I spent a good amount of time over the weekend updating my WR model in anticipation of upcoming rookie drafts, so I figured I would do a post today with some initial ranks. Just to be clear, I expect these ranks to change after the NFL draft, but through the magic of the internet I can actually update them at that time.
My model, like a lot of algo projection systems for WRs is primarily focused on college production measures including some way of normalizing production for scheme, era, etc. and I call that normalized measure Market Share. But this year I wanted to add in a variable for player age as I noticed that it’s probably not fair to compare 23 year old WRs to 20 year old WRs. The former should be more experienced and also physically developed after having spent most of a college career in a college weight room. But in addition to adding age in as a variable, I was also able to add in a variable for the player’s physical measurables, which I hadn’t included before except in a weight adjusted 40 time that was barely significant. In this iteration I’ve actually included a Physical Score which is simple in that it just scales Weight, 40 Time and Vertical Leap and adds those scaled values together (with a favoring of Weight, 40 Time, and Vertical Leap in that order based on finding the best fit). I actually experimented with adding Broad Jump and Height to Physical Score, but dropped them when they didn’t reduce the model’s errors.
The most significant variables in the model ended up being (in order) the two touchdown measures, Physical Score, Market Share Yards, Age, Yards per Reception, Rushing Yards/Game. The dependent variable that I used in the model was average standard fantasy points/game for seasons that the player was less than 26 years old. I’m sure there are any number of problems with using this as the dependent variable, but the advantage it has is that it is an average of a good number of games. So it’s indicative of the overall quality to the start of a WR’s career.
A few notes on the rankings below:
- I don’t actually expect Stedman Bailey to be the most valuable WR in this class. His draft position is going to seal his fate there. But his close to 2 touchdowns per game and 60% of WVU’s touchdowns are frankly ridiculous. He’ll be a steal for the NFL team that drafts him and may provide value in dynasty leagues depending on the situation he ends up in.
- That probably makes DeAndre Hopkins my top WR then. He’ll end up going in the first round, probably in the 20s to a team with a good QB and maybe even a decent WR on the other side of the field.
- You may note that Tavon Austin has the lowest Physical Score of any of the WRs listed. That’s because only his 40 time is good. His VL is well below average and PS favors heavier players. Because standard fantasy scoring is heavily influenced by touchdowns, larger players have an advantage. For most WRs, that advantage carries into PPR scoring as well, although PPR does level the weight-playing-field a little.
- Da’Rick Rogers’ numbers are based on his 2011 season (and reducing his age to the age he was during that season), which may or may not be cheating. Whether or not it’s cheating probably doesn’t matter for an NFL team, assuming that Rogers can be acquired with a late 2nd or early 3rd round pick. He’s only upside at that point.
- Terrance Williams suffers from the age penalty included in the model, which subtracts about 1/2 of a fantasy point per game from the forecast for each year that a player is old.
- Cordarrelle Patterson benefits from two additions to the model which I hadn’t used before, the Physical Score, and Rushing Yards/Game.
- These rankings can’t possibly take into account the fact that Justin Hunter was coming back from an ACL tear.
- Some of the variables in this model are going to correlate with draft position, which means that when I add in draft position as a variable, some of them will get dropped.
- This list doesn’t include Keenan Allen as I don’t have any workout information for him.
- Jon Moore’s guy Charles Johnson would perform extremely well in this model and I would have included him if I didn’t have reservations on comparing apples and oranges related to production measures. But on the physical measures Johnson would be the best prospect by a decent margin.
- Could these rankings be wrong in general? Hell yeah they could. Draft position explains a pretty small part of WR production. This model explains quite a bit more, but also leaves quite a bit unexplained. But also, this is football. Even using a regression to explain year to year fantasy points has limitations – when we’re not talking about players going from college to the NFL. So is this a good system to project WRs? It depends… but it’s probably as good or better than any other.
|Stedman Bailey-West Virginia||193||4.52||34.50||(1.23)||22||0.38||0.60||124.77||1.92||14.23||1.00||22|
|Tavon Austin-West Virginia||174||4.34||32.00||(2.02)||21||0.30||0.29||99.15||0.92||11.31||49.46||21|
|Robert Woods-Southern California||201||4.51||33.50||(0.59)||20||0.25||0.31||70.75||0.92||11.47||6.33||20|
|Quinton Patton-Louisiana Tech||204||4.53||33.00||(0.61)||22||0.34||0.43||116.00||1.08||13.38||(0.42)||22|
|Markus Wheaton-Oregon State||189||4.45||37.00||(0.27)||21||0.32||0.44||95.69||0.85||13.67||10.92||21|