1086: The Truth About Draft Capital

**JJ Zachariason** (0:02)
This is The Late-Round Podcast with your host, JJ Zachariason.
What's up, everyone? It's JJ Zachariason. In this episode, 1086 of The Late-Round Fantasy Football Podcast, thanks for tuning in. All of you have heard me talk about draft capital on this show, especially this time of year. When it comes to the zap model, when it comes to forecasting prospects at the next level, draft capital is the single most important input to consider. And there's a lot of reasons for that. For one, better players generally get drafted earlier. NFL teams, they might be stupid sometimes, but they're spending millions and millions of dollars in an attempt to find the right prospect at each draft pick. They're clearly going to be better at picking those players than most. But also, when a team makes an investment, that player is likely going to have more immediate opportunity, or opportunity in general. When we're playing fantasy football, volume is everything, and if that player has higher draft capital, there's a higher likelihood to see that volume. Now, clearly it's not that simple, but those are two key reasons as to why draft capital gets the signal that it does in models, and it's why it's important. But with all that being said, with all this talk about draft capital being the thing that we should care about most, you probably think that it's close to the only thing to care about. But that's not the case at all. On today's episode, the truth about draft capital, why it's important, but why it's not everything. Now before I get to the analysis, I wanted to send a quick reminder about the Late Round Prospect Guide. It's launching in less than a week. The guide is over 170 pages that covers everything that goes in the ZAP models. There's rankings, there's prospect profile write-ups for every running back wide receiver and tight end at the combine. There's year two profile write-ups for the guys entering their second season in the NFL, and more. You can still pre-order it right now for $14.99 over on lateround.com, and six days, that's going to rise to $19.99. So get on that, lateround.com.
To help explain the truth about draft capital, I of course went into the ZAP models database. That database has every running back wide receiver and now tight end who is either drafted or at the NFL combine since 2011 As I talked about on last week's mailbag episode, the ZAP model looks to predict a player's B2S or his best two seasons and PPR points per game across his first three years in the league. Minimum seven games played in a season for it to count and it throws out the final week of the regular season. So we're only looking at fantasy relevant games. To wrap your head around B2S, imagine a running back scored 10, 12 and 14 PPR points per game over his first three years and he never missed a game. His B2S would be 13 It would average his best two seasons of 12 and 14 That's what the ZAP model is attempting to forecast. It's trying to predict which players will score highest in B2S and it gives a player some prospect score, a ZAP score to help with that. Now I bring that up because when talking about how important draft capital is, B2S is how I'm going to measure that importance. You can do this in a ton of different ways. Best season PPR points per game across his first three years. How many top 12, top 24 seasons the player had. I just do it via B2S. And since B2S features three seasons of data, on today's show, I'm looking at players who were drafted from 2011 through 2023 That way they've had a chance to play those three years. Now I think it's human nature for your brain to go into this linear mode when we talk about draft capital. Maybe not totally linear, but let me explain what I mean. You hear someone say a wide receiver was drafted 16th overall. Then you compare that player to someone who went 32nd overall. And then you compare a third player who went 48th overall. Those are all wide receivers who went 16 picks one after another. There were 16 selections between each pick. But performance versus draft capital doesn't behave in a linear way. For running back some wide receivers, the positions that we're going to focus on today, it's a downward slope where the first part of the first round, that slope, is really steep. But then it starts to level out. And at some points, draft capital is actually completely irrelevant. Now, I want to start with running back first because, quite frankly, it's an easier position to model in general compared to wide receiver. When it comes to big investments at running back in the draft, they almost always give us good fantasy production, which is why if Jeremiah Love's draft capital just keeps going up, he's actually going to end up in the legendary performer category in the model this year. I didn't think that was going to happen at first. But we know that top half of the first round running backs have been unreal over the last 15 or so years in fantasy football. Since 2011, here are the guys drafted in the top 16 Trent Richardson, Melvin Gordon, Todd Gurley, Ezekiel Elliott, Christian McCaffrey, Leonard Fournette, Saquon Barkley, Bichon Robinson, Jamir Gibbs, and Ashton Genti, and probably Jeremiah Love. All of those guys scored roughly 17 PPR points per game or more in one of their first three seasons, except Genti, who's only played one season. Even Trent Richardson got there as a rookie. Needless to say, it's not surprising that when you look at the correlation between draft capital and B2S for first-round running backs, there's decent correlation. But let's take a step backwards first. Like I said, this database has every running back who was at the NFL Combine or drafted since 2011 Anyone who goes undrafted just gets a default draft capital of undrafted. Their draft pick is all the same. The R-squared between draft capital, their raw pick, and B2S is about 0.45, which means draft capital explains roughly 45% of the variance in B2S outcomes across this entire sample, which is fairly strong correlation for fantasy football purposes. But that's across the whole sample. I made a thread on Twitter last week about prospecting through data, and one of the things that I brought up in that thread was R-squared analysis. You hear data people like myself talk about R-squared all the time, and people email me or hit me up and they're like, hey, JJ, what's the R-squared of your model? Now, for those of you who think I'm speaking a different language right now, R-squared, it's a statistical measure that tells us how much of the variation in one variable can be explained by another variable. When R-squared is one, it means it can explain everything. When it's zero, it means there's no linear relationship between those two variables. So, in this case, we're looking at Draft Capital versus B2S. So, we're seeing how much of the variation in a player's NFL production can be explained by a variable. In this case, that variable is Draft Capital. But people will say, hey, JJ, what's the R-squared of your model? And my answer is, that doesn't matter without context. If I tell you that the R-squared from my model is 0.42, you might be like, oh, that's actually not that bad. But actually, that's really bad because that's worse than the R-squared of just plain old draft capital. And anyone can improve that R-squared by adding certain data points. For example, if I removed all undrafted players, the R-squared between draft capital and B2S lowers to about 0.39. Now just so we're all on the same page, think about why that might be the case. It's because I'm removing a bunch of undrafted players who are almost definitely going to do nothing in the NFL. Which means they have super late draft capital, they're undrafted, and B2S's of 0, 1, maybe 2 points. That improves your correlation when they're in the data set. Having the undrafted players in the model improves that correlation. If you change that sample, then R-squared changes. For example, first round running backs. I told you that top half first round running backs have been really good since 2011, and they have. That means that the back half of the first round has crappier running backs, then we'll see some sort of correlation for first round running backs when looking at draft capital versus B2S, right? And that's exactly what we see. When looking only at first round running backs since 2011, the R-squared between where they were drafted and how they performed in B2S has been 0.56, which is stronger than all of these running backs when you look at round one through the undrafted guys. And that should hint to us that later first round running backs might not be that productive, or I should say, they've seen a dip in production compared to the guys drafted in the front half of the first round. Because if they were just as productive, the correlation for first round running backs wouldn't be very strong. As you'd expect as you move through the draft round by round, that correlation weakens. Running back is really top heavy. The top guys of the position are doing so much work. They're almost always hitting. Round 2 running back since 2011, they've had an R-squared of.16. Round 3, .11. And as you approach day 3, you start to get into numbers that are essentially just non-existent. Where you can just throw out draft capital and just pick who you want. You can choose two endpoints here to make your story. To create that narrative. Like for instance, if you look at running backs picked between picks 17 and 50 since 2011 through 2023, we've had 25 of them. You know what the correlation between draft capital and B2S among those 25 running backs is? Again, pick 17 to pick 50 It's.002. That's non-existent correlation. What that tells you is that historically, there hasn't been much of a difference in production for running backs drafted mid-round one through mid-round two. But if you add in those top 16 running backs, whoa, all of a sudden, that equation changes because, remember, those guys are doing a lot of the heavy lifting. The R squared goes from non-existent to very existent. It shifts to 0.25 with 34 running backs now in the sample. Like I said, you can pick and choose your endpoints here. Like for instance, pick 33 through pick 100, essentially day two picks. The R squared, 0.07. That's not very strong at all. Pick 50 to pick 100, the R squared is 0.00. Non-existent. It doesn't matter. Now this isn't to say that draft capital is irrelevant for those players. All things being equal, you'd still probably want the player who goes earlier in the draft. And again, I'm just choosing endpoints here. You can capture different players at different draft capitals and make it say one thing or another. But one thing here is super clear. The further you move through the draft, the less draft capital matters.