Choosing a Quarterback: And This Time I Mean It

I know I’m supposed to start off with a snappy lede to grab your attention but I feel I owe some of you an apology. I’ve spent the last several weeks attempting to develop a quarterback ranking system as I strongly feel nailing the quarterback is integral to DFS success, both for cash games and tourneys. A method I proposed a few weeks back that received some positive feedback has turned out to be rather ineffective and for that I’m sorry. That said, a lot of research and the scientific process is trial and error and if we learn from our mistakes, the effort isn’t a total loss.

As it turns out, I confused cause and effect and was using the wrong filters to rank quarterbacks. By means of reminder, several consistent traits of the top quarterbacks each week were uncovered. The system used these traits to rank the signal callers. These traits are essentially the effect. Instead of focusing on these traits, more focus should have been heeded to the cause.

The real kicker is what will soon be detailed is very much akin to what I do for the DFS baseball projections. I suppose I should have started out in this manner but my reluctance to base this sort of analysis on the sample of a few NFL games proved myopic. While I’m not sure this will be practical early in the season, I now think we’re at the point the abridged process helps mitigate the sample size conundrum.

In short, a baseline projection for each QB will be tweaked based on the match-up. This is identical to what we do for baseball with the exception that the baseline for baseball is based on a projection I generate whereas the football baseline is the player’s performance to date.

The end result is going to be an actual projection of performance that can be converted to fantasy points based on the site’s scoring. This will help fine tune the correlation evaluation used to gauge the success. Instead of comparing a ranking, we’ll compare the projected points to the actual points.

We’ll be projecting the following for each QB:

• Passing yards
• Touchdowns
• Rushing yards
• Interceptions and lost fumbles

PASSING YARDS

1. Start with the player’s average yards per game

2. Calculate the average passing yards allowed of all the teams the QB has faced as well as the league average of passing yards allowed per game

3. Adjust the player’s average as follows: if the QB threw for X yards against defenses that allowed an average of Y yards per game, he should throw for Z yards against an average defense.

Adjusted yards per game = (avg yards per game) X (league avg allowed) / (avg team allowed)

This is how many yards the player should throw against an average NFL defense with respecting to defending the pass.

4. Determine the opposing team’s defense factor by comparing its average yards allowed per game to the league average.

FACTOR = (opposing team’s avg yards allowed per game) / (league avg yards allowed per game)

5. Adjust the neutral passing yards via multiplying by the FACTOR.

Projected passing yards = FACTOR x Adjusted yards per game (Step 3 x Step 4).

6. If the player is at home, multiply by 1.068. If the player is on the road, multiply by .932. The reason for this is on the average, teams throw for 6.8 percent more yards at home as compared to a neutral site.

Let’s carry through a sample calculation for Peyton Manning based on current data. Manning is averaging 322 yards per game. The Broncos have faced defenses allowing an average of 247 yards per game. The league average is 258 yards.

Adjusted yards per game = 322 x 258 / 247 = 335 yards.

This passes the sniff test. Manning has faced defenses better than league average so he would throw for more yards against a league average unit.

This week the Broncos face the Oakland Raiders who sport a defense allowing 237 yards per game.

Projected yards per game = 335 x 237 / 258 = 307 yards

Math majors likely observe that it really isn’t necessary to use the league average as all the above can be combined into one equation with the league average canceling out. However, for the ease of describing the process, it was left in. If you happen to adopt the process, doing this in one step facilitates the calculation.

Finally, the game is in Oakland:

307 x .932 = 286 yards.

The end result is a projection of 286 yards in the air for Manning.

TOUCHDOWNS

1. Subtract all special team and defense scoring from the player’s team total.

2. Determine the percentage of points accounted for by the quarterback

6 x (rushing TDs + passing TDs) / (Team points – 6 x (special team and defense TDs – 2 x safeties)

3. Use the Vegas point spread and over/under to approximate the number of points the QB’s team is expected to put on the board.

4. Determine the number of TD generated by the QB

QB TD = (percent team points by QB) x expected points via Vegas / 6

Granted, there are a couple of assumptions made, but the end result is close enough for jazz. The first assumption is defense and special team touchdowns are random and not predicted by the odds makers. The second is QB rushing TDs are worth the same as passing TDs, which is not the case on all sites. But since running TDs are a much smaller component and somewhat random, the slightly lower points awarded reflect the variance.

Let’s again look at Manning and the Broncos to illustrate this calculation.

Denver has scored 245 points which includes a pick-six and a safety so their adjusted total is 238 points. Manning has thrown for 24 TD and hasn’t run one in.

Percent TD = 6 x 24 / 237 = 61 percent

The Broncos are favored by 11 with an O/U of 49. This portends to a 30-19 final.

Manning TD = (30 x .61) / 6 = 3.05

The end result is Manning is projected to throw (and run for) 3.05 TDs.

RUSHING YARDS

Nothing fancy is done here. The average rushing yards per game for each quarterback is included with no adjustments.

INTERCEPTIONS 

Teams throw ten percent more picks at home so the average per game is multiplied by 1.1 at home and by .9 on the road. Fumbles aren't adjusted.

We now have the four elements necessary to plug into each DFS site scoring system so we can determine the projected number of fantasy points for each QB. Pretty cool, huh. Though, it would have been even cooler two weeks ago.

Here's the Week 10 data, including the projected points on the four major sites.

Admittedly, this may not be as complex a system as some may employ to really fine tune projections, but at the very least it should give you an idea how numbers of this sort are generated along with the practical utility of ranking quarterbacks for DFS use. There's a good amount of stuff to digest so instead of losing you with some further discussion on quarterbacks I like based on the above, please check back later in the weekend for an extra posting where I'll present the consensus industry rankings along with my top-ten complete with commentary and pretty player head shots.