Great article and complement to the free agency piece. Positional value seems to be incorporated into team building to a certain extent, but not enough based on all the information available. Teams spend so many resources to find small edges in scouting but they undervalue the market based comparative edges that exist.
For sure. I’d go as far to say that teams should never draft a tier 3 position in the first round for any reason. The upside is too low and the opportunity cost too high.
Loved this. Will say I'm interested in how this relates to both the "there's no such thing as draft steals only draft reaches" wisdom of the crowds-concept and positional variability from draft to draft. Put in other words and in the context of this upcoming draft, if there's an early run how big of a consensus big board discount would be *too* big when it comes to taking a player in a premium position over (say) one of the TE?
I think the steals/reaches rule still applies for the most part. I'd consider it a reach when we're talking about jumping way up in the positional rankings, not just overall. If you're taking the CB10 over the CB6, that's a reach even if it isn't a low-value position. Just focus on positional ranks more than overall
Great stuff! It's creative analytics like this that totally made the "careers" of sites like FO, OTC and PFF.
It reinforces some ideas I intuitively had and challenges others. But, I think to really get the most meat off this bone, you'd need to combine it with "hit rate", however you define that (for example, I consider 3-4 years as a starter as a hit. Crude, but no less so than AV).
So, what I'm driving at is, that I have always thought late round picks (6-7) should be almost exclusively used on DBs, QBs and OTs, because it "seems" like those positions have the best chance to hit in the late rounds.
Your results on the surface, appear to counter that idea (DL clearly being the optimal choice here) ... but does it? Surplus value isn't the same thing, is it? I mean, if you have, say a 12% chance of hitting on a 7th round OT and a 6% chance of hitting on a 7th round DL, then to really figure out optimal strategy, you have to combine (multiply?) your surplus value with hit rate. Make sense?
One major question: when you put a player into a positional basket, are you using their primary NFL position only? I have a strong preference for prioritizing college OTs in the draft over G/C on the assumption that you effectively double your chances of a hit, because often they can be moved to G and be successful .. same applies to players with C/G college experience over college Gs. So, how many of your Interior OL players were OTs in college? Same question probably affects LBs too, but the lines have been separating between OLB/ILB for a long time anyway.
this is a neat analysis and I wouldn't be surprised if it were correct, but the small-sample-size alarms are going off in my head when I see only seven years of draft data faceted into 10 unique positions. how much variation would you see in these curves, for example, if you bootstrapped a bunch of fake datasets? how much influence do a handful of (un)successful picks have on these conclusions?
Great article and complement to the free agency piece. Positional value seems to be incorporated into team building to a certain extent, but not enough based on all the information available. Teams spend so many resources to find small edges in scouting but they undervalue the market based comparative edges that exist.
For sure. I’d go as far to say that teams should never draft a tier 3 position in the first round for any reason. The upside is too low and the opportunity cost too high.
Loved this. Will say I'm interested in how this relates to both the "there's no such thing as draft steals only draft reaches" wisdom of the crowds-concept and positional variability from draft to draft. Put in other words and in the context of this upcoming draft, if there's an early run how big of a consensus big board discount would be *too* big when it comes to taking a player in a premium position over (say) one of the TE?
I think the steals/reaches rule still applies for the most part. I'd consider it a reach when we're talking about jumping way up in the positional rankings, not just overall. If you're taking the CB10 over the CB6, that's a reach even if it isn't a low-value position. Just focus on positional ranks more than overall
Great stuff! It's creative analytics like this that totally made the "careers" of sites like FO, OTC and PFF.
It reinforces some ideas I intuitively had and challenges others. But, I think to really get the most meat off this bone, you'd need to combine it with "hit rate", however you define that (for example, I consider 3-4 years as a starter as a hit. Crude, but no less so than AV).
So, what I'm driving at is, that I have always thought late round picks (6-7) should be almost exclusively used on DBs, QBs and OTs, because it "seems" like those positions have the best chance to hit in the late rounds.
Your results on the surface, appear to counter that idea (DL clearly being the optimal choice here) ... but does it? Surplus value isn't the same thing, is it? I mean, if you have, say a 12% chance of hitting on a 7th round OT and a 6% chance of hitting on a 7th round DL, then to really figure out optimal strategy, you have to combine (multiply?) your surplus value with hit rate. Make sense?
One major question: when you put a player into a positional basket, are you using their primary NFL position only? I have a strong preference for prioritizing college OTs in the draft over G/C on the assumption that you effectively double your chances of a hit, because often they can be moved to G and be successful .. same applies to players with C/G college experience over college Gs. So, how many of your Interior OL players were OTs in college? Same question probably affects LBs too, but the lines have been separating between OLB/ILB for a long time anyway.
this is a neat analysis and I wouldn't be surprised if it were correct, but the small-sample-size alarms are going off in my head when I see only seven years of draft data faceted into 10 unique positions. how much variation would you see in these curves, for example, if you bootstrapped a bunch of fake datasets? how much influence do a handful of (un)successful picks have on these conclusions?
Value in % of salary cap is surplus per year, right? Not 4-year/contract-length?
Per year
Can you add a graph for QB?
https://twitter.com/KevinCole___/status/1640505221238464514
Maybe I'll will. It's such a unique role with a small sample that it makes any curve less predictive than for the others.
Great work!