My fantasy football season is over. I finished with an 11-3 record and first place out of ten teams after the regular season. Alas, I lost in the first round of the playoffs. The finals will be played this coming weekend without me.

As a look back on this season, here are the players that were selected in the first round and their distribution of fantasy points scored over the first fifteen weeks of the season. All told, there are seven running backs, two wide receivers and one quarterback.

And here are the ten players who currently have the highest average fantasy points per game. Among this group there are eight quarterbacks and two running backs.

In fact, only three of the top ten draft picks ended up in the top ten of average points scored per game!

This is only the second year I’m participating in a fantasy football league, but both times I’ve gotten the general sense that running backs are the most valuable players on any given team. The reasons seem to make sense: teams usually have one main running back who will have the ball on roughly half of the offensive plays, and the running back is more likely to be trusted to score when a team is closer to the goal line. I was lucky to have had the number one pick in my league this year so I went with what was the best fantasy player by near consensus: running back Adrian Peterson.

Seven weeks into the season though, the numbers don’t really add up. I took a look at the week-by-week performances of the top five quarterbacks and top five running backs (by average weekly points earned). Barring unforeseen circumstances, these ten players will definitely be in the discussion of first round draft picks in next year’s fantasy football season. Here are the distributions:

The top five quarterbacks have performed significantly better this year. On average, those quarterbacks earn six to seven more points each than the top five running backs. More telling though is the fact that 75% of the time, these quarterbacks do better than the median score for the top five running backs. Next year, my first pick is going to be Aaron Rodgers.

Last week, I wrote about how ESPN’s Fantasy Football projections were not particularly accurate in my league. That has not changed after the week five match ups. Accurate, however, is not the same as useful. Even if the lines given by ESPN do not predict well the final margin of victory, they may still offer some valuable information for decision making.

For example, how often does ESPN project the correct winner? Out of 25 match ups so far this season, ESPN has picked the correct winner 19 times (76%). Not bad. In the scatterplot included here, I’ve colored the incorrect projections red and marked them differently.

An interesting follow up question is whether or not ESPN gets better at projecting the winner as the projected margin of victory increases. You would think so because the margin of victory can be thought of as an indicator of the relative strengths of both teams. We probably need more data before drawing any conclusions, but in five weeks, ESPN has predicted incorrectly when the line was 7.2, 9.5, 12.1, 17.7, 38.3, and 46.9! This week, my team is projected to win by 31.5 points. A nice margin, but definitely not a guarantee.

Each week, I base a good number of decisions regarding my fantasy football team on ESPN’s projections of how well my players will do. Is this a good idea?

I decided to take a look at ESPN’s predictions of team to team match ups (fantasy teams, not actual NFL teams). In JMP 9, I plotted ESPN’s projection of the difference in final score (called the “line”) with the actual margin of victory at the end of the weekend. The NFL season has been going for four weeks and there are ten teams in my league (five games per week) so there are 20 data points to look at.

If the projections are accurate, we expect only small deviations between ESPN’s line and the margin of victory. There should be a positive association (the larger the line, the larger the margin of victory), and the slope of a linear regression line should be close to 1.

However, we see only a weak relationship between ESPN’s line and the margin of victory. In fact, with a p-value of 0.4028, we can’t even say we’re confident that there is any relationship at all!