Main
Research
Results
Email

We have developed a Ruby based draft simulator. The value in having a simulator based on an interpreted language is rapid prototyping. We can change or modify the draft engine at will. We are interested in refining the language used to describe team needs in a draft. We wish to create something small, easily parsed, and adequate to handle a complete draft.

Okay, good news in that we have a running draft simulator. The source and binaries are available here. This is the same program used to make the following 3 round mock draft. We used the top 99 list from the Sporting News as the source of top players, and we used the team needs lists of Mel Kiper to develop the rules files this program needs, as amended by a number of well-reasoned reader comments. This draft includes the newly released compensation picks in the 3rd round as well as the recent Kevin Carter trade and we give it as an example of what this program can do.

To show off the 'what if' capabilities of this code engine, we present another 3 round mock, the sole difference being that Kansas City's rules were altered to have them draft Drew Brees in the first round. Otherwise this mock draft has an identical top 100 and rules set to the original. The ripple effect of this change can be seen throughout the mock.

A new set of results, with the code *just* released, are the development of stochastic methods. These use the variability of scouting ranking to assign a randomness factor to each and every draft candidate. Using this, we can generate Monte Carlo simulations (NUMERICAL RECIPES IN FORTRAN 2nd Edition, Press et al. pp 685-686) of the draft and generate statistically valid models of draft variability. The upshot? We can predict, within the validity of the data (big if, mind you), where a player will be drafted. Here are a couple runs, the first showing draft variability per player, the second showing the odds of players falling to position 37 of the draft. The data sets used are the default ones in the football data bundle on our code page.

One thing I should point out is that we've received a few very nice letters from people giving us needs lists well in advance. Contributors include Heather Hillis for the Cowboys, Keith Collingridge for the Seahawks, Jim Smith for the Panthers, Mike Coppola for the Buffalo Bills, and Matt Walter for the Washington Redskins. I can only say that we're grateful. More so, the recent letters (and usenet comments) of T. Gryn, Larry Cottrill, Kevin Conry, JayCee1943 and Jonathan Myers have been a big help in refining the rules lists used in the 2 posted mocks.

Grayhaired.