As most of you know, the projection model is my baby, my long-term project, and sort of my niche in the prospect valuation field. I built it originally a few years ago, but at the end of each NFL season, it gets a like wash, rinse, and wax job for fine tuning. With more data in, the formulas need updating. One thing that happens each year is another entire draft class is added to the fold. I wait three seasons of NFL action until I include those former rookies into the formulas. For the 2015 version, that means the 2012 class is finally in the mix. So the player pool used in the formulas is now from 1999-2012 at running back, wide receiver, and tight end.
Another key adjustment going forward is the calibration I use for each comparative statistic. I used to use a generic ‘ranking’ function, which basically orders every prospect in every category from the best to the worst. One perk of that method is that the scoring can be most-easily digested with the static 1-100 scale and it keeps all the larger categories within the same lines.
Why Change?
I did some testing over the past week after asking myself this question: Why penalize the true standouts in a certain drill or statistic by giving them a score of 100, while the second-best option, who may be significantly lower a 99? The scoring can, and will, remain on a 1-100 scale through normalization in each category, but the average will not be 50 overall and the scaling will not be step-wise from 1-100. In short, prospects will be credited and penalized within their metric resume more accurately than in the past. That is a great thing as the predictive strength through regression studies have bumped up as a result.
How Do I Read Projection Model Scores and Data Going Forward?
A ton of the rookie profiles I write will be based upon the projection model scores and data. I use the model extensively for the younger, yet-to-emerge NFL talent pool as well. In general, the average marks in each category are lower. Taking a stroll through the positions already complete, larger categories like ‘Size’ or ‘Athleticism’ or ‘Production’ still operate on the 1-100 scale, but the aggregate average for the player pool is in the low-t0-mid 40s. Why? Because the top players now see more benefit and the bell curve feel to data set is more accurately represented. The overall player scores match that sentiment as well.
In addition to referencing raw scores for a player, category, or drill in the coming weeks and months, I will also give context in terms of the general field in the category. Phrases like ‘20% better than the average’ or ‘in the top-20% of the category since 1999’ will be far more common than in 2014. In general, the ‘average’ player will appear lower in terms of their raw scores and a ’60’ is better than it looked in the model a year ago. I will also reboot some ‘historical look’ posts at each position to clarify some of these overarching tweaks. In summary, the model is better than ever and ready to mine some rookie value in 2015 for startup and rookie drafts. Let’s build some dynasties!
