Carson Palmer, Tom Brady and Drew Bledsoe. Along with Eli Manning, those quarterbacks were the biggest steals at the position in 2005. Those three QBs have something else in common: they had the best three scores in the system I'm about to describe. (But don't believe me, go to the end of this article to see last year's scores.) I hope that this article can once again reveal some great QB sleepers for those fantasy players who like to wait until the middle rounds of their drafts before picking a quarterback.

Here is how a rearview strength of schedule analysis works. Start with the raw fantasy totals for each QB in 2005 and adjust them for two things: actual games played and strength of schedule. Strength of schedule is a topic of controversy among fantasy experts, but there's no disagreement that a rearview strength of schedule analysis is particularly revealing. By normalizing each QB's statistics based on their opposing schedule, and adjusting for the correct amount of games played, we can arrive at which QBs were actually difference makers on the field last year. Ideally, those quarterbacks are the best fantasy players, a trait that we'd hope is relatively constant from year to year.

I wrote a similar article on this last year, but thanks to Doug Drinen I've implemented a new and improved model for 2006. Let's start with a basic chart. This displays fantasy point totals for each QB along with how many adjusted games they played, and their average fantasy points per adjusted game. (The normal Footballguys.com scoring system is in play here: 1 point for 20 yards passing, 4 points per passing TD, -1 for passing INTs, 1 point for 10 yards rushing and 6 points per rushing TD.)

A quarterback's adjusted games total was calculated by dividing each QB's pass attempts in every individual game by his team's total pass attempts by quarterbacks in that game. For example, in week 15 when Kurt Warner threw 10 passes and John Navarre threw 24, Warner was credited with 0.29 adjusted games and Navarro with 0.71 games. Below is the full table for all QBs with at least six adjusted games in 2005.

That list doesn't differ too much from your standard FP/G chart, although a QB like Bulger scores better when you divide his total FPs by 6.8 adjusted games instead of eight games. Intuitively, you might want to penalize Bulger for being injured last year and think he should be downgraded in your rankings because of a perceived risk. But what we're trying to get at here is how good each player actually was when he was on the field. To do that, adjusted games is the way to go.

Now for the second part of the analysis - adjusting for strength of schedule. Playing the Bears (11.6 FP/G allowed) last year was a lot harder than playing the 49ers (21.8), so we have to adjust each QB's weekly score by the stinginess of the opposing defense. What we want to do now is calculate the expected fantasy points for each QB based on his strength of schedule. So if Joe Average played 2 games against a defense that allowed 20 FP/G to QBs, 1.4 games against defenses that allowed 14 FP/G and 0.9 games against a defense that allowed 13 FP/G, we'd expect Joe Average to have scored 71.3 FPs, or 16.6 FP/G. If he scored 19 FP/G, then he's pretty good; if he scored 13 FP/G, he's not very good. Below is the list of the expected fantasy points each QB would have scored if the league average QB played his schedule.

A few interesting things come up already - seeing Marc Bulger at the top of this list might help explain why he was at the top of the earlier list. When you play an easy schedule, you should put up some good numbers. So if Bulger plays a harder schedule and doesn't perform as well in 2006, that shouldn't come as a surprise to any of you.

That table isn't enough by itself though - let's combine the two tables, and rank the QBs by how much value they added. More than any other metric you'll read about this summer, this shows exactly how good of a fantasy player each QB was when he was on the field.

I can pretty much guarantee you that this is the only chart you'll see that shows that Brett Favre was a better fantasy player than Eli Manning last year. But that's a good thing, and you can use it to your advantage. As we saw above, Eli played one of the league's easiest schedules last year, while Favre played one of the hardest. While Manning ranked 4th in total FPs and Favre ranked 13th, we would expect that given their strength of schedules, if the QBs were perfectly equal in ability.

What I always find interesting when I run this study is how few QBs really are difference makers. Twenty-three of the thirty-seven QBs in this study scored within 2 FP/G of what you'd expect based on their SOS. Only a few players every year are truly difference makers; the rest of the top 15 is littered with guys who had easy schedules. You can use this to your advantage by playing QBBC based on weak schedules, and by avoiding those QBs with inflated values based on their prior success.

The last chart you'll see is an important one. I ranked all the QBs by both their actual fantasy points per adjusted game rank, and by their value added rank. This should help to identify which QBs will be sleepers and busts based on ADP, which I think largely mimics the previous season's fantasy points per game ranking. The key to beating ADP is to find useful statistics that the average player ignores, and I think the analysis described above is a great example of that.

Brett Favre, Aaron Brooks and Jake Delhomme are three guys that will be underrated because they faced a tough schedule last year. Delhomme and Favre both started 16 games so they accumulated impressive fantasy totals, which will somewhat mitigate their value. (Favre also led the league in passing attempts, so an expected decrease in attempts will also hurt his fantasy value.) But a guy like Brooks really jumps out here. In both standard scoring leagues and in fantasy points per adjusted games, Brooks ranked as the 16th best fantasy QB. But he had a pretty difficult schedule, and he ranked as the 10th best QB in terms of actual value added over the league average. This means that if Brooks plays exactly the same again this year, he should be about the 10th best fantasy QB. But when you consider that he played most of last year without a home stadium, Deuce McAllister or a healthy Joe Horn, and now he'll be playing with LaMont Jordan and Randy Moss, the 2005 version of Aaron Brooks could very easily be a top 5 fantasy QB in 2006.

Also interesting is how poorly Kerry Collins performed. Collins was second in the NFL in pass attempts, which largely explains his ranking as the 9th best QB. Yet he ranked just 13th in FP/AdjG, and he faced a very easy schedule. If you face one of the easiest schedules in the league and throw a ton of attempts, ranking 9th is far from impressive. I'm not sure if this is a positive or negative for Aaron Brooks (who takes over Collins' role as QB in Oakland), but it sure helps to understand why Kerry Collins won't be back there this season.

One guy to downgrade is Drew Bledsoe. He ranked 6th in standard scoring leagues, based in large part on his ability to play 16 games. But based on FP/AdjG he ranked just 17th, and his schedule was pretty easy too. In reality, he just wasn't a very good fantasy QB last year, but he stayed healthy and didn't have a hard schedule, so he produced good numbers. The addition of Terrell Owens assures that Bledsoe's ADP will remain high, but buyer beware.

Perhaps the most interesting name on the list is Daunte Culpepper. Everyone knows Culpepper had a horrendous year, ranking thirty-third among QBs…or did he? He faced a brutal schedule for the 6.1 games he played, yet he managed to actually do pretty well. Carson Palmer and Tom Brady also stand out because not only did they rank 1st and 2nd in total fantasy points, but they did it with difficult schedules. They added more value than any other QBs in the NFL last season that played in at least ten adjusted games. And something surprising to me was that Mike Vick added more total value than all QBs not named Palmer, Brady or McNabb.

Now that we've talked about a few individuals and I've thrown a lot of data at you, you might be wondering what's the best way to use this. Here's what I think. We almost always start our fantasy projections and rankings based on last year's results. I'd scratch that, and use the above chart - the one sorted by value added - as the starting point for my projections. This way all the QBs are on the same figurative playing field. You now know how much fantasy value each QB added whenever they stepped on the field, adjusted by strength of schedule. Sure Jake Plummer and Mark Brunell rank pretty low on this list, because it's based strictly off 2005 results. That's why it's only a starting point: if you think Javon Walker or Antwaan Randle El and Brandon Lloyd are going to have big seasons, by all means upgrade their QBs. If you think Carson Palmer or Tom Brady will have a hard schedule in 2006, then you downgrade them. But if you look at last year's results (which of course factors in their hard schedules last year) and downgrade them against based on what you perceive to be a difficult schedule, you're penalizing them one too many times.

One last thing to help validate the usefulness of this rearview strength of schedule analysis. We would imagine from that from year to year, a QB's value added point total should stay more consistent than a QB's actual fantasy point per game total. And it does. Of the 26 QBs that threw enough passes in both seasons, the correlation coefficient of the quarterback's FP/AdjG in 2004 and 2005 was 0.40; the correlation coefficient of each QB's value added points per game was 0.59. For those not mathematically inclined, this simply means that a QB's production can be better explained by the value added metric than straight fantasy points per adjusted game. At least in my mind, this goes a long way towards validating this system.

Drew Brees is a great example of this. In 2004 he averaged 19.6 FP/AdjG and in 2005 he averaged 17.7 FP/AdjG. You might think this is because he had a career year in 2004 and he just wasn't as good last year. But based on his schedule, he was expected to average 18.1 FP/G in 2004, and just 16.5 FP/G last year. Passing numbers league-wide were much higher in 2004. Brees was a little better than the average QB both years, but his comparatively easy schedule in 2004 inflated how good he really was.

One last note. This study doesn't factor in how many pass attempts each QB had. So Donovan McNabb ranks a lot higher than Ben Roethlisberger, in part because he averaged 20 more pass attempts per game. This isn't the article for a detailed study of how consistent team or individual QB pass attempts are from year to year, but in general I'd expect both guys to move closer towards the league average in 2006.

2004 End of Year Adjusted Fantasy Totals