Blog Archives

Winning with the Apps: The Similarity Score Apps

joshgordon

Prior to the 2013 season the WR Similarity Score App projected Josh Gordon to finish as a top 15 receiver based on his similarities to other top players like Calvin Johnson, Demaryius Thomas, and Vincent Jackson. Here’s a simple graph that the Sim Score Apps give us for each player that shows the range of outcomes that their comparable players had in followup seasons:

download-17

You can see from the App that comparable players Calvin Johnson (2008) and Vincent Jackson (2008) essentially doubled their fantasy output (the % change is actually a decimal) from the season that they were similar to Josh Gordon 2012. At the same time, players like Matt Jones and Keary Colbert, who had also had seasons similar to Josh Gordon, ended up declining.

At the heart of the Similarity Score Apps engine is a scoring system that grades players based on similarities in a number of box score stats, along with player size (weight) and age. The simple assumption is that if you’re looking at players that are roughly similar on a number of measures, then one player’s follow up season is probably informative of what we can expect from the subject player. By aggregating 20 or more such comparables we can actually produce projections that show a range of outcomes.

Let’s look at a player for the upcoming season and how the Similarity Score Apps can help.

Here is a summary of Michael Floyd’s 2013 season:

NAME
SEAS AGE WEIGHT GMS TRGS RECS YDS YPR YPT TDS
Michael Floyd 2013 24 220 16 7.06 4.12 65.88 15.97 9.33 0.31

Now here are his comparables from the App:

NAME
SEAS AGE WEIGHT GMS TRGS RECS YDS YPR YPT TDS
Dwayne Bowe 2007 23 221 16 7.31 4.38 62.19 14.21 8.5 0.31
Plaxico Burress 2001 24 226 16 7.5 4.12 63 15.27 8.4 0.38
Donald Hayes 2000 25 220 15 7.47 4.4 61.73 14.03 8.27 0.2
Darrius Heyward-Bey 2011 24 210 14 8.21 4.57 69.64 15.23 8.48 0.29
Miles Austin 2010 26 217 16 7.44 4.31 65.19 15.12 8.76 0.44
Andre Johnson 2003 22 225 16 7.44 4.12 61 14.79 8.2 0.25
Andre Johnson 2004 23 225 16 8.62 4.94 71.38 14.46 8.28 0.38
Marques Colston 2008 25 225 11 8 4.27 69.09 16.17 8.64 0.45
Michael Crabtree 2011 24 214 15 7.67 4.87 58.67 12.05 7.65 0.27
Dez Bryant 2011 23 220 15 6.93 4.27 62.07 14.55 8.95 0.6
A.J. Green 2011 23 211 15 7.67 4.33 70.47 16.26 9.19 0.47
Ronald Curry 2004 25 220 12 5.83 4.17 56.58 13.58 9.7 0.5
Derrius Thompson 2002 25 220 15 6.2 3.53 51.53 14.58 8.31 0.27
David Givens 2004 24 212 15 7.07 3.73 58.27 15.61 8.25 0.2
Dwayne Bowe 2011 27 221 16 8.88 5.06 72.44 14.31 8.16 0.31
Pierre Garcon 2012 26 210 10 6.8 4.5 64.4 14.31 9.47 0.4
Dwayne Bowe 2008 24 221 16 9.81 5.38 63.88 11.88 6.51 0.44
Darrell Jackson 2002 24 201 13 8.62 4.77 67.46 14.15 7.83 0.31
Quincy Morgan 2002 25 210 16 6.06 3.5 60.25 17.21 9.94 0.44
Anquan Boldin 2006 26 220 16 9.5 5.19 75.19 14.49 7.91 0.25
Vincent Jackson 2008 25 230 16 6.31 3.69 68.62 18.61 10.87 0.44
Larry Fitzgerald 2006 23 218 13 8.54 5.31 72.77 13.71 8.52 0.46
Javon Walker 2006 28 220 16 7.88 4.31 67.75 15.71 8.6 0.5
Marques Colston 2009 26 225 16 6.62 4.38 67.12 15.34 10.13 0.56
Amani Toomer 2001 27 208 16 8.81 4.5 65.88 14.64 7.48 0.31

If you did some averaging of those comps (all done by the App) you would see that the composite player they produce ends up being really close to Michael Floyd 2013.

AGE WEIGHT GMS TRGS RECS YDS YPR YPT TDS
–Average– ———– 24.7 218 14.8 7.65 4.42 65.06 14.81 8.6 0.38

By looking at the follow up year for each WR in the group, we can create a projection for Michael Floyd in 2014.

- Standard Half PPR PPR
Low 5.9 8.2 10.6
Median 9.1 11.6 13.7
High 10.4 13 15.2

This projection range is the equivalent of saying “For players that were similar to Michael Floyd 2013, what happened in the year after they were similar to Floyd?”

The App also allows you to customize a projection in the event that you think some piece of information is being overlooked. For instance, if you wanted to forecast Cordarrelle Patterson by only looking at his 2nd half of 2013, you could do that. You would just remove weeks 1-7 from his game log.

download (31)

It’s probably worth noting that when you start customizing projections, it’s possible that you’re going to make a more accurate projection, but it’s also possible to become less accurate. Any time you rely on a cherry picked or limited sample size, it introduces potential for errors to your projections.

The Similarity Score Apps create one very specific kind of forecast that we’ve found to be successful in winning fantasy leagues. But it’s also just one way to forecast players and it answers a very specific question: What happened to similar players in a follow up season?

Use the App to look for undervalued players, but always be sure to combine the results of the App with your own logic. In some cases you’ll know that the App is probably offering a value trap, like in the case of a WR that will see diminished opportunity that the App can’t pick up on. Like any source of analysis, the Similarity Score Apps are best used when combined with your own judgment.

Prior to the 2013 season the WR Similarity Score App projected Josh Gordon to finish as a top 15 receiver based on his similarities to other top players like Calvin Johnson, Demaryius Thomas, and Vincent Jackson. Here’s a simple graph