[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Seeking Xtreme speed (here's what to expect...my opening 64 RO and...)

Posted By: neilkaz
Date: Saturday, 21 November 2009, at 7:17 p.m.

In Response To: Seeking Xtreme speed (eXtreme Gammon)

1. Rollout 8/2 6/2 eq: +0.018 Player: Opponent: 50.02% (G:15.83% B:1.03%) 49.98% (G:14.33% B:0.81%) Conf: ± 0.003 (+0.015...+0.021) Duration: 5 hours 07 minutes 34 seconds

2. Rollout 24/18 13/9 eq: +0.015 (-0.003) Player: Opponent: 50.50% (G:13.99% B:0.84%) 49.50% (G:13.87% B:0.77%) Conf: ± 0.004 (+0.011...+0.019) Duration: 6 hours 44 minutes 48 seconds

3. Rollout 24/14 eq: +0.010 (-0.008) Player: Opponent: 50.69% (G:12.27% B:0.70%) 49.31% (G:13.38% B:0.63%) Conf: ± 0.003 (+0.007...+0.013) Duration: 6 hours 20 minutes 28 seconds

51840 Games rolled with Variance Reduction. Moves and cube decisions: 3 ply (I used normal search interval for this)

I posted this at Kit's site during the week when this site was down. This is using my great computer (HP server) on the trading floor at work. Is is like eight 2.4 GHz CPU's so a bit less than a 2.66 GHz i7. XG 1.10 beta calibrates to about 1,600,000/sec on that computer. Note that in little more than 18 hours total XG did over 155,000 total games from the starting position !! That is more than 8000/hour and about 140/minute !

Many positions that occur while studying can yield statistically significant results in a few hundred games which only takes a couple minutes using a HT Quad Core machine (8 effective CPU's) ..ie this doesn't take long at all, in general when studying in real time.

XGR+ comes with the soon to be released 1.10 and based on what Xavier told me the last time we discussed it, it is 6 times slower than XGR, but 7.2 ELO stronger. That makes it about as strong as 5 ply, while being 3 times faster. If Xavier has done more exhaustive strength tests, he can update us. Preliminary New Depreli position data also indicates that XGR+ compares favorably with 5 ply. As I observed, when aiding in its design, XGR+ makes fewer bad takes than XGR.

On the 8 core machine at work, I can play games vs XGR+ and it moves at human-like speeds. It can think a while frequently, but I doubt that it would time out if playing an 11 pt clocked match at ABT time controls although it would certainly use some reserve time. Sometimes it is good to play against the best possible opponent, and I cannot think of one stronger than XGR+ that will move at an acceptable pace.

.. neilkaz ..

Messages In This Thread

 

Post Response

Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:

 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.