tag:blogger.com,1999:blog-38600807.post2407199056631540939..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Running In Style: Vikings Game-Tying DriveUnknownnoreply@blogger.comBlogger3125tag:blogger.com,1999:blog-38600807.post-71672040314890193762011-11-01T13:34:32.788-04:002011-11-01T13:34:32.788-04:00I agree that smoothing the curve isn't meaning...I agree that smoothing the curve isn't meaningful or useful. I disagree that bars would be best, though - I think an unsmoothed line plot would be by far the easiest way to view these results.<br /><br />While I get the origin of the "absorbtion" term, it seems like unnecessary technical jargon that surely confuses some of your audience. "Drive outcome probability" is both a precise and accurate term in this case, and it is much more clear to the semi-layman, I think.<br /><br />It would be interesting to see some further research on your argument at the end of this piece, that they should have stuck with the run the entire drive. Of course, looking at the results of one specific drive isn't very meaningful, due to variance (and the fact that passing is inherently higher variance than running). But with enough data, the results could be interesting.<br /><br />Brian has done some research on optimal run/pass balance in the past. I'd love to see the Markov model applied to this. When play-calling is pushed more heavily towards the more efficient choice, to what extent does this hurt the efficiency of that choice?Tarrhttps://www.blogger.com/profile/14368810359650066790noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-48419965655887908352011-10-31T23:57:02.483-04:002011-10-31T23:57:02.483-04:00Interesting comment about using a more discrete pl...Interesting comment about using a more discrete plot, I'll definitely look at it. I like the smoothed curve because I think it does a better job highlighting the evolution and patterns of the drive.<br /><br />As for absorption probabilities, it is based on an absorbing Markov chain, and that is technically where the calculation comes from. It is the probability that the chain (or drive) is absorbed (ends in) that specific drive-ending state.Keith Goldnerhttps://www.blogger.com/profile/16510947295485321744noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-22885283413684048242011-10-31T20:49:27.237-04:002011-10-31T20:49:27.237-04:00Keith, why do you use a smoothed curve to plot the...Keith, why do you use a smoothed curve to plot these results? There is no meaning to play 3.7 or 4.2, and there should be no expectation of continuity between plays. Bars would probably be the best format, but if you insist on lines, at least let them indicate the discrete nature of the underlying process. <br /><br />And a nit..."absorption probability?" These are not generic state transitions, they are "drive results" or "drive outcomes" or "likelihood drive ends in..." or something that indicates that these numbers have a meaning.Willhttp://www.edwardtufte.com/tufte/books_vdqinoreply@blogger.com