Any math whizzes on here?

[Swedish Extrovert]

I've always thought that if someone could formulate a statistical model for winning based on numbers - everything from using combine results (for bringing in players) to play calling...

Could this be the start of a "statistical" NFL model?

http://blogs.denverpost.com/broncos/...ce=rsshomeblog

[Swedish Extrovert]

Another:

http://harvardsportsanalysis.wordpre...-of-tim-tebow/

[Bacchus]

Originally Posted by Mightysmurf

I've always thought that if someone could formulate a statistical model for winning based on numbers - everything from using combine results (for bringing in players) to play calling...

Could this be the start of a "statistical" NFL model?

http://blogs.denverpost.com/broncos/...ce=rsshomeblog

You can't base it on numbers because sports are more than numbers. You can not put a number to work ethic, desire or just being a good teammate. Jeff George had great numbers I'm sure.

[Swedish Extrovert]

Tebow has a much worse EPA – meaning that he has objectively produced worse results over all of his plays – but a significantly better (though still negative) WPA – meaning that his plays have contributed more to the chances of his team winning. While Tebow’s plays have produced much fewer expected points than Orton’s, Tebow’s positive plays have come at crucial points in the game, when they have a much larger impact on the outcome: think his 20 yard TD run for the lead with 58 seconds left last week, or his 56 yard TD pass for the lead with 6:44 left the week before. Likewise, Tebow’s negative plays have come at points where their effect was less harmful.

EPA - Expected Points Added
WPA - Win Probability Added

[cutthemdown]

Tebow + Ball / Juice = winning

[jet19]

I am kind of a math whizz, I took a look at that info and I think I have the formula you are looking for

Tebow+Miller+Elway=Win

[Abqbronco]

Originally Posted by jet19

I am kind of a math whizz, I took a look at that info and I think I have the formula you are looking for

(Tebow+Miller+Elway)-orton=Win

Math error corrected. Analysis complete

[Swedish Extrovert]

I think the board is more in line with:

[(Von + Doom + Bunkley) x (Champ + Bailey / Goodman)] + {[(Tebow/McCoy) x McGahee] + (-Orton)} = win


The front seven compounds coverage (they play off eachother), Champ and Dawk are divisible by the weakness = Goodman.

Added to...

Tebow, divisible by McCoy's play calling (not my opinion, the forum's ;P). McGahee is independent from Tebow's parenthetical because he can just pound it but Tebow's inability to open up limits McGahee's potential, and adding a negative Orton (addition by subtraction).

About right?

[gyldenlove]

Ultimately there are far too many variables and far too few games to make a model with any kind of decent behaviour. Simply put the number of input variables is much greater than the number of possible data points (games) which means you can not find anything that even resembles a unique solutions (there will be infinitely many solutions) unless you introduce arbitrary ad hoc contraints the limit the solution space.

In this analysis the choice of input variables is highly questionably as every metric used is highly subjective. EPA for instance fails to take into account the situation you are playing in, if you are facing a very strong defensive team in really bad weather with a lot of injuries on your offense you are not nearly as likely to score points as you are if you are facing a very bad defense with a lot of injuries in a dome. EPA since it is based on historical averages is a metric that entirely negates these much more important factors. WPA is like EPA in that it is based entirely around historical data which washes out trends, some QBs and teams are very good in the 4th quarter (Elway was) and some are very bad (Orton), this means that if Elway had the ball at midfield with 1.11 to go down by 4, he was all else being equal, substantially more likely to pull out the win than Orton would be in the same situation - WPA because it is based on averages fails to take this into account and so Elway would have a surprisingly high WPA when studying his results compared to Orton - this is the same with Tebow, he in many games plays much better in the 4th quarter than he does in the first.

Using DVOA as a metric of how the defense has played is absurd at best, the DVOA metric is highly subjective and includes evaluation of each play and as such has as much to do with whoever is reviewing the game and possibly what that person had to eat as it does football.

Returning to the actual analysis, what it shows is that Tebow is good when it matters the most and not very good when it has the least impact. The author concludes that this is an aberration and therefore not likely to last, however I would say it is a result of the metric failing to account for players who consistently can elevate their game in crunch time and has no predictive value at all. I am sure if you compare different QBs over their careers to the model you will see outliers who are no more extreme than the Tebow case.

But given that Tom Brady and Aaron Rodgers haven’t been able to do that over the long term, I wouldn’t bet the farm on it.

That is not only a poorly written sentence it is also wrong. What has happened is not that Tom Brady and Aaron Rodgers are winning less than they used to, they are scoring more than they used to, so rather than winning more than their production would warrant given his model they are now producing enough for his model to justify their very high win percentages. The Brady and Rodgers examples show that is not only possibly but in fact not particularly unlikely for a young QB to win despite not producing as much as expected.

[Bacchus]

What would be the numerical rating be for being blessed by Jesus?

[Swedish Extrovert]

Originally Posted by Wobet Selur

What would be the numerical rating be for being blessed by Jesus?

Tebow ∞

[crawdad]

Originally Posted by Wobet Selur

What would be the numerical rating be for being blessed by Jesus?

Your name should be selur wobet to be honest. But the Broncos Rule in my house!

[That One Guy]

I think the current equation is:

{avg[(Miller's pass rush)+(Doom's pass rush)]-avg[(RT pass block ability)+(LT pass block ability)]}*(opp QB resilience value)/(opp ability to stay awake*[Tebow+McGahee rush success]*(Tebow luck factor)

[KipCorrington25]

If you need math whizzes try Chiefs Planet... oh wait no, that's meth whizzes on Chiefs Planet I always get that confused.

[Swedish Extrovert]

Originally Posted by KipCorrington25

If you need math whizzes try Chiefs Planet... oh wait no, that's meth whizzes on Chiefs Planet I always get that confused.

There are some smart dudes on this forum.

[Rohirrim]

Originally Posted by KipCorrington25

If you need math whizzes try Chiefs Planet... oh wait no, that's meth whizzes on Chiefs Planet I always get that confused.

[Vegas_Bronco]

Our company does this daily on other market oriented studies and spends millions each year backtesting coefficients...trust me when I say...there will always be outliers anomalies to the model that really make the model more of a probability and speculation tool than an firm investment by which you can guarantee a certain result. I can tell you who is most likely to spend and on what and when but whether that actually happens or not is speculation.

[Broncoman13]

Originally Posted by cutthemdown

Tebow + Ball / Juice = winning

I thought juice went to the 10th power

[Willynowei]

Originally Posted by Mightysmurf

I've always thought that if someone could formulate a statistical model for winning based on numbers - everything from using combine results (for bringing in players) to play calling...

Could this be the start of a "statistical" NFL model?

http://blogs.denverpost.com/broncos/...ce=rsshomeblog

There are extremely well paid geniuses who spend their entire life's work trying to predict things with significantly more historical data than the performance of Tim Tebow (and all of the advanced NFL stats they used there); and who have managed to, despite all that, repeatedly **** it all up.

Then you have two guys from a school with a nice name draw a cute graph after guessing at the effect of a few variables on the game of football - a sport full of Chaos - of motorcycle injuries, I-hop related injuries, gun shootings, idiotic actions leading to suspensions, egotistical coaches pulling players to show power, teams running offenses and plays no one has seen before, and a league refereeing committee that comes up new **** to call/ignore every day.

I don't have a Phd in economics from MIT, but I won't need one to tell you that the study those two guys are doing is likely no better at predicting Tim Tebow's future performance than you or me after a dozen shots of Jameson.

EDIT: I feel like i said a whole lot of nothing related to the Topics main question, which seems to be whether the study hints that this style of modeling might be valid for future use etc.,

Answer is no, study looks pretty meaningless to me. All they did was draw a relationship between how many points a QB is expected to add (how they got that i have no idea) vs. the chances of winning the QB adds to a team(how they got that, who knows). Thats two pretty ****ty and arbitrary numbers considering the amount of other factors in a football game, how the heck do you know whether it was the QB that day or the rain? or the offensive coordinator, or receiver/corner matchup? the answer is you can't control for these things and theres likely not enough info on the NFL to get these numbers in the first place.

Also, the R-squared (correlation/relationship they did find) is 82% and thats kind of ****ty, I could probably find stronger correlation between microwave use and erectile disfunction.

Although, i'd be shocked if far better, more complex, and more accurate models didn't exist in the hands of Vegas odds makers.

[Requiem]

Math sucks.

[Vegas_Bronco]

Originally Posted by Requiem

Math sucks.

Says the twin brother left on earth that isn't traveling at light speed.

[That One Guy]

Originally Posted by Willynowei

Although, i'd be shocked if far better, more complex, and more accurate models didn't exist in the hands of Vegas odds makers.

Exactly what I thought, too. If the game of football could be predicted, Vegas would have the equation.

[theAPAOps5]

The problem with using stats to project winners from the draft is that this is a team sport based on 11 players who play as a unit and are dependent on each other. Nearly impossible

[broncosteven]

Originally Posted by Mightysmurf

I've always thought that if someone could formulate a statistical model for winning based on numbers - everything from using combine results (for bringing in players) to play calling...

Could this be the start of a "statistical" NFL model?

http://blogs.denverpost.com/broncos/...ce=rsshomeblog

I just read where SpaceX did a model on their rocket engine (The Merlin) and they found that by feeding the fuel in a stream through a needle like injector rather than the traditional "shower head" injector plate format they got better performance and less chance for combustion instability (this is the one issue that troubled Von Braun's team and lead to the shower head config) than current liquid fueled rockets.

Math and Science stuff is fun.

[fdf]

Originally Posted by Mightysmurf

I've always thought that if someone could formulate a statistical model for winning based on numbers - everything from using combine results (for bringing in players) to play calling...

Could this be the start of a "statistical" NFL model?

http://blogs.denverpost.com/broncos/...ce=rsshomeblog

I worked those models about 20 years ago. You can beat the spread reliably. But by such a small amount that you would have to have a lot of money and patience to have a reliable winning strategy over the long run. The reason, you would have to be able to stand up to several losing weekends in a row.