Mullayo's metrics approach explained

Here is a blurb taken from 2015 that explains the thinking behind the formation of the OL athleticism metric of today. RAS wasn't around and is still far too simplistic anyway, being basic percentile aggregation. The metric is maths-only neutral and grades the players according to positional thresholds across all measureables.

------------Feb 2015-----------

Last year, I analysed our O-line measurables and found commonalities in stats that must be desirable under our particular zone blocking scheme. Last week I decided to take it further and make a metric that matches our values so I could project our potential persons of interest using this coming Combine's data.

Our team ZBS (and probably all others) values length, cleverness and nimbleness to control whole areas and make in-game decisions, which correlates to high scores in the Wonderlic, arm-length, short-shuttle and three-cone stats which are usually above average. Also, hand size is important for me because it correlates with natural size and power which leads to a higher ceiling and eventual production. Combine bench press is also important because it shows the athlete's application, especially when coupled with the natural disadvantage of long arms. Eventually, weight room exposure can increase the strength but a decent score shows either natural power or dedication, which are both plusses.

Even weight is a positive, a 332lb beast like Greg Robinson having a 1.69 10-yd split is much more impressive than 302lb OC James Stone's 1.72. Afterall, F=M x A. So that was given an albeit reduced value compared to other stats, although it would be more significant for a power blocking scheme.

I was going to make a simple equation adding the positive attributes (ones where a high score is good) then dividing them by the sum of the negative attributes (ones where a high score is bad) until I realised stats are not all created equal and each attribute needed to be weighted against the others to accurately reflect our scheme's values.

Also, each attribute needed to provide a suitable range between good and bad, (with exceptional outliers adjusted for). The more important the stat, the bigger the variance should be—leading to more separation in the final answer. To do this I needed a bigger data set than just our players and free agency interests so I imported the entire OL class of 2014 - tackles, guards and centers into Excel—which I am learning to use this week, solely for things like this.

I kept the overall metric simple by making a formula for each individual Combine stat, so I could tweak them individually to get the desired range between top and bottom scores and overall weighting against the other stats while still affecting the overall result.

Thus, the positives in order of importance (and therefore assigned the greatest range and weight) are: Arm-length, hand-size, bench press, weight*.

Meanwhile the negatives in order of importance are 3-cone, short shuttle, 10-yd split and 40yd dash. To get the ZBS value result, you simply add the now weighted positives together and subtract the total weighted negatives.

Then for optics I normalise, by multiplying the result by 50 to give it a score out of approximately 100 with 90 so far being the top and 51 being the bottom. Under 50 could be construed as sub-NFL.

Generally tackles should get higher results because they are generally the better, more sought after, athletes and consequently higher paid. So even within my results, more leeway should be given to guards and more again to centers - i.e. their scores can have a lower floor.

Also, scores are affected by outlier stats, which should be taken into account. AND of course on field play (intelligence/instincts) is the ultimate disclaimer. There is no accounting for idiots. The main use this metric has is to sort the wheat from the chaff in terms of who deserves a closer look at game tape based purely on their athletic physical profile compared to the positions' thresholds.

After entering the data, I was pleasantly pleased to discover most of my favourites dominated the scoring – entirely incidental, as I worked out the weighting and requisite range on paper before tweaking the individual stat formulas.


Ryan Groy, who I believed we should chase as an UDFA before my metric comes first overall with 90. Maybe he's a bonehead or injured, but the stats back him as someone to bring in. Stephane Milhim who was/is on our practice squad is another phenomenal prospect who maybe can't play but was worth the look with 87.9.

As were some of my draft favourites Jake Matthews, Trai Turner and Joel Bitinio all up the top end. Although others I liked as players Dakota Dozier and Chris Watt didn't fare well at all.

My analysis fave FAs: Evan Dietrich-Smith (89.1), Jon Asamoah (88.9) and Alex Mack dominate as well. Mack's 86.6 was held back by a 20 bench which is weird when he was considered really strong coming out.

Two centers my initial research told me fit our scheme, Gabe Ikard (80.4) and Wesley Johnson (81.4), also score well.

Jags are mostly in the upper half, led by Luke Joeckel (83.7 albeit a bit behind the better athlete Jake Matthews 89.9), Josh Wells and Brandon Linder, then Sam Young, Austin Pasztor, Luke Bowanko and Tyler Shatley (71.3). Bearing in mind most of them are guards that's great scouting.

The icing on the cake is Will Rackley (67.2), Zane Beadles (57.8) and Mike Brewster (57.5)rank amongst the worst. In Beadles defense, he was never a dynamic athlete, has a phenomenal Wonderlic and has 5 years in a NFL weight room but was still a strange choice. In Rackley's and Brewster's defense - no comment.

Note: I filled in a few blanks to make the formulas work, but only 10-20 squares which I guesstimated based on the player's other results etc.

However, overall, I'm really happy with it. It confirms mainly what my eyes told me but holds enough surprises to pull a few gems out and could have avoided a fair few busts. It may need tweaking, but—when you adjust for the odd outlier -the results ring pretty true for me. Post Combine I'll add in the new data and reveal prospective fits for our scheme.

*Height can be neutral, positive or negative. Much more important is arm-length, with the ideal stat being total wingspan and accounting for shoulder width as well.

FanPosts do not necessarily reflect the views of the authors of Big Cat Country or SB Nation.