A few days ago I posted an average rankings position spreadsheet and as promised I've updated it (added the Bruno Boys' top 100) and added two new analyses to it. First, in addition to averaging all of the rankings, I ran the standard deviation for each player's rankings to determine how much of a consensus there is across all the experts; the lower the standard deviation, the more consensus on the ranking for that player. I then added the standard deviation to the average ranking position to get a better picture of these rankings - this tells me both a player's ranking and the level of agreement on that ranking, which changes the list I posted earlier ever so slightly. For example, Adrian Peterson is now #3 since there's more agreement on the ranking of Brian Westbrook, who is now #2 - so it's interesting stuff and I think the average+consensus gives a more holistic view of the overall rankings.
Second, I did some average rankings within positions; I'm running these off of the same overall rankings that I mentioned last week, so keep in mind that I'm not averaging positional rankings, just the overall top 100 rankings.
Finally, I ran a cluster analysis on the averages+ consensus numbers, too, which is basically just a way of tiering/clustering players. Keep in mind this is just a statistical tool that I used to create tiers; for better or for worse there was no additional human input into the tiering (outside of the initial rankings from the experts of course). I'll be talking about this in more detail in my weekly column for the Bruno Boys' A Librarian's Touch, which will be posted mid-week.
I've saved the full spreadsheet in Scribd again and also as a google spreadsheet this time, but once again feel free to e-mail me if you'd like an excel version. Also, I'm sharing this data with Sports Data Hub, who will load it into their database and Kevin at SDH and I are probably going to write up some thoughts on the rankings in the next week or so. And a hearty thanks to my brilliant husband for helping me run all this data in Stata.
Second, I did some average rankings within positions; I'm running these off of the same overall rankings that I mentioned last week, so keep in mind that I'm not averaging positional rankings, just the overall top 100 rankings.
Finally, I ran a cluster analysis on the averages+ consensus numbers, too, which is basically just a way of tiering/clustering players. Keep in mind this is just a statistical tool that I used to create tiers; for better or for worse there was no additional human input into the tiering (outside of the initial rankings from the experts of course). I'll be talking about this in more detail in my weekly column for the Bruno Boys' A Librarian's Touch, which will be posted mid-week.
I've saved the full spreadsheet in Scribd again and also as a google spreadsheet this time, but once again feel free to e-mail me if you'd like an excel version. Also, I'm sharing this data with Sports Data Hub, who will load it into their database and Kevin at SDH and I are probably going to write up some thoughts on the rankings in the next week or so. And a hearty thanks to my brilliant husband for helping me run all this data in Stata.
Comments
This is good stuff. Thanks!
-Marc
I've been compiling my own average rank list for years, and I've been dominating my fantasy leagues for several years as a result. Well, perhaps it's my obsession in managing my teams through the season that also is a factor. Thanks for saving me the time!
Mike
P.S. Will you be updating this as the pre-season unfolds? My drafts are not until after the last pre-season game.
But then again, if your site gets too popular
I personally find the ADP lists a much better tool for pin pointing value in the later rounds, when players fall due to teams drafting needs and the uncertainty surrounding guys creates vastly differing opinions regarding their impact.
What drew me to yours is your statistical comparison of many of the major websites ADP reports. I consider it a nice cross reference. Maybe I can persuade you next year to roll out a list of 300.
I draft in a 16 team league with 22 rounds, so knowing your late value is crucial.