Hello everyone! Check out my blog here! Let’s get right into the updates.
NFL Draft Picks
This week I worked on a smaller project about the relative value of draft picks. For those of you who aren’t aware, before each NFL season, all 32 teams receive draft picks. With the worst team (by record of the previous season) getting the first pick, the second-worst team getting the second pick etc. This process repeats 7 times over for a total of 7 rounds of picks. (There’s also some additional compensatory picks given to teams when their players leave in free agency or coaches are hired by other teams.) It makes intuitive sense that overall, players picked higher in the draft are generally better than those picked later on in the draft. After all, there’s a lot of very smart people running these teams and they generally can figure out the players with talent.
The question here is what is the relative value of these draft picks. Once again, it makes intuitive sense that the difference in perceived value between the first and second overall picks is larger than the difference between the 200th and 201st overall picks. By the time you get that deep into the draft, players get more and more similar in terms of skill level/talent. We can get a sense of this relative value from draft pick trades. Teams are allowed to trade draft picks just like players, often as some teams want to move up (get a higher pick) to get a player that may not fall to them, and other teams are content to move back in exchange for additional draft capital. Take the recent trade the 49ers made with the Dolphins, trading their #12 pick, 2022 first round pick, 2023 first round pick, & 2022 third round pick to move up 9 spots to #3 overall. The 49ers got a higher quality draft pick, while the Dolphins got a greater number of draft picks.
This trade is just one of many, and with a history of hundreds of these draft-pick shuffling trades, I hoped to be able to generate relative values for all of them. I would do this by having variables for each overall pick and using the trades to create linear equations for each trade. The 49ers-Dolphins trade would look something like this:
Value12 + Value1-yr out 1st + Value2-yr out 1st + Value1-yr out 3rd = Value3
The Packers move up to get Jordan Love last year would look like this:
Value30 + Value136 = Value26
After converting all of these trades into these linear equations, all that’s left is a simple matter of solving all those equations to get relative values for each of them, assuming the last pick has a value of 1. This can easily be done in R or other math programming languages by converting this system of linear equations into an augmented matrix, and row reducing to solve the system. See Link for a bit more detail on the math behind this; the same logic applies just with a much, much, much larger matrix.
As I was about to complete this step, I did a little bit more research into these trades, and stumbled upon Jimmy Johnson’s draft pick value chart. The great Cowboys coach in the 90s created this chart assigning a value to each pick so that he could easily determine how to get compensation when he was trading back or how much to pay to move up. The reality is that this chart is now used by almost every NFL team. The correlation between real NFL trades and the value of the draft picks by the draft value chart is 0.972. Therefore, all my approach would be doing is back solving for this draft value chart, which is far from interesting given we already know the correct answers.
Despite putting in a lot of time but not being able to get out an article about this, I’m still looking forward to continuing work on other articles. I’m still working on the longer article about NBA Gravity that I explored in more detail in this post. Please leave a comment below if you have any questions or if you have ideas for other basketball or football topics you’d like me to explore. Stay tuned for future updates!