As an avid and serious fantasy football player, I’m always looking for the edge that will put me ahead of my opponents and take me to the championship. As you know, football and fantasy football are games of inches (or fractional points).
When you’re stuck at 12:45pm on Sunday trying to decide between starting Terrell Owens or Johnny Knox in that WR3 or Flex position, and the projections from ESPN and Yahoo are varying (and never accurate anyway), what do you do?
Luckily, I recently discovered a program called numberFire that projects weekly fantasy points using an algorithm that actually makes logical sense (compared to random projections that I think Yahoo and ESPN make).
The algorithm essentially looks at two variables: player and matchup. The formula takes into account the player’s statistical data and skill set as well as the matchup between the two teams.
For example, when evaluating Peyton Manning, the program will consider his past statistics, the type of offense the Indianapolis Colts use, Manning’s playing style, his willingness to throw down the field (as determined by YPA), whether he’s a clutch player (as determined by a higher or lower passer rating in the fourth quarter), and so on.
After that, a list of results come up comparing Manning to similar type of players along with a match percentage: Drew Brees in 2009 is a 96.51% match, Philip Rivers in 2009 is a 94.23% match, Drew Brees in 2006 is a 92.92% match.
In addition, the program will evaluate the offense vs. defense matchup when Manning’s Colts play the Denver Broncos in Week 3. The Broncos use a 3-4 defense, which will be taken into account, as well as the personnel, packages, formations, playcalling, etc. The program will then look at similar offense vs. defense matchups from past NFL games as a blueprint for projection.
The highest match percentage is the Cincinnati Bengals offense vs. the Cleveland Browns defense in Week 2 of the 2007 season. The result yielded a 94.67% match.
After gathering all of the information, the program will come out with a list of players that Peyton Manning is most similar to by comparing Manning’s statistics and skill set to other players while factoring in the type of offense he’s running and the type of defense he’s playing against.
Here is a screenshot of what I mean:
As you can see, after considering all factors, the program lists the most similar player as Peyton Manning vs. the 2009 Denver Broncos with a 97.9% match. Another good comparison would be Philip Rivers vs. the 2008 Tampa Bay Buccaneers with a 95.86% match.
The Randall Cunningham result is arbitrary because the data was pulled from a game in 1998. The further back the data goes, the less trustworthy it is because offensive and defensive schemes have changed a great deal since then. Therefore, the most recent results are the most reliable.
The projection is made by finding all of the times that someone similar to Manning is playing for an offense similar to Indianapolis and against a defense similar to Denver. The application then uses the results to generate a projection for what we can expect Manning to do against the Broncos defense.
For this week, numberFire projects 20.54/31.45 completions, 258.19 yards, 2.07 TDs, and 0.83 INTs for 17.14 points. The point system is based on a standard scoring league.
In addition, what the program allows you to do is compare numberFire’s projections to those of Yahoo and ESPN’s. The user can actually use their Yahoo ID to log in and import their fantasy teams into the application. The team below is one I’ve imported from a Yahoo league:
To see how accurate the program is, numberFire has come out with a list of popular players with their projections and actual results from Week 1:
The program beat ESPN 13/22 times (59%) and beat Yahoo 17/22 times (77%). Here are the projections and results from Week 2:
The program beat ESPN 14/22 times (63.6%) and beat Yahoo 13/22 times (59.1%).
At this point, the sample size is too small to determine just how good the program is. However, the early results clearly show a marked improvement in fantasy projection over both Yahoo and ESPN.
Because numberFire is new, there is currently a free trial going on that will last for two more weeks. After that, users must pay in order to receive premium services, which would be projections for all players. I would highly suggest signing up right now to take advantage of the free trial.
The jury is still out on numberFire. But, if it continues to show more accuracy in projections than Yahoo and ESPN, the program will definitely be worth the small amount that users will have to pay to become premium subscribers.
Check out the program at numberFire.com and let me know what you think.