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

BGonline.org Forums

XG v SW v GNU Opening / Second Move R/O

Posted By: Nack Ballard
Date: Saturday, 2 January 2010, at 4:58 p.m.

In Response To: XG v SW v GNU Opening / Second Move R/O (Michael Depreli)

Is someone Stick?

An interesting philosophical question :) Actually, this is the way I read your question before I realized that what came after it was part of the same sentence.

Is someone Stick? planning on posting all the XG 1st and 2nd move rollouts online at one location ?

It wouldn't surprise me in the least if he's planning to add XG rollouts to pages he already has (currently with Gnu and sometimes Snowie rollouts posted).

On a related topic, I've been meaning to recommend to Stick that he change one of the template pages of his rollout navigation from

21S-11
21S-21
21S-41
21S-51, etc.

to

21S-11 [Snw data] [Gnu data] [XG 3-ply data] [XG 4-ply data]
21S-21 [Snw data] [Gnu data] [XG 3-ply data] [XG 4-ply data]
21S-41 [Snw data] [Gnu data] [XG 3-ply data] [XG 4-ply data]
21S-51 [Snw data] [Gnu data] [XG 3-ply data] [XG 4-ply data], etc.

The columns would be entitled something like: "Position," "Snowie," "Gnu," "XG 3-ply" and "XG 4-ply." (I tend to order the bots chronologically, but you may prefer otherwise.)

For the data format, I suggest nacbracs (because I don't know an alternative nearly as compact.) Depending upon space and choice of layout, you could include all the plays in the rollout, or only plays within .03, or just the top two or three, whatever. A single k number could be listed that refers to the top two plays, or the k trials could be omitted.

The idea is that the reader could see the encapsulated result at a glance, and also for which bots a rollout has been included. If he wants to instead or also see rollout details, he can click on the "21S-21" just as he does now. (The reader might want to have a closer look only at the results that surprise him.)

Another benefit is that a reader can easily see which bot-rollouts are not yet posted on the site (there will be empty spaces). If he has done a rollout, he can send it to you, Stick (and you can see if you really don't have it or if it's just not posted, if it's better than yours or supplemental to yours, etc.). He can also ask you if you would like him to do particular rollouts that are missing and interest him.

I, for one, would send you rollouts faster if I could work off such a list rather than having to tediously click an extra level for each position and then scroll and find numbers that are far apart.

Would anyone else like to see the navigation pages laid out this way?


I think Nack alluded somewhere that XG seems to be aligning itself more to Snowies rollouts. Has someone got any overall data on that? If it's very heavily weighted towards Snowie then should we be concerned that the gnubg rollouts may not be reliable or is it just a matter of playing styles?

My observation that XG 3-ply seemed to align itself with Gnu and XG 4-ply seemed to align itself with Snowie is based on too few sample points to yet draw any meaningful conclusions. I value both Snowie's and Gnu's rollouts so it's hard to know what to make of it. I just brought it up because I started to notice a pattern and I thought it bore watching.

My larger concern is that XG 4-ply has disagreed with XG 3-ply a high percentage of the time. Again, that could be variance from too few sample points and/or too few 4-ply trials. To be fair, I have been pushing XG rollouts (and others have gravitated towards them as well) for positions in which Snowie and Gnu disagree, and one would expect a greater chance than average that XG rollouts of different plies will disagree with each other on those positions.

Different bots different plies different results..should we be averaging the results or if we believe say XG 4-ply is the strongest then weight it much more heavily in that direction?

For purposes of determining what I consider to be current "value" I do average all the data that I have available in some way (and my choice of weights is always being reconsidered and in subtle flux).

That said, I agree with Tim that it is better to present the data bot by bot, ply by ply, at least until a broad consensus has been reached about the most useful and reliable and practical (speed-wise) information.

For example, I am going to trust a XG 3-ply rollout of 25k more than a XG 4-ply rollout of 10k because of variance (though I would absorb both, along with the other bot data, into my perception of value). But someone else might feel differently: he might trust one bot's data above all else, or he might suspect a bug in one of the plies, or whatever, causing him to discount or ignore information I might not.

Nack

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.