GILLESPIE Projection: RGIII vs. CIN
This is just a test run of GILLESPIE as I work on getting all of the formatting down.
GILLESPIE (Game Level Similarity Projection) is based on looking at similar players against similar defenses. It’s something of a lazy way of forecasting. Instead of trying to figure out the relationships that a number of variables might have with fantasy scoring (and the relationships that those variables have with each other) we just make an assumption that if we just look at similar players vs. similar defenses, we’ll get a decent projection. If you test GILLESPIE, or a k-nearest neighbor projection, which GILLESPIE is based on, against a linear regression, you’ll find that it can improve upon the linear regression. To be fair, the improvements that GILLESPIE will have over a linear regression will be modest.
The first table below contains a summary of the subject player’s season to date stats, along with a summary of the defense’s stats. The columns on the right hand side of the table are devoted to the projection. Then in the second table I show all of the games featuring similar players against similar defenses. The results of those games are shown in the columns on the right side of the table.
In order to save space I’ve combined the player and defense’s season summary stats into one column. In order to understand them you need to know the format, which is as follows: Att/G, Yards/G, Yards/Att, TD/G, Int/G, Rushing Yards/G.
You’ll see that through two games RGIII has been really close to Michael Vick’s 2010 season. So GILLESPIE is going to look for defenses that Vick faced that are similar to the CIN defense. Again, this is just a test post so I think it’s important to keep in mind that a number of the summary statistics are going to change. RGIII won’t average 9.6 yards/attempt through the season. That number will come down for sure. So basically, any projection we make today is going to have even larger forecast errors than usual (and forecasting the future is always a difficult endeavor).
Next week we’ll have more data so I’ll be posting more things like some “Start of the Week” type stuff, as well as maybe some sneaky starts as well. Some of these posts may go up on Rotoworld, but I’ll always try to provide links at this site.
Note for fantasy scoring I generally use 5 points for passing touchdowns in order to split the difference between the two major scoring formats.
Subject Player
| Season Summary | Game Projection | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Name | Yr | Player Summary | Def | Def Summary | Att | Yds | Y/A | TD | INT | RYDS | RTD | FPS |
| Robert Griffin | 2012 | 27.5/263/9.6/1.5/0.5/62 | CIN | 34.5/325.5/9.4/2/0/6 | 29 | 252 | 8.7 | 1.9 | 0.5 | 45.5 | 0.5 | 26.1 |
Comparable Matchups
| Season Summary | Game Stats | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Name | Yr | Player Summary | Def | Def Summary | Att | Yds | Y/A | TD | INT | RYDS | RTD | FPS |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | HOU | 34.2/281.1/8.2/2.1/0.8/10.8 | 33 | 302 | 9.2 | 2 | 1 | 48 | 1 | 30.88 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | JAC | 31.6/261.9/8.3/1.8/0.8/6.9 | 31 | 290 | 9.4 | 3 | 0 | 30 | 1 | 35.6 |
| Michael Vick | 2004 | 21.4/154.1/7.2/0.9/0.8/60.1 | KC | 32.6/277.3/8.5/2/0.7/14 | 21 | 119 | 5.7 | 0 | 2 | 62 | 0 | 6.96 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | DAL | 33.8/259.2/7.7/2.1/1.2/8.9 | 26 | 270 | 10.4 | 2 | 2 | 16 | 1 | 24.4 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | WAS | 36.1/271.4/7.5/1.4/0.9/15.2 | 28 | 333 | 11.9 | 4 | 0 | 80 | 2 | 53.32 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | WAS | 36.1/271.4/7.5/1.4/0.9/15.2 | 7 | 49 | 7 | 0 | 0 | 17 | 0 | 3.66 |
| Cam Newton | 2011 | 32.3/253/7.8/1.3/1.1/44.1 | MIN | 33.7/272/8.1/2.1/0.5/10.5 | 35 | 290 | 8.3 | 3 | 0 | 53 | 0 | 31.9 |
| Tim Tebow | 2011 | 20.9/132.8/6.3/0.9/0.5/50.8 | MIN | 33.7/272/8.1/2.1/0.5/10.5 | 15 | 202 | 13.5 | 2 | 0 | 13 | 0 | 19.38 |
| Michael Vick | 2002 | 28.1/193.5/6.9/1.1/0.5/51.7 | MIN | 33.9/268/7.9/2.1/1/15.1 | 28 | 173 | 6.2 | 1 | 1 | 170 | 2 | 38.92 |
| Michael Vick | 2002 | 28.1/193.5/6.9/1.1/0.5/51.7 | DET | 36.9/275.1/7.4/1.7/0.6/12.4 | 38 | 336 | 8.8 | 2 | 1 | 42 | 0 | 25.64 |
| Michael Vick | 2004 | 21.4/154.1/7.2/0.9/0.8/60.1 | OAK | 31.6/252.8/8/1.9/0.6/12.8 | 20 | 145 | 7.2 | 0 | 0 | 31 | 0 | 8.9 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | DET | 32.6/235/7.2/1.4/0.9/16.6 | 35 | 283 | 8.1 | 2 | 0 | 37 | 0 | 25.02 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | IND | 33.5/225.6/6.7/1.4/0.6/11.5 | 29 | 218 | 7.5 | 1 | 0 | 74 | 1 | 27.12 |
| Michael Vick | 2011 | 32.5/253.7/7.8/1.4/1.1/45.2 | DAL | 34.1/259/7.6/1.5/0.9/11.8 | 28 | 279 | 10 | 2 | 0 | 50 | 0 | 26.16 |
| Michael Vick | 2011 | 32.5/253.7/7.8/1.4/1.1/45.2 | DAL | 34.1/259/7.6/1.5/0.9/11.8 | 32 | 292 | 9.1 | 2 | 0 | 10 | 0 | 22.68 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | NYG | 33.8/228.8/6.8/1.5/1/17.1 | 35 | 242 | 6.9 | 3 | 1 | 130 | 1 | 41.68 |
| Michael Vick | 2010 | 31.1/251.2/8.1/1.8/0.5/56.3 | NYG | 33.8/228.8/6.8/1.5/1/17.1 | 38 | 258 | 6.8 | 0 | 0 | 34 | 1 | 19.72 |
| Aaron Rodgers | 2009 | 33.8/276.9/8.2/1.9/0.4/18.9 | DET | 34.1/271.2/8/2.2/0.6/4.2 | 37 | 357 | 9.6 | 2 | 1 | 8 | 0 | 23.08 |
| Aaron Rodgers | 2009 | 33.8/276.9/8.2/1.9/0.4/18.9 | DET | 34.1/271.2/8/2.2/0.6/4.2 | 39 | 348 | 8.9 | 3 | 0 | 5 | 0 | 29.42 |
| Donovan McNabb | 2002 | 36.1/228.8/6.3/1.7/0.6/45.9 | ARI | 33.4/249.8/7.5/1.8/1.1/3.6 | 25 | 255 | 10.2 | 4 | 1 | 0 | 0 | 28.2 |
