A Brief History
You’ve probably seen it mentioned a few times since my arrival to Pride Of Detroit, and it’s possible you saw it previously from my work on Twitter. I’m talking of course about Relative Athletic Scores, or RAS (pronounced however you like, I like Raz, like Raspberry). You know that it’s a number, and you could probably figure out that it’s a 0 to 10 scale, which is all well and good if you know what it means. Do you know what it means? That’s what we’re here for today, to explain how the metric is calculated, how it works and why it has enough statistical importance that I make a point to blow up Twitter every time I can with those little black, green and red tables.
I’m going to start with a brief history lesson. Once upon a time in Texas, 2012, I was talking with some friends online about what else but football. It was draft time and the NFL Combine was on. The Lions, of course, needed a cornerback and offensive line help because this is the 2000s and they’ve always needed those things. We were arguing the merits of players like the “ridiculously athletic” Morris Claiborne, or the “slow and unathletic” Alshon Jeffery. Now, for full disclosure, I liked neither of these players pre-draft, though for different reasons. That’s not important, but what is important is that at some point we started looking at how “explosive” these guys were. Holy crap, man, Mo Claiborne had a 34.5 inch vertical! Dude can just explode from his zone to break on a receiver!
You may notice some of the words I put in quotes because you’ll hear things like that every draft season, or some variation thereof. What I started to realize is that those terms don’t really mean anything. That 34.5 inch vertical for Claiborne that was being used to show how explosive he is? That’s below average for a cornerback. That slow and unathletic receiver? He measured above average in almost every category (he had a weirdly slow 10 yard split). All these terms we hear talking about player athleticism and they meant pretty much the same thing, which is nothing.
To make matters more confusing, I would later find out, is that the measurements mean even less when you don’t have context. My mother loves football. She’s the one who introduced me to the game. Love it as she does, she doesn’t understand anything about the draft and has never been interested (read: Obsessed) enough to get into detail with it. In a conversation with her about the combine, I switched from talking about cornerbacks and mentioned how USC’s Nick Perry ran a 4.64, but aside from speed there wasn’t much else to him. It was an offhanded comment, but my mother (who has the attention span of a goldfish) only half heard me and thought I was still talking about corners. She had heard me talk about Stephon Gilmore running a blazing 4.4 flat, so she got confused when I said Perry’s 4.64 was fast. After some explaining about the difference between what those times meant at their respective positions, I got this idea.
Wouldn’t it be great if we could just put a number on these measurements that would show the context right in the number? Like you could see a score for a 40-time and know whether it was good or bad for their position? From that simple idea, Relative Athletic Scores was born.
I decided very early on to put it on a 0 to 10 scale. Why 0 to 10? It’s a neat little trick used in statistical polling. Give someone a scale of 1 to 10 and they will sometimes confuse what constitutes good. Is a one good or is a 10 good? Am I rating this correctly? Change that scale slightly so it’s 0 to 10, and it is inherently understood that zero is bad and thus 10 is good. Using this scale offered that context I wanted to provide. I wouldn’t have to worry about whether a score was above or below average anymore. If I saw a guy had a 6.5 out of 10, and 5 is average, then it’s obviously above average. Math!
Once I settled on what I wanted the system to look like, I needed it to be functional. If this were a movie, there would be some kind of a montage here. Probably set to some upbeat 80s classic with motivational lyrics. Ooh, or maybe that “She Blinded Me With Science” song by Thomas Dolby, that would be a good one. Math and Science work together, that would fit. You’d need the music because the actual action at this point was pretty boring. Trial and error until I found a system that would rate the players how I wanted them and also fell on a 0 to 10 scale. Fast forward to late-2015 and I had the calculations perfected. So plot twist, the montage wasn’t very exciting but it covered almost four years of work. Way to go out dated pop culture references!
So how does it work? RAS is essentially a ranking system, so it roughly correlates to percentile. While I always say that 5.00 is average, that doesn’t quite describe it accurately. A better way to describe that 5.00 middle mark is that it is the score for the average player at that position. Shouldn’t that be the same thing as an average, you ask? Why no, no it is not. It’s an oddity of studying these measurements so much, but in most cases the actual mathematical average (the mean) tended to be a good deal above what the average player at a position would score. So for instance, a player who had an actually average score may end up with a 7.50 for that measurement. That wasn’t what I wanted at all. I used an incredible amount of math to get to the current system, you guys, and I don’t mean incredible as in “wow, that’s amazing!” but more in the “wow, that’s overly complex and cripplingly unexplainable” kind of way. Using the calculations as they are now, we get to see what the average player at a position scored at each measurement (closer to a statistical median than a mean).
Now it’s pretty simple. The actual numbers correspond loosely to percentile, so a player with a 9.87 score for his 40 time managed to be in about the 98.7th percentile of his position group for that measurement. The final number — the one we actually refer to as the player’s RAS — is gained by averaging the individual scores for each player at a position. This raw average is then compared to the raw averages for every other player at the position to come up with the final Relative Athletic Score for a player.
It isn’t a perfect system, and like the measurements themselves, there are plenty of outliers. What it gives us, though, is a way to put a number on a player’s measurements as a whole when compared to several hundred other players at their position over the past 17 years (2000-2016). It’s a lot of data and I’m continually adding to it. Adding more data does affect existing scores, but not very much. After adding over 1,500 players this offseason, the biggest change was only about 4%, with most scores being affected by about 1%.
Because a player may not have completed all measurements, I set the minimum required to calculate a score at six. Why six? I didn’t want any particular area to take up more than half of the score. At most, speed takes up about 30 percent of the total score, but if we only had five measurements and three of those were the 40-yard dash and its splits, we would end up with speed covering about 60 percent of the score. That was unacceptable. With a minimum of six, speed can only ever take up half the score, not even that when considering the 10-yard split (more later).
Relative Athletic Score Weighting
To break it down further, RAS is calculated from 10 different measurements, each of which account for 10 percent of the average. Height and weight make up the size portion, accounting for roughly 20 percet of the score. The 40-yard dash and its splits of 20 and 10 account for speed, about 30 percent of the score. This is slightly misleading because the 10-yard split is also used as a measure for explosiveness. Depending on your leanings in that regard, speed could be considered anywhere between 20 and 30 percent, so let’s just call it 25 percent.
Bench is a measurement of upper body strength, which only accounts for 10 percent of the grade. Positions like wide receiver and quarterback rarely even bother with the bench, but we still use it where available until there is something better. Vertical, broad jump, and the aforementioned 10-yard split account for explosiveness. Using the same logic we used for speed, we’ll call it 25 percent. The 20-yard shuttle and three-cone drills measure hip and ankle bend and agility, the final 20 percent of the final score. It ends up looking like this:
Why is it Useful?
Ah, the big question. What makes RAS different than any other measurement, including the ones it represents? Why is it any more useful than just using the base measurements? That answer is what has made me so excited in its application, and also why I can’t post the whole gamut online anymore. See, most statistical representations in the NFL, in fact most in general, can be represented on a bell curve. Bell curves show what is known as “Normal Density”, and a majority of those measured will tend to fall closer to the average. The further you are from average, the fewer players you will encounter with scores out there. When represented on a table, it gives you that nice bell shape.
Each of the individual scores in RAS, for 40 time, weight, bench press, etc. will also end up on a bell curve when put in chart form. This makes sense since the individual scores are just a different way of looking at the original data. The final RAS, however, does not follow Normal Distribution. Once RAS is calculated fully, the scores are equally distributed from 0 to 10. What that means is that if you have 100 players, you would have 10 from 0-1 RAS, 10 from 1-2, 10 from 2-3 etc. all the way to 9-10. Since it’s equally distributed, we can get a much clearer picture of trends since there is no longer a need to account for volume. There will always be exactly the same number of players above 5.00 as there are below 5.00 for any position. So when I say something like “defensive tackles rating 5.00 or above make the pro bowl FIVE TIMES as often as those rated below 5.00,” I’m speaking of sample sizes that are equal in volume. It’s tough to find in statistics, and it’s valuable for football.
So that, in a nutshell, is what Relative Athletic Scores are, how I came to this metric, and why it is valuable. Next time, we’ll take a look at some high level statistics, showing the correlations between different positions, RAS and NFL success by various measures. The stat I quoted above about defensive tackles is true, for instance, which is why defensive tackle is one of those positions I believe measurements are very important for. Are there positions where measurements don’t really show much meaning? I’ll show you what I know sometime soon, but in the meantime give me your predictions in the comments of which positions you think measurements mean the most for and for which you think they don’t matter at all.