Rearview QB article (1 Viewer)

[Chase Stuart]

http://footballguys.com/articles/07stuart_rearviewqb.php

This is probably my favorite article to write each year. Thought I'd open it up for discussion in the pool here.

 

[Sylira21]

I like getting the adjusted PPG stat. I knew McNabb was having a magical season, but not really just how good it was...

The SOS piece. I don't know, I tend to get a little dubious about this. How well are you able to predict something like QB SOS? And overall, it doesn't have a huge impact, according to this format. However, it does tell you one interesting thing, which you do note Chase - comparisons between QBs that might be fighting for a job. That is interesting to nore, in my mind.

Your other main point that leapt out at me was this:

"There’s another cluster of pretty good QBs, but very few real difference makers at the position. Players like Rivers, Favre, Brady and Roethlisberger ranked in the top ten in FPs, but none were really that impressive. They either played lots of games or had an easy schedule, or both."

This seems to be true to me, regardless of what you do to tweak scoring systems, or what nuggets of data you mine. As long as you play only 1 QB, you have to take this effect into account...which means it's hard to get too excited about paying top dollar for a QB.

 

[Chase Stuart]

How well are you able to predict something like QB SOS? And overall, it doesn't have a huge impact, according to this format.

There's no predicting involved. It involves adjusted the scoring for the strength of the defenses from last season. So a player gets equal credit for scoring the what an average QB would score against the Raiders as the player gets for scoring what an average QB would score against the Bengals. Adjusting for the strength of defenses is pretty important, IMO, when analyzing last year's results.

 

[Joe T]

Do strength of defenses from prior years have any correlation to the strength of the defense in the current year?

 

[Chase Stuart]

Do strength of defenses from prior years have any correlation to the strength of the defense in the current year?

I believe the answer is some, but not much. I'm not home right now so I can't run the numbers, but I'll try to do so tonight. I have the strength of defense ratings from the past four years, which will give us three correlations to check. It's important to note that the defenses are also adjusted for strength of schedule; I think that should help a bit in terms of improving the correlation.

 

[fishergm31]

Excellent article. I look forward to reading it every year. All of my leagues are 2QB leagues, so finding any QB advantages, no matter how small, can be huge.

I think it's important to note that not only are you adjusting the QB ranks for SOS, but also the Defense ranks. It's difficult in a "closed environment" to normalize stats. Did Manning have a good game because the Ravens defense is bad? Or did the Ravens defense have a bad game because Manning is so good?

I found it interesting that some of the best defenses (Bears, Ravens, etc) gave up more than the avg fantasy passing pts. Is that because teams were typically playing from behind against them and thus had to throw more?

I'd be curious to see a lookback of SOS predictions from last year. The teams that we thought would be easy to pass against, how did they actually fare? Were Chicago and Baltimore projected to be in the top 10 easiest teams to pass against? I doubt it.

 

[Chase Stuart]

Excellent article. I look forward to reading it every year. All of my leagues are 2QB leagues, so finding any QB advantages, no matter how small, can be huge.

I think it's important to note that not only are you adjusting the QB ranks for SOS, but also the Defense ranks. It's difficult in a "closed environment" to normalize stats. Did Manning have a good game because the Ravens defense is bad? Or did the Ravens defense have a bad game because Manning is so good?

I found it interesting that some of the best defenses (Bears, Ravens, etc) gave up more than the avg fantasy passing pts. Is that because teams were typically playing from behind against them and thus had to throw more?

I'd be curious to see a lookback of SOS predictions from last year. The teams that we thought would be easy to pass against, how did they actually fare? Were Chicago and Baltimore projected to be in the top 10 easiest teams to pass against? I doubt it.

Thanks! Glad you enjoy it.I think you're reading the defensive list upside down. Oakland was the toughest defense to face last year; Chicago and Baltimore were among the stingiest defenses.

Listed below are all the team defenses, ranked by their adjusted fantasy points allowed value, which is how many FPs per game less you would expect the league average QB to score against them than against a league average defense.

So a positive number means you'd expect the defense to allow less points than average. The Colts didn't allow many passing FPs because teams rarely passed against them; the Raiders were great for the same reason, and because they've got a talented defense.

 

[fishergm31]

You're right I was reading it backwards. I got so excited for the numbers that I didn't read the paragraph that closely... and I figured the Raiders were the worst at everything else, so it seemed logical that they were the easiest team to pass against, when in reality they were the toughest. Probably partially due to everyone running the ball against them so much.

So let me make sure I understand the QB part right. McNabb's FP/G adjusted for SOS would be 25.1 (26.0 - .90)? The league average being 16.45 (Rivers 16.0 + .46 SOS - .001 value). McNabb's 8.65 above avg + 16.45 league avg = 25.1 seems to confirm that. And this means that had McNabb started all of his games against the "average team" he would have scored 25.1 pts instead of 26.0.

Do you think 6 pts/TD would change the findings at all? I could see it moving some players around a little bit, but not drastically changing the results.

 

[BroadwayG]

This is the goods.

 

[Chase Stuart]

You're right I was reading it backwards. I got so excited for the numbers that I didn't read the paragraph that closely... and I figured the Raiders were the worst at everything else, so it seemed logical that they were the easiest team to pass against, when in reality they were the toughest. Probably partially due to everyone running the ball against them so much. So let me make sure I understand the QB part right. McNabb's FP/G adjusted for SOS would be 25.1 (26.0 - .90)? The league average being 16.45 (Rivers 16.0 + .46 SOS - .001 value). McNabb's 8.65 above avg + 16.45 league avg = 25.1 seems to confirm that. And this means that had McNabb started all of his games against the "average team" he would have scored 25.1 pts instead of 26.0. Do you think 6 pts/TD would change the findings at all? I could see it moving some players around a little bit, but not drastically changing the results.

Your analysis is correct. Whether or not the SOS modifier should be through addition/subtraction or multiplication/division is a debatable point, but I don't think there's a debate that there should be a modifier for SOS. But yes, the league average was 16.5. McNabb averaged 26.0, which was 9.5 points better than average. Because LeagueAverageQB would have scored 17.4 PPG if he had played McNabb's schedule, McNabb was really "only" 8.65 points better than league average.The people who come away badly when seeing this analysis are Eli Manning and Alex Smith. I'll post my thoughts on them in a second.

 

[Chase Stuart]

Going through the adjustments, you'll see why I think Eli Manning and Alex Smith may be overrated based on last year's stats.

1) Eli Manning ranked 12th last year, Alex Smith ranked 18th.

2) BUT, Eli Manning and Alex Smith both played 16 games. Manning ranked 15th in FP/G (min: 8 games) -- McNabb, Hasselbeck and Delhomme jumped him (and Romo would if he hadn't been the holder in each game). Alex Smith ranked 21st in FP/G (min: 9 games, to not count Garcia and McNabb): Leinart, Delhomme and Hasselbeck passed him.

3) BUT, Eli Manning and Alex Smith both took every snap for their teams. They played 16.0 adjusted games. Manning ranked 20th in FP/adjG (min: 8 games), with Leinart, Grossman, McNair, Huard and Romo all passing him. Smith ranked 26th in FP/adjG (min: 8 games), with Huard, Romo, Harrington, Frye and Garrard passing him.

4) BUT, Manning and Smith both played slightly easier than average schedules. Manning drops to 24th in adjFP/adjG (min: 8 games), with Harrington, Pennington, Losman, Garrard and Frye passing him (and he passes Delhomme, whose schedule was even easier). Smith drops to 28th, with Carr and Plummer passing him.

So Manning goes from 12th to 24th and Alex Smith from 18th to 28th. Those latter rankings show how good each QB really was last year. Their overall all rankings could easily induce someone into thinking they're better than they really were.

 

[Chase Stuart]

Do strength of defenses from prior years have any correlation to the strength of the defense in the current year?

Correlation Coefficient of the raw defensive strength ratings from Year X to the raw defensive strength ratings in Year X+1:
03-04 0.26304-05 0.29905-06 0.102Avg. 0.222Correlation Coefficient of the adjusted defensive strength ratings from Year X to the raw defensive strength ratings in Year X+1:

03-04 	0.27104-05 	0.39105-06 	0.099Avg.	  0.254

The 2004-2005 CC increased quite a bit when adjusting for strength of schedule in 2004. This suggests that there were several teams in 2004 that played extremely tough (or easy) schedules, that skewed the results enough that you'd make an inaccurate prediction the following year based on those results. This is exactly what happened.Twenty teams over the four years had strength of schedule ratings of more extreme than +/- 1.00. What that basically means is twenty teams played a whole lot of great or terrible QBs due to random chance. Fourteen of those tough (or easy) seasons occurred in 2004. The '04 Bengals are the best example, with a SOS of -1.90, meaning they had an incredibly easy schedule.

4.49 Billy Volek 1.90 Jake Plummer 1.41 Tom Brady-0.35 Chad Pennington-0.52 Jeff Garcia-0.96 Ben Roethlisberger-0.96 Ben Roethlisberger-1.99 Vinny Testaverde-2.33 A.J. Feeley-2.88 Drew Bledsoe-4.44 Patrick Ramsey (79%)-4.96 Kyle Boller-4.96 Kyle Boller-5.74 Mark Brunell (21%)-6.12 Eli Manning-22.93 TotalThat, of course, is an incredibly easy schedule -- and it doesn't even show two full games out of Kelly Holcomb, Koy Detmer and Jeff Blake, which drops the schedule's strength to -30.34. In other words, the '04 Bengals faced QBs on average that scored 1.90 fewer points per game than league average (which is how we got the -1.90 number originally). The '04 Bengals ranked 13th in YPA allowed, but as we've seen, it's because Cincinnati faced a ton of terrible QBs. In 2005, Cincinnati faced an easier than average schedule (-0.79) but it was still a lot harder than the previous year; Cincinnati subsequently dropped to 28th in YPA allowed in '05.Correlation Coefficient of the adjusted defensive strength ratings from Year X to the adjusted defensive strength ratings in Year X+1:

03-04 0.23704-05 0.35505-06 0.135Avg. 0.243 I'm not so sure why these numbers aren't stronger than the adjusted to raw numbers. I suspect they should be, so I'd probably side towards the theory that the sample size here is too small to show the true results.

Overall, I'd guess that the correlation coefficient is somewhere between 0.25 and 0.30 from year to year. This roughly means that between six and nine percent of the Year X+1 results can be "explained" by the Year X results.

 

[Q-Bert]

So by looking back on last years article that also included a ranking of the 2004 season we find that Donovan McNabb has finished 3, 2, 1 in Value.

Starting McNabb when he is healthy and plugging in whoever his backup is when he is out would have resulted in the following overall finishes for the last three years.

2004 - 3

2005 - 2

2006 - 2

That shocked me.

 

[Chase Stuart]

So by looking back on last years article that also included a ranking of the 2004 season we find that Donovan McNabb has finished 3, 2, 1 in Value.Starting McNabb when he is healthy and plugging in whoever his backup is when he is out would have resulted in the following overall finishes for the last three years.2004 - 32005 - 22006 - 2That shocked me.

The Phi TM QB ranked 3rd, 7th and 1st in 2004, 2005 and 2006.