tag:blogger.com,1999:blog-38600807.post345834828854111784..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Comparing The 3-4 And 4-3 Defenses Part 2Unknownnoreply@blogger.comBlogger22125tag:blogger.com,1999:blog-38600807.post-31493647377610467322010-10-16T10:07:30.161-04:002010-10-16T10:07:30.161-04:00Hey Brian, had a few comments:
- How are you coun...Hey Brian, had a few comments:<br /><br />- How are you counting so-called hybrid defenses? What about the teams that had a transitional year where they played 3-4 only part-time while staying mostly 4-3?<br /><br />- It seems fairly commonly believed around the league that a big advantage of the 3-4 earlier in the decade was a Moneyball-type cost of talent one. Attributes and body types needed/not needed for the 3-4 were not in high demand and competed for, so players that could be effective in a 3-4 but not so much in a 4-3 were comparatively plentiful and cheap. With the fast-rising popularity of the 3-4, this advantage is now disappearing (or gone). I.e. perhaps it wasn't _better_ personnel but rather leveraging under-utilized personnel that explains part of the discrepancy?<br /><br />- Furthermore, the AFC turned to the 3-4 much earlier and in greater numbers, thus familiarity (i.e. preparation) was high in the AFC for years before the NFC. Can the 3-4 advantage be fully accounted for simply by inter-conference games? How different is the advantage if you limit analysis to only intra-conference AFC games?<br /><br />- I wonder what a similar analysis of the shift from majority 3-4 to ubiquitous 4-3 in the 80s would show... I realize you don't have play-by-play for games back then, but it makes me curious.<br /><br />- Lastly, I think it makes sense to include spikes and maybe aborted plays. A defense may be willing to give up a bigger play that keeps the clock running. Likewise, a defense that gives up the same play but lets an offense get out of bounds should be punished. Perhaps it makes sense to combine the previous play and the spike as a single "play" for EPA purposes. As for aborted plays, at least some of them are caused by the defense (if indirectly via confusion or rattling). Probably too infrequent to affect the final results, but worth noting anyway.Michael Lnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-43759287139638017562010-09-28T01:25:31.383-04:002010-09-28T01:25:31.383-04:00Actually, I'm not a hater. I love your resear...Actually, I'm not a hater. I love your research, but I think it would be more eagerly accepted if it didn't come with such a pretentious attitude and unwillingness to accept criticism. And I know others associated with the league feel the same way, FYI.<br /><br />FootballCommentary published the full source code for its WP model. And SportsQuant and GridIronMine have also had WP calculators on their sites in the past. GIM also lists WP changes on a play-by-play basis. Not necessarily saying their implementations are more accurate, but then again, how would we know?<br /><br />Clearly you have the best presentation (although the GIM charts were nice too) and apparently the most time to devote to spreading the word and keeping things up-to-date. I just think you need to be more careful when nearly every research article is now based on EP or WP, which are very powerful concepts, but very difficult to implement correctly. If you can't show the accuracy of the base estimates, I think it reduces the effectiveness of all the other studies you do.<br /><br />BTW, your description of your WP methodology matches almost exactly how ProTrade described theirs a couple years ago ("regression with Markov chains"). Yet there are game situations with very different estimations. So again, how does one determine which number is better? Without looking into the black box, we'll never know.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-69331271965050373272010-09-24T22:21:02.349-04:002010-09-24T22:21:02.349-04:00Why do people with personality disorders always en...Why do people with personality disorders always end up at my site? <i>"Oh wait, I forgot, it's all done by hand."</i> Who's the 7-yr old, loser? You're a hater with an agenda, and it's transparent. <br /><br />Show me another WP model that's more open. <br /><br />-Does it provide its full methodology, like ours?<br /><br />-Does it provide a calculator for determining the WP at any game state, like ours?<br /><br />-Does it provide graphs and play-by-play of a decade of games? <br /><br />-Does it provide free and full data to anyone who would like to replicate it, or even build their own?<br /><br />I thought so.<br /><br />For those without severe personality disorders who are interested, the WP model is not a black box. It is primarily based on several dozen LOESS regressions. There is no magic formula or code. LOESS regressions produce tables of data. The WP uses a combination of multi-dimensional look-up tables and Markov models to interpolate game states to estimate WP.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-48074374011954433192010-09-24T20:55:09.892-04:002010-09-24T20:55:09.892-04:00Just pointing out that your WP implementation is a...Just pointing out that your WP implementation is a proprietary black box. Fair enough, but it would be hypocritical to imply otherwise.<br /><br />There have been many other football WP models published (with varying degrees of openness), yet you come across as having invented the concept, unwilling to cite or even acknowledge your predecessors in this area of research.<br /><br />And the name-calling exposes you as having the maturity of a seven year old.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-62168278524974500742010-09-23T22:47:03.500-04:002010-09-23T22:47:03.500-04:00Hey, jerk. That's not how it works. Replicatin...Hey, jerk. That's not how it works. Replicating research doesn't mean you get spoon fed all the analysis and then you press print. <br /><br />WP is a regressed average of how often teams have won given particular game states from all actual games from 2000-2009. You want my source code? Right! Build your own model. I've given you all the data you need, for free.<br /><br />Nice try. Buzz off.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-85777731654552819422010-09-23T22:38:54.679-04:002010-09-23T22:38:54.679-04:00Hmm, I can't seem to find the defensive format...Hmm, I can't seem to find the defensive formation listed in your play-by-play data.<br /><br />Nor can I find the source code for WPA. Oh wait, I forgot, it's all done by hand.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-17852018503995548962010-09-23T21:11:41.232-04:002010-09-23T21:11:41.232-04:00Yes it can.
By the way, here you go. Get crackin...Yes it can. <br /><br />By the way, <a href="http://www.advancednflstats.com/2010/04/play-by-play-data.html" rel="nofollow">here you go</a>. Get crackin'! Let me know your results. You can publish them <a href="http://community.advancednflstats.com/" rel="nofollow">here</a>.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-71908746163179845802010-09-23T18:14:54.202-04:002010-09-23T18:14:54.202-04:00This can't be taken seriously unless you share...This can't be taken seriously unless you share the data so that this study can be replicated.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-51252418508310373442010-09-23T12:08:48.996-04:002010-09-23T12:08:48.996-04:00But what you aren't taking into account is how...But what you aren't taking into account is how changes in the front dictate changes in the strategy of the Defense -- specifically the differences between teams that one gap and teams that two-gap.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-78923423600912190962010-09-22T17:53:21.728-04:002010-09-22T17:53:21.728-04:00Most "3-4" teams run 4-3 fronts in varyi...Most "3-4" teams run 4-3 fronts in varying ways. They just do it with 3 lineman and 4 LBs.<br /><br />There's "3-4" the front, and "3-4" the baseline personnel package.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-28127519475009419612010-09-22T17:41:43.326-04:002010-09-22T17:41:43.326-04:00Is this a play-by-play or a game-by-game analysis?...Is this a play-by-play or a game-by-game analysis? I think that's an important question because many teams that use the 3-4 as a base defense use a 4-3 in some situations. In fact, I believe the Patriots last season actually used the 4-3 more than 3-4, though I may be wrong about that.Ralph Hickokhttp://www.hickoksports.com/noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-92106564152720251882010-09-22T00:59:07.226-04:002010-09-22T00:59:07.226-04:00You should post all the statistics of your regress...You should post all the statistics of your regression. For example, you can't tell if the difference in the pass and run coefficients are significantly different. My guess is no, they aren't significantly different. The only thing I can tell because of different signficance levels is that the standard error for the pass coefficient is probably larger than the standard error for the run coefficient. <br /><br />I'd like to see the r-squared, too.<br /><br />Finally, I thought that another argument for the 3-4 was that down linemen were hard to find: getting good down linemen is harder than getting good linebackers.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-72459332051733080442010-09-21T22:06:35.674-04:002010-09-21T22:06:35.674-04:00"Each individual team was also assigned a dum..."Each individual team was also assigned a dummy variable intended to capture the team-specific strength of their defense."<br /><br />Can you further explain what these dummy vars were and how they were derived?skyjonoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-16123835342450749822010-09-21T17:03:20.695-04:002010-09-21T17:03:20.695-04:00If it's true that the main thing NFL defenses ...If it's true that the main thing NFL defenses do very well, and differently than college defenses) is disguise coverages, 3-4 defenses give teams more options to overload one side of the offensive formation by only adding one blitzer. 4-3 defenses I think accomplish the same feat by employing tactics that take longer to develop (e.g. lineman stunts). The Tampa-2 and its' progeny are somewhat unique here as they rely on getting single blocking for one dominant lineman, and very rarely blitz linebackers or DBs.<br /><br />I also have the impression that 3-4 defenses ask less of each position than 4-3 defenses, perhaps it is easier to identify talent and to teach.Unknownhttps://www.blogger.com/profile/11189860457582272745noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-46908635384843328562010-09-21T16:02:36.693-04:002010-09-21T16:02:36.693-04:00Wow - this was unexpected. I certainly thought tha...Wow - this was unexpected. I certainly thought that 3-4 defenses were better because the teams running 3-4 were "defense first" kind of teams. (Pit Bal NE)<br /><br />I do wonder what the explanation is - having one more guy standing up is better - why?JURInoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-54243636986959529562010-09-21T16:01:47.359-04:002010-09-21T16:01:47.359-04:00Are 3-4 defenses more "expensive" than 4...Are 3-4 defenses more "expensive" than 4-3 defenses? My perception is that they are and therefore teams who use the 3-4 have to give up something with their the offensive unit. Could you compare the offenses of this same data set to see if the teams with the 4-3 defenses have better offenses? If so, does the offensive improvement exceed the EPA they are giving up by using the 4-3?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-39845576093814974642010-09-21T15:27:38.254-04:002010-09-21T15:27:38.254-04:00I know this isn't really related to this post ...I know this isn't really related to this post but are you going to do an article on the Saint's decision to kick the field goal that gave them an 8 point lead rather then go for it on 4th down? Thanks and great work as always.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-64908421728141331182010-09-21T15:20:48.574-04:002010-09-21T15:20:48.574-04:00One note about pass vs. run effectiveness from a S...One note about pass vs. run effectiveness from a Steelers fan: effectively handling the run game from "base" packages (which seems to be something 3-4 defenses are adept at) allows a team to be better against the pass.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-77011067080907862282010-09-21T11:21:42.366-04:002010-09-21T11:21:42.366-04:00James-I'll have a full analysis on that shortl...James-I'll have a full analysis on that shortly. It's probably going to end up at the Wash Post.<br /><br />The bottom line is that the FG from the 34 is generally better than the punt.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-15264759254575852162010-09-21T11:15:09.914-04:002010-09-21T11:15:09.914-04:00Okay Brian, I'm trying to work out the Texans ...Okay Brian, I'm trying to work out the Texans OT punt/FG/go-for-it probabilities but having trouble with the WP calculator. Is there some way you could add an option or list the different WPs depending on whether the offense kicks, punts or attempts to convert?Jameshttps://www.blogger.com/profile/01838293735141324662noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-52776467865742852942010-09-21T09:51:40.073-04:002010-09-21T09:51:40.073-04:00That's a good question. It's hard to answe...That's a good question. It's hard to answer definitively because so few teams ran the 3-4 in the early 2000s. Your point is important. As offenses become more familiar with the 3-4, they should improve, and the advantage disappears.<br /><br />We're probably approaching what game theorists call an Evolutionary Stable Strategy (ESS) mix. Suppose the 3-4 really is better. Most teams will become 3-4s, and the minority will be 4-3. Now, offenses would be less familiar with 4-3s. At some point, the overall average effectiveness of each strategy would equalize at an equilibrium.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-83074523689981704772010-09-21T08:19:43.976-04:002010-09-21T08:19:43.976-04:00Can you tell if this advantage has widened, narrow...Can you tell if this advantage has widened, narrowed or remained the same over the last ten years? Is it possible that the 3-4's advantage is that teams see it less often and as more teams switch to that alignment its advantage will disappear? I just stumbled across this site, so I apologize if you have addressed this elsewhere.Anonymousnoreply@blogger.com