
14 years ago
-
6 years ago
-
15 years ago
-
12 years ago
-
5 years ago
18 comments:
For lack of a better place to post, when do you want to play our match? This is to determine who goes into the gold medal round against Ivan. :-)
How's tonight 7:15 PST?
I would only make that time if I have a horribly short game at the club tonight. I usually don't get home until after 11 pm my time.
How would Friday work for you?
Unless you crush your opponent in 15 minutes, I don't want to be responsible for such a short game at your club ;-)
Friday won't work for me... how about tomorrow 7:15 PST?
Thursday no good for me. How does the weekend look? With the time difference I might be able manage Saturday, Sunday or even Monday evening.
Sweet! :)
Polly: the weekend's no good for me... how about Tuesday evening?
We'll connect sooner or later :)
I could play on Wednesday because I'm going to play my club match in the afternoon. I'll probably stop by the club, but won't be in a prolonged match.
Wednesday it is! Is 7:15 PM PST ok for you? It'll be hard for me to get home any earlier.
Looking forward to our rematch :)
I just ran my first Monte Carlo analysis using Aquarium this morning (it runs 33% slower than the Fritz version of Monte Carlo). How neat for complex endgames, where engines are usually quite clueless even with TBs. :)
likesforests: cool! I've been reading up a lot over at the Rybka forums as to when Monte Carlo analysis makes the most sense, and why. I hope to write up a post soon with some examples of positions where Rybka's own evaluation deviates significantly from the results of its Monte Carlo analysis. There's quite a bit of informative commentary from Larry Kaufman in the forums(this is a pretty interesting thread, for example).
One example he gives is: Let's say that you are analyzing some position in two variation mode on Rybka 3, and you want to know the best move. If the two choices show as almost the same value (let's say 1 or 2 centipawns different), then run the MC on the two positions arising after each of those two moves, at whatever depth will allow you a few hundred games in the time you want to allot for this. If the MC shows a much better score for one position than the other, it is probably a better move.
And as you say, MC is definitely very useful in complex endgames, fortress situations etc.
good stuff :)
I found just such a game. At this page, Teichmann-Nimzowitsch, Karlsbad 1911.
I gave Rybka3/Human 6 hours to analyze the endgame (move 31+) and this was its output:
31...Qe6: -0.04
43...Be6: +0.18
51...Ne4: +0.56
55...Ke7: +5.50
57...Ng6: +5.50
Pretty useless.
In some positions there were two moves Rybka3 rated as equal (+/- .10). I ran Monte Carlo on them both and the output was:
Move 1: ==================
Move 2: ==================
Not too helpful either. I looked at some of the games and saw the stronger side kept trading down into a Bs of Opps ending... silly. I upped the search depth from 7 to 14 and that stopped happening. Now the output looked like this:
Move 1: +=+++=++==++++=++=
Move 2: ==================
Useful info! Clearly one line makes a draw easier than the other. Practically, move 1 is better. The Monte Carlo runs take maybe 10-20 minutes.
Now I'm curious whether Move 1 is a force win or not. So I use IDeA and it builds a tree, analyzing deeply at each node. It might goto depth 40 or 50 in its analysis (taking 4-8 hours).
If the final output is:
Move 1: 0.06
Move 2: 0.04
I know both moves lead to a theoretical draw but move 1 makes things easier for White.
So they are both interesting and complementary for analysis. :)
Both Monte Carlo and IDeA provide you with the good lines and bad lines tried I should mention... which is a big improvement over normal analysis output that only shows the "best" line.
Chess: 7:15 PST is fine. I'm a bit of a night owl so starting at 10:15 my time works.
likeforests: very interesting! IDeA is a feature I'd love to have, it sounds like a great tool. For now, I'll have to make do with Rybka's regular eval plus Monte Carlo. But as you say, both IDeA and MC are big improvements over regular engine output alone, so I'm a happy camper :)
Have you come across any fortress type positions that Monte Carlo evaluates significantly different than Rybka's engine eval?
I'll interrupt the chess geek speak, just to confirm that we're on for tonight at 7:15 PST for our game.
Hehehe... yeah I know, as if chess alone wasn't geeky enough :)
Yup, we're on for tonight, 7:15 PST. Not sure if I've got the black or the white pieces...
How many Americans dislike Kobe now?
Post a Comment