Redskins trade for Jason Taylor (updated)

[Schneed10]

Originally Posted by GTripp0012

Well, the exact win total is based on some work I've been doing to convert value to team into a win estimate.

For example, if we consider Jon Kitna to be the standard for a starting QB, with most teams having better options, (I just picked a guy who's got a readily available skill set, and doesn't do too much to hurt his team), we can estimate that Tom Brady was worth about 5 wins more than Kitna-level last year alone. Essentially saying that the Patriots, with the same schedule, same team, but Jon Kitna at QB, would have been an 11-5 team.

The value metrics are readily available at places like footballoutsiders.com, profootballreference.com's blog, KC Joyner's articles on ESPN.com, and Brian Burke's Advanced Football Stats blog.

There's always going to be slight disagreement among the systems about certain players value, but generally these four sources paint me a pretty good picture of just how valuable a player is.

Anyway, I'm most concerned with how valuable a rookies' first 4 years in the league are for a few reasons: 1) it's a good indicator of how good he will be, and 2) that's how long in which his draft position will determine how much he exceeds his value by.

For a second rounder, in his 2nd-4th years in the league, we know that he's likely to be better than a replacement level player, and a key contributor on this team. Everything I've done says that a second round pick who doesn't bust out of the league, but also fails to become a superstar will be worth not quite a win per season on his rookie deal.

So I'm estimating over a large sample when I say the average second rounder is worth 2-3 wins over his rookie deal. About a third of the second round selections will be less than that, and a third will be more than that. But that's the expectation for a second round selection, and expectation is really the only thing we are dealing with here.

I'd be interested to hear what kind of analysis you're doing to strip out the covariances involved with the multiple variables driving team performance. That's the big reason sabermetrics has not caught on in football. With baseball, you can normalize your data set rather easily because in the end, it largely boils down to a pitcher vs hitter matchup, with a few variables like day vs night, score, and situation to adjust for. But in football, the QB's performance depend's upon the line's ability to block the defense, the WRs' ability to get open, the effectiveness of the running game, the score, the quarter, etcetera. It's hard enough to quantify some of these variables, and even harder to mathematically formulate the covariance quotients to effectively tease them out and normalize your data. Have you done anything on this front?

[That Guy]

Originally Posted by Schneed10

I'd be interested to hear what kind of analysis you're doing to strip out the covariances involved with the multiple variables driving team performance. That's the big reason sabermetrics has not caught on in football. With baseball, you can normalize your data set rather easily because in the end, it largely boils down to a pitcher vs hitter matchup, with a few variables like day vs night, score, and situation to adjust for. But in football, the QB's performance depend's upon the line's ability to block the defense, the WRs' ability to get open, the effectiveness of the running game, the score, the quarter, etcetera. It's hard enough to quantify some of these variables, and even harder to mathematically formulate the covariance quotients to effectively tease them out and normalize your data. Have you done anything on this front?

yeah, i watch the game.




but i did make a neat equation if you're interested:
good players = win more games.

yay!

[jdlea]

Originally Posted by That Guy

good players = win more games.

That is some funny sh#t!

[Schneed10]

Originally Posted by That Guy

yeah, i watch the game.




but i did make a neat equation if you're interested:
good players = win more games.

yay!

This was basically the point of my post. Sabermetrics is relatively useless in football because nobody's been able to do any of the things I mentioned.

[That Guy]

Originally Posted by Schneed10

This was basically the point of my post. Sabermetrics is relatively useless in football because nobody's been able to do any of the things I mentioned.

i understand that, but when you go from saying there's a probability that brady would win you five more games to stating it as a fact that's annoying some i think.

those stats aren't proven, and while it is useful, they're not definitive, it's just probabilities vs average (and generally, it's average production, not average talent)... the actual game scorers that provide their basic data feed are NOT accounted for either among other things (ie, philly's scorers tend to count passes throw directly out of bounds as PDs for the nearest defender, but most scorers don't do that).

because most of the work doesn't factor out all that much (it's better than nothing, but on a scale of 1 to 10, it might be a 3 or 4) so quoting it as definitive or fact is really pretty misleading.

*the team stats are actually pretty decent, it's just the individual stats and finding specific answers based off vague input data that's not great.

[mredskins]

Originally Posted by Schneed10

I'd be interested to hear what kind of analysis you're doing to strip out the covariances involved with the multiple variables driving team performance. That's the big reason sabermetrics has not caught on in football. With baseball, you can normalize your data set rather easily because in the end, it largely boils down to a pitcher vs hitter matchup, with a few variables like day vs night, score, and situation to adjust for. But in football, the QB's performance depend's upon the line's ability to block the defense, the WRs' ability to get open, the effectiveness of the running game, the score, the quarter, etcetera. It's hard enough to quantify some of these variables, and even harder to mathematically formulate the covariance quotients to effectively tease them out and normalize your data. Have you done anything on this front?

This paragraph drips nerd. =)

[GTripp0012]

Originally Posted by Schneed10

I'd be interested to hear what kind of analysis you're doing to strip out the covariances involved with the multiple variables driving team performance. That's the big reason sabermetrics has not caught on in football. With baseball, you can normalize your data set rather easily because in the end, it largely boils down to a pitcher vs hitter matchup, with a few variables like day vs night, score, and situation to adjust for. But in football, the QB's performance depend's upon the line's ability to block the defense, the WRs' ability to get open, the effectiveness of the running game, the score, the quarter, etcetera. It's hard enough to quantify some of these variables, and even harder to mathematically formulate the covariance quotients to effectively tease them out and normalize your data. Have you done anything on this front?

The other big issue is sample size: one football game provides about two weeks of equivalent data in baseball.

To an extent, you have to leave the team oriented variables in there often. One of the biggest allies of statistical analysis in football is roster turnover: parts come and go, and production changes accordingly, so over multiple years, you can see which positions and players are more valuable, and exactly what the numbers are showing.

As far as the normalizing of the data, I'm not really working on the frontlines there. There are people out there who are much better than I am working with correlations and statistical significance, and that type of work (such as FO's DVOA metric or Doug Drinen's Approximate Value metric).

A lot of the work I do is simply taking the results that other systems churn out, and trying to work through team and coaching variables to better translate those numbers into wins and losses. Basically, I'm best when working with the statistical defenses of roster moves (or lack thereof).

Like I said earlier, my current project is trying to find out how many wins Jason Taylor adds over Phillip Daniels. Once I figure this out, I'll know just how long Jason Taylor has to play at a high level to be worth a second round draft pick.