For the past four weeks, like many other Americans (mostly men) who are simply not satisfied enough with just watching their favorite teams play, I have been drawn into the world of Fantasy Football. This being my first year playing, there was certainly a learning curve for me (which I continue to attempt to overcome). For those of you who don’t know what Fantasy Football is, it is simply choosing the best (i.e. your favorite) players from all the many players in the NFL, and forming them into a motley team under whatever clever name you’ve come up with. Then, as the weeks progress, if your players do something good on the field, your team gets points; if your players do something bad on the field, you lose points. At the end of each week, the team with the most points is the winner.
Obviously, what makes this pseudo-sport difficult is selecting the players for your team. Everyone is choosing from a limited supply of players, and you want to make sure you have as many high scoring players as possible. The real problem, however, comes from how unpredictable sports players can be. While a player on your team may have seemed amazing before, they could totally fail you and get few points when playing against a different team each week. Similarly, the player who was ignored because of bad past performance may suddenly step up and leave everyone wishing they had him on their team. Therefore, being able to predict the performance of each player is crucial to a good Fantasy Football team.
Having never played Fantasy Football before, and being slightly uniformed (read, “nearly completely ignorant”) of football players and performance, I was very happy to discover that both NFL.com and ESPN.com provided their own “predictions” of each players upcoming week performance. For instance, if a player was playing a team with a good defense, their score might be low this week and vice versa when facing teams with poor defense. While this is just one example of what goes into the prediction algorithm, there are certainly many other more complex aspects of the equation. Therefore, when choosing my own team, I relied heavily (nearly 100%) on these score predictions so at least if I hadn’t ever heard of a player, I knew that someone had and thought they’d be good.
As I mentioned earlier, however, players can be…inconsistent (which is why we watch and are entertained by sports). Even with delicate algorithms to predict success, the highest ranked player on your team could have a mediocre game that causes you to lose this week.
This can be very frustrating.
For those of you who have read my past posts, you will know that I like to spend a lot of time talking about the problems with expectation. It is a trap that seems nearly impossible not to fall into the moment you think you’ve climbed out, and Fantasy Football proved to be no exception. Though it was highly enjoyable when a player you thought would be mediocre does well, it was even more disappointing when a player you thought would do well lets you down. For the past four weeks, watching the games and my players miss throws and drop the ball has only led to increased stress.
I am not okay with this.
Watching sports is supposed to be fun.
This was not fun.
So, if you can’t predict a player’s success, can you at least reduce how much disappointment they create? After the first week of playing, this became my purpose.
WARNING: Statistics ahead.
After creating a spreadsheet containing each players’ week’s predictions, I compared them to the week’s actual scores to find the difference. By taking the standard deviation of these differences, I was able to determine the standard error for each role. For instance, by my calculations using NFL.com predictions, Quarterbacks who played in week 1 were off from their predictions by an average of -0.77 pts and a standard error of 6.68 pts. By comparing the Standard Error to the player’s score difference, I am able to find out where they fall in a normal distribution of score inconsistency. By using this calculation, we can discover how likely a player is to play as well as we expect.
For example, in week one, Matt Ryan (Quarterback for the Atlanta Falcons) was predicted to score 13.32 points. Instead, he had an amazing week and scored 32.46 points, putting him 19.14 points more than expected. Since the Standard error for QBs that week was 6.68 points, this means his error was 2.86 standard deviations (2.86 = 19.14 / 6.68) away from how off most QBs were. That week, Matt Ryan was the most satisfying QB in terms of going above and beyond expectation.
On the other hand, that same week Matthew Stafford (Quarterback for the Detroit Lions) was predicted to score 27.52 points and only managed 12.50 for a difference of 15.02 points. This put him at 2.25 standard deviations BELOW the mean (2.25 = 15.02 / 6.68). That week, Matthew Stafford was the most disappointing QB in terms of living up to expectation.
Over time, team owners can use these formulas to make sure they don’t have any players on their team who are regularly disappointing (and get players who regularly go above and beyond). However, the reliability of a player isn’t the only thing that matters. A player who plays below his predicted value could still get more points than a player who plays above his predicted points. For instance, Matthew Stafford still got more points than over a dozen other QBs in the league, even though he was a severe disappointment in terms of expectation. Somehow we must factor in the predicted score so that we not only get highly predictable, but also highly scoring players. To do this, we simply rank the players based on their predicted score for this week minus their average deviation over the year (which gets more precise after each week). Depending on ability to handle disappointment, team owners can then rule out any who often prove to be highly disappointing or go for players who often exceed expectation.
Of course, this is not fool proof, and even players prone to success can have horrible weeks. Also, many of you might point out that I’m just creating additional expectation on top of already existing expectation. I suppose, then, that the point of this is that statistics are awesome, but they can only predict so much. Or maybe this should turn into an article about how desperate we become to try to predict even the unpredictable. Or maybe the point is just that football and sports in general can be stressful when you get too wrapped up in them.
Well, at least for the next week I can rest easy thinking that I’ve somehow predicted my way out of a stressful weekend. And how am I doing at Fantasy Football, you ask? Well, I’m getting better…