I quoted my post because I want to expand on what I said there. I just got done with what amounts to about a month of work on this, not full time of course, but still a ****load of hours so I was kind of excited to finally have the results in front of me.
This figures to be a long *** one, as I will explain the whole thing and how *I think* it relates to team building.
IMO the salary cap is the actual amount of talent you're allowed to have on your team. At least theorically. And the number of picks you get per year is your pipeline for adding talent. In order to quantifiy the value of a draft pick, we need to connect the 2. This was the main idea that drove me to start this research. What is the value of a draft pick?
Expected value is a well known term,
And so I took 30 years worth of draft picks, classified them into 4 categories: Elite, Good, Jag and bust. This part was a grind, I tried alot of different ways to get those right, at some point realized that no matter what metrics I'd use for different positions, the ratios would change very marginally.
Then it was all a metter of getting odds for every category and positions for each draft picks. We cant use the raw data by itself, we need to smooth out the curves a little bit. So I built a multinomial logistic regression for that part. This is how I got the % used for the EV formula.
Then I needed the outcome, which is player salary. Here we assume that money is directly associated with talent. Which can be flawed on a case by case analysis but seems to hold true when used on a whole population. All I needed was the average salary for 4 different groups[elite, good,jag,bust] and around 10 different positions.
Then its just filling up the EV formula. I substracted the rookie salary associated with each pick to result. The slope of the rookie salaries was simply to steep to ignore, had it been more linear, I might have just ignored it.
Now what I think it all means?
Those results pretty much confirmed what I originally thought...
- QB obviously
- T theorically I know it makes sense but I was somewhat surprised it actually came out 2nd
- WR again, Im surprised, but this just pretty much follows the theory that the passing game is king
- Bunch of positions really close to each other
- RB, TE and interior OL basically undraftable at the top of the draft. Does that really surprise anyone?
Now to elaborate on my previous post...
For a team starting from scratch, we could view the actual expected value of draft picks for that team to be the average of all of the positions available to them's expected value.
This average EV is shown as the red line on the chart
The average EV when a team already has QB sloved is the yellow line on the chart
The average EV when a team has both QB and T solved is the green line
View attachment 35374
Now nevermind that those positions are QB, T and WR in that order, the real takeaway here is that by nailing higher value positions, you are lowering the actual value of a draft pick to your team. While this might sound counter intuitive, this is actually a good thing. We've all heard the saying getting value from the draft, IMO the way to get value from your draft picks is to get players who have more value than the pick you used to get them.
Sure as a GM, you are looking at this problem and trying to fix your end of it, get better at picking good players... But fundamentally, I think trying to increase the value of the players you pick might just be thougher, and more out of your control, than just trying to nail high priority positions first, thus lowering the overall value of your draft picks, making it easier for you to actually get value.