Rewind: Prospecting Through Data

00:00:00
Speaker 1: What's up, guys, JJ here with another late Round podcast rewind. Hopefully you can't hear my daughter in the background, but I hope you're doing well. I hope you're staying safe, and I hope that you're washing your hands. Last year, I brought draft analyst Anthony Amiko on to talk about prospecting through data. Anthony and I talked about why numbers matter when evaluating prospects, and then he gave some insight on some of the data points that he looks at when he's scouting. Now, the show today does abruptly end. You'll hear that in a second because I asked Anthony questions about last year's draft class, which clearly isn't relevant today. But I think there's a lot of good information from this show. I like bringing on guests during the off season who's scout in different ways because it's a process where there aren't any defined answers, and listening to how other people do things may be able to help you in the way that you're viewing certain prospects. So, without further ado, here's the first large chunk of the interview with Anthony, and for any of you not following him on Twitter, you can find him at a Meeksta. That's a and I see Sta. Enjoy the interview and I'll be back in your ears on Friday with the Weekly mail Bag Show. All right, let's start by getting a little bit philosophical. Why do you think so many scouts are against heavily weighing production when prospecting wide receivers and running backs? Like why does the eye tests have to mean absolutely everything to them?

00:01:21
Speaker 2: Yeah, it's a really good question.

00:01:23
Speaker 3: I think the heart of it is, like it's just like anything else, right, Like you put a lot of time into something and you want your work to be meaningful. I mean, I can't imagine the man hours it takes to grind all that tape, to have to do all that, and then you know, you go online and people are telling you, like what you're doing maybe doesn't matter because like I can do X, Y or Z and that's maybe quicker or better. So puts you in a spot to like have to anchor and defend your work instead of maybe being like open to the new research. So I think, you know, if you're looking at all that nuance all the time, and that's the basis for your takes, you know, the idea that you can like clip a few buttons or something and look at some numbers probably seems a little weird. And it probably doesn't help that maybe some of us in the numbers community take things a little too far in terms of propagating the value of numbers versus the film. I mean, I'm I'm definitely strongly in the camp that both matter. I think you probably are too, Yeah, And it's, uh, you know, we end up having these like weird holy wars on Twitter because like everyone just wants to like make sure that their stuff matters.

00:02:24
Speaker 1: But you know, right, yeah, part of me thinks too that like the approach that numbers driven analysts take, you know, given the fact that like, let's just look at the last fifteen or twenty years, as analytics have gotten bigger and bigger, it's almost like folks who are are into numbers and are are prospecting via numbers, you know, we might we might be able to go about it in a little bit more of a gentle way, right like like we're they're there. And I'm not saying you were mean necessarily, I just mean the community in general, where we discover something and instead of being understanding in a little bit more empathetic as to what the other side has done for the past fifty years, you know, looking at these at these at these guys via film. You know, I think that we can approach it in a way where it doesn't have to be such an attack. You know, it can be. It can come from a place where we're saying, you know what, numbers can can help. Right, it doesn't. We're not We're not We're not saying that what is being done is a horrible, horrible thing. We're saying that here's some added information to make prospecting a little bit more accurate.

00:03:28
Speaker 3: Absolutely, And I feel as though in football in particular, but like in a lot of sports, like people are just resistant to to things that are like outside of the realm of the sport. I mean, like people don't even really hire like coaches that they haven't played this sport before, right, let alone listen to analysts about like how to draft prospects.

00:03:46
Speaker 1: Yeah, I mean there have been many, many times, and I mentioned people saying have you ever even played football before? So I definitely feel that. So, Anthony, you've been working on your prospect models for some time now, can you sort of talk through your your approach and build those models and sort of how you've improved the model over the last year or though.

00:04:04
Speaker 3: Yeah, so I mean at first, just obviously just looking for some answers. I mean, the draft is probably my favorite sporting event of the whole year, but figuring out who's going to be like hashtag good and who's taking fantasy drafts is pretty tough. So I started trying to build you know, like a simple linear aggression model using some statistics to predict a player's best PPR season within their first three years. I chose that markers because I felt like, at least in terms of dynasty, like if a player, if a player is good and he doesn't do it, it doesn't be good for fantasy. If he doesn't do it in those first three years, it's it's very likely that that player is no longer on the roster of the team that drafted him.

00:04:43
Speaker 2: So I thought like that early success was really important.

00:04:47
Speaker 3: Always, you know, like we was saying before the film versus stats thing always pops up. So I was kind of like, there has to be a way to maybe quantify like what the game watchers are seeing, even if they're not quantifying it.

00:04:57
Speaker 2: For US.

00:04:58
Speaker 3: So I just looked at I just started adding Draft Scout rankings to my models from NFL Draft Scout, and you know, big gains were seen in predictiveness. Even bigger gains were seen when I started taking the log of those ranks. Not to get like two into the weeds, but rankings are obviously like very linear, Like the difference between you know, nine and ten is the difference between ninety nine and one hundred, But with logs, it's you know, the difference between one and ten is the same as the difference between ten and one hundred, So it like more highly values the players that we would consider it to be really good, and that ended up proving to be valuable as well. So those differences have been big over the last year or so. I've changed up the statistics in those models. We'll probably talk a little bit about that later in the show. But I've also just used some different approaches outside of like you know, your basic R squared hunting as I'll sometimes call it.

00:05:52
Speaker 2: You know, I've done some.

00:05:52
Speaker 3: Threshold testing to see where certain stats matter and how much, and most recently I've used logistic regression and to come up with some probability as a player success at the next level to give us probably you know, a little bit better of a picture of maybe how likely it is that these players are going to hit for us.

00:06:10
Speaker 1: Yeah, so you mentioned the Scout rankings as an input in this model. I'm assuming the Scout rankings standalone or not as good as this model mean or as good as what the model would say. Meaning the data that you're inputting alongside with the Scout rankings is really what's doing in for you.

00:06:24
Speaker 2: Yeah?

00:06:25
Speaker 3: Absolutely, I mean that the Scout the Scout rankings give us a really good baseline to work with, and then we're adding, you know, a couple of those key stats that are you know, driving our success even higher.

00:06:36
Speaker 1: Right, right, So what are some of the statistics that you found to sort of be worthwhile when you're looking at running backs? Then? Have you found that there are things that you think that people just tend to overrate from a statistical standpoint.

00:06:47
Speaker 3: Yeah, running back the stats that I've found matter the most have been a breakout age, which for me is the age of back hits one hundred and thirty or more adjusted all purpose yards per game. Just a kind of an number that I picked arbitrarily. But you know through testing it, it was working final year scrimmage yards per game and final year yards per scrimmage play, which.

00:07:10
Speaker 2: Something I was just talking to you about, you know, a couple of weeks ago.

00:07:12
Speaker 3: Yeah, that proved to be probably the most important out of all those, uh, you know, gives us again like that nice blend of sharing the offense, but also efficiency in terms of what I think people are overrating yards per carry I think is probably overused market share of team rush attempts, not necessarily rush yards, but just the attempts. And obviously, like all the raw statistics in the Russian department, I think that's probably the biggest misconception in productive production based analysis. Like you start talking about it and people bring up, oh, like look at this guy who had all these rushing yards and he was terrible, But that's that's not really what it's about. I mean, I think that if you're just looking at those raw numbers, you're probably not gonna find a ton of things that are useful.

00:07:55
Speaker 1: Yeah, that's really interesting. The the attempt share is really interesting. I mean, I've seen it matter to a degree, but I've also found that that sheer volume just seems to matter more we're trying to look at how these running backs looked from an attempt standpoint within their offense. Obviously when we're looking at things like market share of rushing attempts, but it just seems like the overall the amount of workload that these players are getting just seems to matter more.

00:08:26
Speaker 3: Yeah, like raw rush attempts per game was much more valuable, just because, I mean, the ability to carry a heavy workload I think is a skill. And you know, if you demonstrate that skill in college, it's it's likely that you could do that at the next level.

00:08:39
Speaker 1: Yeah, And I wonder if you could also even look at it from the perspective of if you have a really good running back on your team, and he's going to get more volume regardless of what that offense looks like. So if it's let's say it's a very past first offense, well, if you have this game changing running back, you're probably going to change that offense for that running back, right for sure. I think that's that's another reason why maybe the market share numbers didn't look as good as the overall volume. It's really interesting though, but I'll ask the same for wide receivers. You know, what are sort of the some of the worthwhile stats that you've seen that seem to work with wide receivers in their transition to the NFL, and then maybe some numbers that folks tend to overvalue.

00:09:17
Speaker 3: Yeah, I mean, I know you've talked about this before on this show, but receiving yards per team pass attempt is fantastic. I mean, it gives you the blend of market share and efficiency. I think that those things matter probably a little bit more at receiver than they do.

00:09:30
Speaker 2: At running back. Breakout age again important.

00:09:34
Speaker 3: That's for me year that or age that a player makes a thirty percent dominator. So the share of receiving and receiving yards and touchdowns. Market share of receiving yards on its own is pretty powerful, as is receiving touchdowns per team pass attempt. So, I you know, breaking it down per team pass attempt definitely proved to be valuable. But some of these bigger share numbers obviously do matter. And then the overrated statistics. I mean it's similar to running back. I mean, like the raw efficiency like yards per reception not incredibly important. I mean, if you if you have, like if you're breaking it down and you find like this group of really efficient receivers, Like, maybe that's valuable if they don't meet some other thresholds, but certainly you want to hit some other numbers before you get to that raw efficiency. And I kind of think that we may overrate total dominator too much. So again, like that average of receiving yarded share and touchdown share, I think, like there's thresholds that you want to hit, but I think too often maybe we look at somebody that has like a forty percent dominator, and we look at someone who has a thirty five percent dominator and we say, oh, well, the first player player is like definitely better. But I think really, as long as you're hitting some kind of minimum, that stuff ends up being like a little noisy once you get past like certain numbers.

00:10:49
Speaker 1: Yeah, I totally agree with you. The other thing too about Dominator that I haven't really talked that much about on this show for whatever reason, is that I tend to find more success by by breaking up the statistics and the market share numbers instead of combining them, like like dominator does you know, it just gives you a better idea of what that wide receiver sort of did. I mean, you can look at reception share versus yardage share and just find a difference there and just sort of sort of get a feel for what that not that that's predictive, but to just get a little feel of what that wide receiver did. And then on top of that, I found that standalone and to your point you talked about this as well, standalone touchdown numbers are really really important and and the fact that that combined with with yardage is really what is bringing together with dominator. When you're able to separate that, you can see there are certain players who might not have crazy, crazy good yardage marks, but they're able to make that jump because they're touchdown numbers are just so out of control or vice versa.

00:11:47
Speaker 3: Yeah, I totally agree. I mean, you want to know what makes these players good. I mean, a guy like you don't want to compare a guy like David Sales to a guy like I don't know, like Niki Harry or someone like just like their dominators might end up being like relatively close, but like Sales is scoring like tons and tons and tons of touchdowns and you know Harry is accumulating a ton of yards in his offense.

00:12:10
Speaker 2: Yeah.

00:12:11
Speaker 1: Yeah, have you found that any of this is different for tight ends. I do a lot less prospecting with tight ends. It's just not something that I've really jumped into quite yet. So I'm curious as your thoughts on that.

00:12:21
Speaker 3: Yeah, definitely haven't been as extensive with the tight ends like you were saying, But a lot of the stuff that I've looked at has actually said that just getting.

00:12:28
Speaker 2: The ball is predictive of success.

00:12:29
Speaker 3: So like receptions, receptions per game, whether it's final season or career, like, those are the numbers that pop the most for me. So it just seems like if you're in it, if you're a college offense, is throwing to the tight end at all at like a decent rate, like that's that means the guy's probably pretty good.

00:12:45
Speaker 1: Yeah, that makes sense. I'm gonna switch gears a little bit and ask something that's not necessarily related to just the statistics and the box score scouting, but it is the combined meaningful. Have you found that there's any meaning within your models?

00:13:01
Speaker 2: Yeah? I think the combine is really meaningful.

00:13:04
Speaker 3: I mean athleticism speed score in particular is like really important for running backs, Like that's actually in my running back model. Same but yeah, so but even just like aside from the models, and I think without doing too much testing of it, I do think that athleticism probably matters. A lot of tight end UH samples are a little harder to get there. But even if you're the athleticism is not popping in a lot of the models. Otherwise, Like I've just looked at some of the stuff with SPARK, and like you want guys to hit like some minimum thresholds. Like it again, like it may not be important to run like a four to three, but like if you're if you're trying to draft the receiver that runs like a four to nine obviously being you know, excessive, But like that's that's a problem, right, Like you want guys to hit minimum levels of athleticism, you know, in some way. I mean this is using SPARK, but like just looking at the SPARK stuff, like if you hit a couple of minimum thresholds, like your your success rate does go way up. It's just that like reaching the ends of the ranges isn't really important. You just have to be like above replacement or above a certain level. And I think that makes sense considering you know, these guys are entering the most athletic league in the sport that they play.

00:14:11
Speaker 1: Right, right, Yeah, no, it absolutely makes sense. I mean when Elijah Holyfield did what he did in the forty, I think that there's a lot of reason for people to freak out, you know, it makes a lot of sense to freak out at that bad of a forty yard dash, Whereas someone like David Montgomery, you know, he runs his slower than what we would want from a top back in the class, but at least we've seen successes within that range before.

00:14:35
Speaker 2: Definitely