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

BGonline.org Forums

Best online data source for studying opening replies?

Posted By: Matt Ryder
Date: Tuesday, 22 December 2009, at 9:25 p.m.

I'd like to embark on a close examination of the 2nd roll replies. There appear to be two excellent, readily available sources of rollout data on the web:

1. Tom Keith's Backgammon Galore archive.

2. Stick Rice's collection of materials here.

Tom's data is immediately attractive because of the uniform way in which it is presented and the meticulous attention to detail evident everywhere on his site. With some effort, I have managed to extract this dataset and have imported it into a personal database for analysis and viewing.

However, Stick's rollouts are much more extensive, and (perhaps more importantly) are cubeful not cubeless.

Further, Tom derives his DMP, GG and GS estimates from the cubeless equities, whereas Stick offers actual rollouts (albeit only DMP at present) AtS -- clearly a much more precise approach.

So I'm torn. Tom's data is more 'complete' but potentially less accurate. However, from a data processing perspective, it is substantially easier to manipulate in bulk.

I had assumed (erroneously) that the differences between Tom and Stick's outcomes would be minor. However upon comparison, the variances seem quite large. For example, in the case of 21$ 11, Tom's GNU money rollout suggests the Hit (just ahead of the Near play), while Stick's Snowie analysis suggests the Hit is a 0.03 mistake. In this regard, the recent 21$ 11 analysis conducted by Nack Ballard and Neil Kazaross (here) is also relevant.

The most accurate rollouts seem to be coming from XG these days. Will this new data render the existing information obsolete?

Has anyone conducted a careful comparison of Tom Keith's and Stick's data? If so, where are the largest anomalies? As I've already taken the trouble of importing Tom's data, should I use it as a base and manually update the numbers with the more up-to-date information posted on this site (and indeed in this forum)? Or should I start from scratch with Stick's data (leaving out GG and GS)? In cases where Snowie, GNU and XG differ, whose opinion should I favour?

Does it make sense to mathematically infer DMP, GG and GS figures from the base equities (as Tom Keith does), or is that approach too inaccurate?

Note that I'm not necessarily after absolute 'precision'. I'd be perfectly happy if I could fix upon a set of data that reasonably accurately suggests where each reply option lands on the Ballard-Weaver Error Scale. This allows some tolerance for variation between methodologies/bots etc. Doubtless there will always be debate amongst experts (and bots), but I'd like to give myself the best possible chance that the bulk of what I'm studying is unlikely to be radically overturned in the near future.

(Or perhaps I should just sit tight and await the other 20 volumes of Mr. Ballard and Mr. Weaver's opus :-)

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.