Actually it is dumb given the number of measures you get to estimate any one player's athleticism. The assumption that every player's combine shuttle times etc. are perfect indicators of that player's true ability is just bad statistics. Perhaps if you had the player run a 40 or whatever on several different days under different conditions, you'd get something reasonable but basing analyses on a single measure like this is doing completely ignore concepts of variability

There is nothing dumber than not critically evaluating information. Yes, someone gave you some numbers. Just suck that down and don't ask any questions.

What do you mean whether that is an indicator of a "player's true ability?" It is a proxy for their athletic ability. Not a substitute for their game tape. Nothing dumber than not being able to see that distinction.

That's not my argument at all. Basic statistics. A single combine 40 time will likely vary from a player's average 40 time. Two thirds of the time you'll be within a standard deviation of their average - but you've got no idea how much they vary from 40 to 40, so largely you've got no information. Taking single measure and using it to represent their true 40 average is an incredibly noisy measure. This gets compounded by using a ton of other messy measures on other performances. Yet it gets used in a formula like we should have confidence in the value.

You realize a player doesn't run just one 40 right? Between the combine and pro days I feel comfortable in a players tier for the type of speed they have. And even if we had just a single data point. Let's say a wr ran a 4.3. That's good enough because even if he's a bit slower on average you have an idea of their high end. You can't luck into a 4.3 due to standard deviation error. That's not how it works. Silly argument.

Unless the player is running 30 times and you are averaging those times, you aren't getting anything estimate that comes with any precision. Wow, a "standard deviation error" - sounds like you are using words you don't really understand. It is entirely possible for a 4.5 guy to put up a 4.3. If he averages a 4.5 but his runs have a standard deviation of 0.1 seconds, he'd hit 4.3 once every 20 times. That's not a particularly rare outcome. What you call his "high end" is, for this player, something he'd bring to just a handful of plays a game.

No. No. No. A guy who runs a 4.7 as his precise and accurate measure of ability isn't by chance going to run a 4.3 due to normal statistical variation. And no I don't need an N=30 in this case. And what do you mean just bring to a "handful of plays?" We are approximating a player's athleticism. That is all and comparing it relative to players of his draft class and historical averages. This is a simple and useful exercise. Oh and I have an engineering degree and a MBA from a top 10 US institutions. I know statistics.

That completely untrue. You have no idea how much variation that player shows in his runs. It is statistical ignorance to make that claim in any manner. Hahahaha a business student think he know statistics - that's consistently the worst department on any campus to receive such instruction.

I received a B.S. in Engineering. I received an masters in business administration. At a top 10 institution in the U.S. I currently work at the most prestigious investment bank in the world but yeah...I don't know stats. And again if you run a 4.7 as a precise and accurate measure as I said earlier...you're not magically waking up and running a 4.3 based on statistical variation. That's a joke.

I have a Ph.D. in statistical methodology. Please don't think the handful of courses you took make you qualified to argue here. I've spent the last 25 years of my life doing statistics every day. Typical arrogant MBA.

Typical arrogant *wealthy* MBA. And where did you get your doctorate? Where are you now? Anyhow please tell me how by statistical variance someone who has the true ability to run a 4.7 may trip fall and land on a 4.3 just by statistical happenstance?

I went to a university that is well regarded as providing excellent applied training in quantitative methodology. I am a professor at a well regarded public university. I've taught statistics for over 25 years and use statistics extensively in my scholarly work. In fact, I've written books and journal articles on a number of statistical topics (as well as many others using complex statistical methods). So please don't compare the one or two stats courses you may have taken as equivalent. A players whose true ability is 4.7 would naturally have runs that are faster and slower. True ability is the long term average for that player's 40 times. If this particular player were particularly inconsistent - say producing a standard deviation of 0.2 in their runs then, as I noted, 1 out of every 20 would hit 4.3. Problem is, you don't have enough measures to get an estimate of the standard deviation. More generally, point estimates are completely useless.

The problem with what you're doing (typical of those who teach in a field rather than act in one) is that you're applying a statistical profile..in this case a normal distribution...to a physical situation in which it doesn't apply. There are physical limitations to an athletes straight line speed and acceleration. This isn't a stock that moves at random or an electron that can be anywhere and everywhere within a cloud. Typical. Now as to this conversation there are certain tests at the combine, even without an n=30 sample size, that have proven moderately predictive. That information is worthwhile. There is nothing more dumb in ignoring it.

A physical situation does not apply to a normal distribution. REALLY? There are literally millions of physical and performance attributes that do follow normal distributions. The idea that an athletes speed does not move around is just plain laughable and indefensible given the many aspects that influence performance on a minute-to-minute basis.

I said an athletes speed doesn't move around? No, I said that a 40 time for an athlete isn't likely normally distributed and you're not getting a 4 one hundredths of a second variance just by statistical happenstance. Do you know how big 4 one hundredths of a second is? It's the difference between you being Dez Bryant vs Tyron Smith.

I suspect that's exactly correct. The Force Players by Mosqueda mentioned that he was mid tier because he didn't do all the drills. And I'd imagine an algorithm like SPARQ doesn't return a useful answer if you leave out two data points.

30 times.....hate to be in your combine. 'Hey Mr. Larry Allen, I need you to run the 40, 30 times to get an accurate number.' Then after he hits you, you wake up and it wasn't really that important anymore. The draft isn't really high on precision, also makes the draft unpredictable. Could it be improved, yes, but the combine would last a month or two if was that precise.