First of all, hi, I'm new. One of my interests in fantasy football is in statistical analysis, so this post is going to be somewhat technical. I was thinking of a way to analyze what position to draft first (or second, third etc.). Anyway, this is what I came up with. If this type of analysis exists somewhere, please point me towards it.
Rather than trying to predict how well a player will do by looking at his previous years stats, match ups, etc. (this type of prediction is really hard and prone to large errors), it might be useful to predict a players performance by his ADP. For instance, by ADP Arian Foster is the number one RB this year. Historically, how have #1 RBs done?
For each of the past 5 years I assembled players' preseason ADP and their in season stats. To help eliminate injuries skewing the data, I looked at per game fantasy point production, rather than whole season point production. For each ADP spot (by position, so QB1, QB2, QB3 etc., same for RB, TE and WR) I calculated the median points per game for the past 5 years (I used the median to eliminate particularly bad or good seasons, which would show up in an average). I then plotted ADP versus PPG, and fit a straight line. So the equation for this line gives me median PPG as a function of ADP for each position. I can than use this equation (a different equation for each position) to predict, for instance, how many PPG the 5th ranked TE will get.
So after doing this, I got this year's ADP, and wrote a program to simulate an n-owner snake draft (my leagues use 12 players). By default, owner draft by highest remaining ADP (assuming the team needs that position). I then used the above equations to calculate the PPG for each team. No surprise, the 1st team to draft winds up with the best team.
This score, by draft position, I used as a baseline score. So this is the score you'd expect from a certain draft position by just drafting by ADP, assuming everyone else does the same. Then one at a time, for each draft spot (1 through 12), I used a different draft strategy and computed the new score to the baseline. Some of the strategies I tried was RB, RB RB, QB RB, WR WR, etc.
As it turns out, RB RB was the best strategy, outperforming the baseline for every draft spot (except those that were already drafting RB RB just by ADP). One spot (10th I think) outperformed the baseline by a full 5 PPG. And this was with 6 fantasy points for a passing touchdown. So even in a league where QBs dominate the top scoring spots, RB RB still does better than drafting QB early.
Wow, that was long, Hopefully people that are interested in this sort of thing could follow that. Anyway, a few questions: Have you seen an analysis like this before? Do you have any suggestions on how to improve this / make it more useful / make it more accurate? How can I account for injury rate (I assume QB is less likely to get injured than RB and WR, an effect I did not account for). One thing that I think is particularly elegant about this analysis is that it naturally takes into account the idea that the top QB is more likely to finish near the top of QBs than the top RB is to finish near the top of RBs, since it compares ADP to actual in season production.
Yeah first post!
Rather than trying to predict how well a player will do by looking at his previous years stats, match ups, etc. (this type of prediction is really hard and prone to large errors), it might be useful to predict a players performance by his ADP. For instance, by ADP Arian Foster is the number one RB this year. Historically, how have #1 RBs done?
For each of the past 5 years I assembled players' preseason ADP and their in season stats. To help eliminate injuries skewing the data, I looked at per game fantasy point production, rather than whole season point production. For each ADP spot (by position, so QB1, QB2, QB3 etc., same for RB, TE and WR) I calculated the median points per game for the past 5 years (I used the median to eliminate particularly bad or good seasons, which would show up in an average). I then plotted ADP versus PPG, and fit a straight line. So the equation for this line gives me median PPG as a function of ADP for each position. I can than use this equation (a different equation for each position) to predict, for instance, how many PPG the 5th ranked TE will get.
So after doing this, I got this year's ADP, and wrote a program to simulate an n-owner snake draft (my leagues use 12 players). By default, owner draft by highest remaining ADP (assuming the team needs that position). I then used the above equations to calculate the PPG for each team. No surprise, the 1st team to draft winds up with the best team.
This score, by draft position, I used as a baseline score. So this is the score you'd expect from a certain draft position by just drafting by ADP, assuming everyone else does the same. Then one at a time, for each draft spot (1 through 12), I used a different draft strategy and computed the new score to the baseline. Some of the strategies I tried was RB, RB RB, QB RB, WR WR, etc.
As it turns out, RB RB was the best strategy, outperforming the baseline for every draft spot (except those that were already drafting RB RB just by ADP). One spot (10th I think) outperformed the baseline by a full 5 PPG. And this was with 6 fantasy points for a passing touchdown. So even in a league where QBs dominate the top scoring spots, RB RB still does better than drafting QB early.
Wow, that was long, Hopefully people that are interested in this sort of thing could follow that. Anyway, a few questions: Have you seen an analysis like this before? Do you have any suggestions on how to improve this / make it more useful / make it more accurate? How can I account for injury rate (I assume QB is less likely to get injured than RB and WR, an effect I did not account for). One thing that I think is particularly elegant about this analysis is that it naturally takes into account the idea that the top QB is more likely to finish near the top of QBs than the top RB is to finish near the top of RBs, since it compares ADP to actual in season production.
Yeah first post!
