If I had unlimited time and resources on my hands I am certain there is one thing I would spend a large chunk of infinite (OK, you can't really have a large chuck of infinite b/c anything/infinite=effectively 0) figuring some sort of computer model that would tell a fantasy football drafter what player to select when their chance came up based on all past evidence. I recall attempting to do this once for a few hours with another math loving buddy of mine, and after my mind was completely boggled I gave up and watched TV. That does not mean of course that we can't test some of the commonly held logic used in fantasy football drafts mathematically.The #1 general rule in any fantasy football draft is that running backs are the most important thing that you must draft. Most drafts will feature an entire first round of running back selections (with the exception of an occasional QB or WR), and nobody every really questions that if it is the smart thing to do. Everyone can give you a reason as to why they do it (hard to find good RB's late, there consistent players, b/c since everyone else is doing it I have no choice), but are there any strong mathematical reasons behind the thought process. Lets look at the numbers:
| QB | Points Total | Standard Deviation |
| Peyton Manning | 358 | 10.35 |
| Drew Brees | 302 | 11.69 |
| Michael Vick | 286 | 10.80 |
| Marc Bulger | 284 | 11.62 |
| Carson Palmer | 282 | 9.63 |
| Tom Brady | 250 | 8.63 |
| Phillip Rivers | 240 | 8.64 |
| Jon Kitna | 236 | 9.77 |
| Eli Manning | 226 | 8.54 |
| Donovan McNabb | 216 | 11.11 |
| Vince Young | 212 | 9.73 |
| Brett Favre | 212 | 10.34 |
| Average | 258.66 | 10.07 |
| Average-PM | 249.63 | |
| Top-Bottom | 146 | |
| 2nd Top-Bottom | 90 | |
| StanD | 44.57 | |
| StanD-PM | 33.30 |
| RB | Points Total | Standard Deviation |
| LaDainian Tomlinson | 420 | 12.80 |
| Larry Johnson | 318 | 10.07 |
| Steven Jackson | 316 | 10.01 |
| Frank Gore | 250 | 8.03 |
| Willie Parker | 246 | 10.68 |
| Brian Westbrook | 238 | 9.29 |
| Tiki Barber | 220 | 9.82 |
| Maurice Jones-Drew | 200 | 8.75 |
| Rudi Johnson | 196 | 6.54 |
| Deuce McAllister | 176 | 6.21 |
| Marion Barber | 176 | 7.42 |
| Joseph Addai | 172 | 9.92 |
| Average | 244 | 9.13 |
| Average-LT | 228 | |
| Top-Bottom | 248 | |
| 2nd Top-Bottom | 146 | |
| StanD | 74.44 | |
| StanD-LT | 52.12 |
| WR | PointsTotal | Standard Deviation |
| Marvin Harrison | 198 | 8.80 |
| Terrell Owens | 192 | 5.89 |
| Reggie Wayne | 184 | 7.52 |
| Lee Evans | 176 | 9.47 |
| Donald Driver | 176 | 6.66 |
| Torry Holt | 170 | 8.11 |
| Chad Johnson | 181 | 10.79 |
| T.J. Houshmandzadeh | 171 | 6.55 |
| Javon Walker | 170 | 9.28 |
| Steve Smith | 166 | 7.86 |
| Roy Williams | 166 | 7.16 |
| Plaxico Burress | 152 | 5.89 |
| Average | 175.16 | 7.83 |
| StanD | 12.40 | |
| Top-Bottom | 46 |
To gather my data I relied on what was readily available to me, the NFL.com scoring point system and results. For those not familiar with the NFL.com scoring system, it rewards 1 point per 10 rushing/receiving yards, 1 point per 25 passing yards, and 6 points for touchdowns. It deducts for fumbles and interceptions. It is a rather standard league, but if you would like the full details they are available here. In the above charts I took the top 12 players (12 is the number of teams in an NFL.com draft) at QB, WR, and RB according to NFL.com from last year, gave their total points, the standard deviation of their weekly points, and other break downs on their numbers. So what do all those above numbers tell us?
First of all, from the data above we can see that QB's in the Top 12 score slighly more points than the RB's even if we allow the RB's to keep LaDanian Tomlinson and we remove Peyton Manning from the QB's. It is not a big margin though. What is quite clear though is that WR's score much less on average than the other two possible first round position possibilities. That is one strike against taking them in the first round. Does the fact that the QB's score slightly more than the RB's mean we are wrong to put our focus on the halfbacks so early in the draft? The others numbers suggest the exact opposite.
The RB's came in as slightly more consistant than the Quarterbacks when a standard deviation was taken for their week-by-week totals was taken. A QB supporter could chalk this up to the fact that the QB's score more points so they naturally have a wider possible range, which could be a valid point, but there is much more evidence still left to consider. The standard deviations amongst the top 12 at each position even when taking out top players from each side or allowing one side to keep their top player are massively different. With LT factored in the standard deviation of the RB scores is 30 points higher than that of the QB's, and even with him out the RB's still remain 8 points higher with the QB's allow to keep Peyton Manning (there top performer). What this means is that each of the quarterbacks is much closer to an average score for their positions than are the RB's. Every single RB picked before you will result in more points lose of the course of a season than will each QB picked. Even more evidence comes to support this when you look at the #1 or #2 QB or RB vs. the #12 QB or RB. The difference between Manning and the 12th QB is only 146 points compared to 248 points for the RB's (LT to #12). Even if you throw out LT's amazing year (you can't plan a draft around getting LT) the RB's still match the QB's (w/o Manning the QB's would be back behind 50 points again).So, the idea that running backs are more consistent than QB's proved true last year (I cannot conclusively say this is always true based on the data I've collected so far) and the drop off from RB to RB is largest than that among QB's so getting the 12th back is much bigger lose than getting the 12th QB. With those facts in the book though, should a WR, who we already know score lower than these other two substantially, even be considered? According to our data, the answer is a definite no. WR's week-by-week come in as more consistent than RB's and QB's by standard deviation (7.83), but once again this could be due to the fact since they score less their possibility of score ranges is smaller.; however, the drop off from the #1 to the #12 WR or anything inside of that range is simply not a very big deal compared to the other positions discussed. Their total score standard deviation was pitifully small compared to RB's and QB's (12.40) and their top minus bottom was only 46 points meaning that there is not a big comparative drop-off among the top 12 wide receivers like there is at the other positions. Unless you've got a time machine and know a WR is catching 20 TD's to go along with 1500 yards in the next season it just would never make sense to think of taking a WR extremely early in the draft (1st round is out of the question).
This study is by no means comprehensive and even on this year's data there are many more observations that could be taken on it. It does seem from what our current data situation tells us though that the "RB, early and often" strategy is a mathematically sound means of determining who a person should draft. It would be wonderful if anyone reading this would like to pick up the ball and run with this idea to test is further as I would like to. In the coming weeks I'm going to try to at least gather more data and put it on the site even if I do not grow each post into a full length article. There is so much luck in fantasy football that any reason based edge a player can gain is huge boost to their cause (especially when money is involved), so I feel this information is truly valuable. For now though, if you want to become the uber-nerd of your fantasy football draft party, when someone selects Terrell Owens in the first round of your local draft, whip the good old spreadsheet out and explain to them just why they are going to finish in last place this season.
