Creating Simple 2013 Projections

In a few weeks, I will join fellow dynasty writers Bryan Fontaine and Chad Parsons in their monthly composite dynasty rankings. Traditionally, I have arrived at rankings in various formats before my drafts, after which I’ve made adjustments less formally. For my sanity, I decided that now was the time to improve the efficiency of that process, and so I have started to translate that production into database queries that will make it easier for me to duplicate the work.

I have done a lot of research into player value in future seasons based on risk classifications, which I plan to include down the line. First, I need 2013 projections. Previously, I have relied on available projections like those we publish on the site or on previous season value-based drafting assessments that I have manually adjusted. Those methods are fine for near-draft preparation, but I need a system that is updated to include the most recent statistics on my schedule, and so I decided to create my own projections.

I am unabashedly fascinated by projection systems, and I fully intend to create a sophisticated one of my own in the future. However, Bryan probably wants some rankings from me before 2015, so I am going to use as simple a method as I believe is appropriate.

First, our dynasty rankings are intended to be universally applicable. The only declared specific of the rule set is PPR. Otherwise, it is a blank slate. Of course, I am a pedant, and so I feel compelled to spell out that I will rank players based on a 10-team league with 16-man rosters and standard PPR scoring. Here is the roster breakdown:

The VBD total column identifies the player rank of each position that I designate as the replacement level. I reallocated the 70 total bench slots among the four major positions, assuming each team would carry only one kicker and one defense. The 16 unallocated bench slots represent flier or handcuff selections that do not affect the replacement level. For example, if I needed a running back in a given week because of byes or injury, I would not pick up Bernard Pierce. In a 10-team league, there will be better options on the wire. However, if I owned Ray Rice, I might prefer to own Pierce over someone on the wire I would expect to outscore him that specific week.

For the projections, I opted for a 5-3-2 weighted average of player production in 2012, 2011, and 2010. That method works well for the Roddy Whites of the world that never miss a game. However, I needed a way to handle players like Peyton Manning. His missing of the entire 2011 season does not dramatically impact my opinion of his prospects in 2013, so it would be unfair to allocate him a zero for 30 percent of his average.

Strangely, I feel more confident in Manning than I do in someone like Michael Vick or Ahmad Bradshaw, neither of whom has missed a season in recent years but both of whom routinely miss a few games each year.

Injuries threatened to make this exercise complicated, but simple was my directive, and so I made some arbitrary determinations. First, if a player missed all of either 2011 or 2010, I used a weighted average of 2-1 in 2012 and the other healthy season. For rookies or players that only had 2012 data, I gave it the full weight.

I handled the in-season injuries a bit more subtly. Since I deemed those more likely to repeat, I wanted to penalize players that had missed time. When a player is out, you replace him with the best available option, which I would define as it applies universally as the replacement level. Well, I already outlined where the replacement level is at every position, which means I had everything I needed to create a formula. Here is the one for a player that has not missed a recent season:

2013 Points =

0.5 * (2012 Games * 2012 PPG + (16 – 2012 Games) * 2012 Replacement-level PPG) +

0.3 * (2011 Games * 2011 PPG + (16 – 2011 Games) * 2011 Replacement-level PPG) +

0.2 * (2010 Games * 2010 PPG + (16 – 2010 Games) * 2010 Replacement-level PPG)

That formula does not tell you how many points the player will score, but rather the points you expect out of his position because you drafted him. And, really, that is the more important information.

I only used points scored for the first 16 weeks of each season since most leagues exclude week 17, and so every player has at least one week of replacement-level production in their projection to account for their bye week. To clean up the results a bit, I added filters so that a player had to play at least six games to be included in a season and that a quarterback had to combine for at least 20 pass and rush attempts to be considered in a game.

Here are the results by position:

The top of the quarterback board looks similar to everywhere else. Michael Vick is the first surprise at six, which in a way makes sense. In recent seasons, Vick has always been productive when he’s been on the field, and you should be able to combine Vick with a replacement-level player in the weeks he misses to create a top-10 quarterback. However, this method likely still overrates him a bit because of the pass and rush attempts filter. It does not consider games where Vick leaves early, and those are killers.

Colin Kaepernick is at 12, which is pretty good when you consider he became the starter in week 10 and so is saddled with replacement-level production for the first half of the season. Kaepernick is a prime example of the type of player this method will systematically undervalue, which is one that sees his role expand in the middle of a season. I expect to see names like C.J. Spiller with a similar bearish projection.

The top of the running back board comingles Doug Martin, Trent Richardson, and Alfred Morris with the likes of Ray Rice and Jamaal Charles despite the differences I have in confidence in their projections given their disparate track records. That is fine. The valuation method I use to price players accounts for risk, and so that will be accounted for later.

There is Spiller down at 23, and a few spots later, Stevan Ridley. Ridley suffers a similar but slightly different problem. One third of his projection is based on a 2011 season before he was a featured player, but he still played, and so that (non-)production is still a part of his average. Accounting for that in the query would be too complicated, and it affects few enough players that I can manually adjust them later.

A few prospects like Lamar Miller and David Wilson start to show up in the 50s. This method is built on past production, and so it is useless for these types of players. I will have to return to them later, as well.

While Reggie Wayne nearly cracks the top-10, guys like Larry Fitzgerald and Steve Smith fail to crack the top-20. Sometimes, consistently elite players have a bad season, and this method will really punish them if that season was last season. I’m not so sure that is predictive.

Even with five games of replacement-level tight end production in 2012—which was terrible, by the way—Rob Gronkowski still easily paces the position. After the first six guys, the separation between players is very small. When you consider that Gonzalez and Gates may be on the way out, it appears you will need to invest an early pick to snag one of the top-four names. Otherwise, you’re better off waiting until the last round.

Overall, the projections seem realistic. That’s a start. Over the next few weeks, I will make the necessary adjustments and incorporate future value based on risk. Hopefully, it won’t be too long before I have some rankings.

—

Scott Spratt also works for Baseball Info Solutions and writes for The Hardball Times.  Feel free to ask him questions on Twitter – @PFF_ScottSpratt

Find more of our Dynasty Content here.