1091: Mailbag - ZAP Model Misses, Zero RB, and More

**JJ Zachariason** (0:02)
This is The Late-Round Podcast with your host, JJ Zachariason.
What's up everyone, it's JJ Zachariason, and this is episode 1091 of The Late-Round Fantasy Football Podcast. Thanks for tuning in. Before getting to the mail back questions, I want to send a reminder for you to check out The Late-Round Prospect Guide. It's 170 pages of analytical information about this class, and it includes information on second-year players as well. All for $19.99. If you purchase now, you'll also get the post-draft version of the guide emailed to you just hours after the NFL Draft is over. So check it all out, lateround.com. Okay, let's get to these mail back questions.
The first question this week is from Patrick on Patreon. It says, What's the biggest miss in the ZAP model database? And what is the takeaway? So I get this question a lot, and I do think my answer in the last mail bag about publishing the database, publishing historical scores, it kind of goes back to this question being asked so frequently. And I'm not saying that you're guilty of this, Patrick, but there is this fixation on finding these individual instances where something was wrong. I'm going to be honest with you guys, the ZAP model has gotten tons of things wrong, but the NFL has too. Teams get this stuff wrong all the time in the NFL Draft. It's no different for me. When it comes to prospecting and when it comes to the ZAP model, I get stuff wrong all the time. I don't know exactly how to answer this one though, just to be honest, because I'm not sure exactly how to measure biggest miss. It's all relative to expectation. Like the current model has Jaylen Polk with a negative draft capital delta, and that's good, because it means that the model likes Polk less than the NFL did, or where he was actually drafted. But it's not like the model said, run for the hills, Jaylen Polk is going to be one of the biggest busts that you'll ever see at wide receiver. So was that a win or a loss? Is it a huge miss because the model didn't get the exact outcome? I think there have been guys who have come out in recent history, and I want to keep this to recent history just to make the content more engaging for you guys. But there have been players in recent history who I've just completely whiffed on. You can see that in the prospect guide each year. Bucky Irving. I did not expect Bucky Irving to be as good as he's been. Now, I've made some model changes and some iterations over the last couple of years, and he does look better than he did coming out a couple of years ago. But where my process has also changed is more on the subjective side, not even talking film watching here, but with the analysis that I do that isn't ZAP model focused. I've leaned more on the advanced metrics in my write-ups and in my analysis that I didn't have as much access to even just a couple of years ago. And some of those advanced numbers definitely paint a more favorable picture for a player like Irving. And hopefully, I would have caught that this time around, if another player exactly like Bucky Irving were to come out. A good example of that is with Zachariah Branch this year. Technically, the ZAP model doesn't like despise Branch. I wouldn't say it likes him, he has a negative draft capital Delta, but it's not so horribly negative that he'd be an obvious avoid when you only look at the model's inputs. But a lot of the advanced data paints a scary, scary picture, which is what the analysis shows in the prospect guide itself. Now, I know a lot of people are probably thinking, then why don't you just include that data in the model? The reason I don't is because modeling isn't about throwing as many data points together in an attempt to find the best answers. That usually creates a lot of issues. You're double counting things, you're overfitting, it can get messy. Usually, the best models, they're relatively simple, when it comes to the inputs that they're dealing with. Second, a lot of the data that people use just doesn't go back far enough. So you can't reliably trust it, at least very easily, in a model if you're looking to test it properly. And I'm not even getting into the issue of proprietary data, which isn't something that we need to talk about on today's show. But back to the question, I think for a player like Irving, it's about understanding what the model says and then looking beyond those inputs. If any of you guys read one of the original Late Round Prospect Guides when I first launched this business, the analysis in those are like night and day compared to what I'm doing now. Just objectively, I don't want to say the old guides were horrible analysis, but I've learned a lot over the last five years, and I try to get better every year that I do this. Now, as for some other misses, Rashid Rice was a recent miss. I couldn't boost him simply because of landing spot, and I don't think that was necessarily wrong. We've seen Kansas City wide receivers fail pretty frequently in this Mahomes era. I do think Rice could have been another case of leaning in more on the advanced metrics, though. But like, he had a 31% slot rate in college. His wide yards-per-route run was pretty average for a wide out in that database. His yards after the catch profile, that was great. And I think landing spot, it ultimately mattered a ton for Rashid Rice. It's just hard to know exactly how a team might deploy a player like Rice. Just last year, the model didn't mind Caleb Johnson. I wasn't quite as bullish as the model was, and there were fantasy managers who really liked him. But it's a big, big model miss. Even with some of the fixes that were made to the model this offseason, he still has a pretty good zap score. Maybe the low explosive run rate, the kind of more middling combine, maybe those things should have been reasons to cool off of Caleb Johnson. But I don't know, I think with prospecting work, oftentimes we see a miss like this, and we overstate what it actually all means. We take it so far and we live or die by that one player. That's just not how I view this stuff, or how I think anyone should view this stuff. One player shouldn't change your approach to prospecting, or how you should model things. It shouldn't make you completely change your process. We have no idea what's going on behind the scenes, how these players are off the field, what kind of work ethic they have. Usually that all comes out in age-adjusted production and other predictive analytics, but that's never a given. Again, NFL teams are missing on these guys all the time, and they're not nerds living in their mom's basement like I am. They're spending millions of dollars on this stuff. But those are just a few of the names that came to mind when I saw this question. I know there are plenty more, and I'm sure some of you are screaming, what about this guy that you missed on? But I think those three players sort of capture my process and how I've evolved pretty well. Just remember, there's going to be a lot of misses. That's how it goes with prospecting.