tag:blogger.com,1999:blog-38600807.post3935993535108104336..comments2024-10-28T16:59:39.881-04:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Are Teams Going For It More In 2009?Unknownnoreply@blogger.comBlogger11125tag:blogger.com,1999:blog-38600807.post-36468152451094738202009-12-21T10:55:02.220-05:002009-12-21T10:55:02.220-05:00Jonathan, I'll take a shot:
p = .124 (expect p...Jonathan, I'll take a shot:<br />p = .124 (expect proportion of "going for it")<br />q = .876 (1-p)<br />p1 = .147 (observed proportion of "going for it")<br />n = 3224 (sample size)<br />z = (p1-p)/sqrt(pq/n) (test statistic)<br />z = (.147-.124)/sqrt(.124*.876/3224)<br />z = (.023)/(.005805) = 3.96<br />For a 1-tailed test, p(z=3.96) = .0000375. So, yes, 14.7% is very much a significant difference IF that sample size of 3224 is correct (seems high to me - I didn't try to verify it).<br /><br />This says nothing about cause and effect, however. The higher percentage of conversion attempts could be caused by more short yardage downs that might be caused by something else, as pointed out earlier.LamKramhttps://www.blogger.com/profile/11026152257890059515noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-4107561931027938882009-12-20T18:29:39.469-05:002009-12-20T18:29:39.469-05:00By my count, there have been a total of 3224 fourt...By my count, there have been a total of 3224 fourth down "decisions" this year. 474 of them resulted in the team going for it. <br /><br />If 12.4% is the true expected rate, then we would expect that number to be 400. So it's a difference of 74...I don't know if that is statistically significant though, as it's been awhile since my stats class. ;)Jonathannoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-4892511056127324992009-12-20T11:11:19.281-05:002009-12-20T11:11:19.281-05:00You guys are completely right. Significance is imp...You guys are completely right. Significance is important, and it's highly unlikely the intra-2009 post-Belichick rate is significant with the current number of cases. Further, I haven't accounted for to-go distances yet, which could be important. <br /><br />That said, my point was to only confirm or deny the appearance that teams are going for it more often, or was it just our imagination. The answer is yes, they are, and it has increased since the notorious 4th and 2. <br /><br />I did make the 'revolution' crack, which was just my own irrational exuberance showing through. I probably should have left that out...<br /><br />When the season's over, I'll take a more rigorous look at 4th down rates.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-85107115959731492832009-12-20T09:02:46.924-05:002009-12-20T09:02:46.924-05:00Anonymous - that's a great point. Fortunately ...Anonymous - that's a great point. Fortunately Brian has his Win Probability calculator so could, if he wanted, do some filtering to see if teams are going for it more often in non-desperation situations (for instance, a look at the % of Go For Its when the WP is 25-75%). It would be interesting if coaches are going for it more when the game is still close.Iannoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-80754758921140310702009-12-19T14:07:20.096-05:002009-12-19T14:07:20.096-05:00I would think you'd also have to adjust for th...I would think you'd also have to adjust for the score. 2009 has seen a lot of one-sided games, which would lead to more losing teams going for it out of desperation.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-25367472960674248862009-12-19T05:29:44.730-05:002009-12-19T05:29:44.730-05:00One point Brian. Although it may average out over ...One point Brian. Although it may average out over the data set, is it possible that random variations have meant teams have faced shorter than average 4th downs this year? This would result in a similar phenomenon to 3rd down conversion percentages. A team who faced 3rd and 1 every 3rd down would have a higher conversion rate than a team that faced 3rd and 5 for instance.<br /><br />Of course, shorter than average 4th downs could themselves be an indicator that coaches are more prepared to go for it. If you decide before a series that you will go for it on a 4th and 1 then your play calling will be slightly different (e.g. on 3rd and 5 you would be happy to call a draw because you could easily get to 4th and 1 with it - a coach who didn't go for it would try passing more and would face a longer 4th down in the event of a failed 3rd down)Iannoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-71762709928310929322009-12-19T00:05:19.609-05:002009-12-19T00:05:19.609-05:00That's what the test for significance is. Beca...That's what the test for significance is. Because there is random variation from year to year we need to know if this percentage from this year is actually significantly different than the previous years. If it's not a significant different (say, a p-value above .05, or even .10) than it's just the random fluctuation from year to year.Paulhttp://stats4you.blogspot.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-6416484283110742892009-12-18T18:19:32.045-05:002009-12-18T18:19:32.045-05:00Yeah, any chance we get some more colors (like 200...Yeah, any chance we get some more colors (like 2002, 2003, 2004, 2005, 2006, 2007) added to the first graph? I have a hard time believing that 2.5% isn't anything more than standard year-to-year fluctuation.Marverhttp://pigskintelligence.blogspot.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-29920377044596120132009-12-18T17:47:05.249-05:002009-12-18T17:47:05.249-05:004th down go-for-it rates tend to fluctuate slightl...4th down go-for-it rates tend to fluctuate slightly from one year to the next. For example, NFL teams went for it more often in 2007 than they did last season. There has been very little change since Romer's paper made headlines in 2002.Jim Anoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-67437161534465431672009-12-18T14:35:12.946-05:002009-12-18T14:35:12.946-05:00Adam, that was my first thought as well. Brian, ca...Adam, that was my first thought as well. Brian, can you confirm the data is significant? Can you supply the p-value for this?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-2116933281444801892009-12-18T11:16:55.454-05:002009-12-18T11:16:55.454-05:00I assume all the differences noted are significant...I assume all the differences noted are significantly significant at p=0.05 (or p=very small). Is this correct?Adam Cznoreply@blogger.com