tag:blogger.com,1999:blog-38600807.post4875963964175880628..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): In-Game WP Model ExplainedUnknownnoreply@blogger.comBlogger7125tag:blogger.com,1999:blog-38600807.post-39234107205260387732008-11-15T15:13:00.000-05:002008-11-15T15:13:00.000-05:00This is a great site; i love your analysis. i can...This is a great site; i love your analysis. i cant watch an NFL game without looking up the WP or efficiency rankings. I had a question about in game Win Probabilities.<BR/>I read this somewhere on your site but cant remember where i found it.<BR/>Somebody asked why you dont use the game probabilities that your efficiency model calculates at the begining of each game instead of 50/50; your response was something like "where do you merge the pre-game predictions, which may or may not be accurate, with the real game?" i completely agree.. with one second left the 12-0 team doesn't have a 95% chance of winning if they are down 14. I was wondering if you could take the two probabilites side by side (in game and eff. model) at the start of the game. multiply the efficiency model prediction by 1; add that quantity to the in game prediction (which is .50 right now) times 0. creat a linear relationship so that at the end of the game you multiply the in-game prediction by 1; add that to the efficiency model times 0. what im trying to say mathmatically is:<BR/>WP_displayed = Eff_prediction * ((game time - time elapsed)/game time) + In_game_prediction * (time elapsed / game time)<BR/>You dont actually have to do it; i just wondered if that made sense.<BR/>Thanks - AndyAndyhttps://www.blogger.com/profile/13033784825091918451noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-7003366124900410492008-11-11T10:23:00.000-05:002008-11-11T10:23:00.000-05:00Have you noticed any differences of significance g...Have you noticed any differences of significance going back to 2000 between the HTML play-by-play and the PDF gamebooks on the NFL site?<BR/><BR/>How about using the play-by-play that is available via internet archive sites? I do believe 1999 and 1998 nfl gamebooks are complete as far as play-by-play goes. That could increase the database.<BR/><BR/>Also, has anyone found the HTML version of the gamebooks that have been available through the 2007 season on sportsline.com?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-77886538416551425182008-11-10T22:57:00.000-05:002008-11-10T22:57:00.000-05:00Buzz-that's a great research question. How does a ...Buzz-that's a great research question. How does a single sack kill a drive? I'll be sure to do an article on that. If I forget to credit you, kindly poke me!Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-13400331481460538982008-11-10T22:26:00.000-05:002008-11-10T22:26:00.000-05:00I really appreciate all your efforts for this proj...I really appreciate all your efforts for this project as well. It has been pretty cool watching it develop this year. Also I just love all of the other analysis that you have done. I have been starting to work on some projects as well, nothing close to what you are doing. I know I would certainly be interested in even paying for some type of data tool like you are talking about. Here's one question that I have in regards to the play by play data that you might have. What percent of drives stall after a qb sack as compared to the average (I believe you said about 66% of 1st and 10 are 1st downs). It sure seems to me that drives with sacks are killer. Keep up the great work.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-76534585829951502932008-11-08T15:35:00.000-05:002008-11-08T15:35:00.000-05:00Cool! I wonder how Retrosheet got the data rights...Cool! I wonder how Retrosheet got the data rights for Baseball. Have you had any discussions with http://footballproject.com?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-4608774818139833462008-11-08T12:05:00.000-05:002008-11-08T12:05:00.000-05:00Hi, Miles. Yes, that's a major goal of mine. I'd l...Hi, Miles. Yes, that's a major goal of mine. I'd love to spread the wealth, and I really don't care for other sites with "proprietary" bunk. But there's a problem. Team and player stats are considered public domain, but putting a giant play-by-play database into the wild isn't so clear cut. The NFL vigilantly enforces its property rights to the games themselves. <BR/><BR/>So I can't give out the raw play-by-play data directly. It is however available on a game-by game basis at the NFL site itself, or at others such as Yahoo. It's just a pain to manually collate hundreds of games into something useful.<BR/><BR/>Here's my plan. I'm building a play-by-play parsing tool that can convert html play-by-play tables into a usable flat database file. That way I'm not giving anyone any data, just a tool to better interpret the data already freely offered by the NFL. I'll probably have something available this off-season.<BR/><BR/>My other idea is to create an online query tool that will predigest the data. So you wouldn't get raw play-by-play data, but you could find out the average gain on all 3rd downs by field position, or by time remaining, or whatever. It would be useful for research purposes while still not giving away "accounts of the game."Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-57540412854765685172008-11-08T11:23:00.000-05:002008-11-08T11:23:00.000-05:00Have you thought about releasing your parsing code...Have you thought about releasing your parsing code or your database (perhaps it could be exported to a sqlite db). Having the play by play data available might help spur some more NFL research (or allow independent verification of other work).Anonymousnoreply@blogger.com