00:00:00
Welcome, welcome everybody to the podcast edition of Wharton Moneyball. This is Eric Bradlo, professor of
00:00:06
marketing, statistics, and data science here at the Wharton School. Some combination of myself, my co-host who is
00:00:11
here today, professor of statistics and data science Audi Winer, Kade Massie and
00:00:15
Shane Jensen are here every week on Wharton Moneyball. Audi, first, it's great to see you. It's always great to
00:00:21
be back with the show. 11. Oh, actually we're approaching 12 years of the show
00:00:25
now. But one of my favorite parts of the show, I know it is yours, is when we get
00:00:30
to interview people that actually apply statistics and data science in the real world. And today, not only is today no
00:00:37
exception, but probably someone that has impacted you'll talk about this audi I'm
00:00:42
sure, which you can talk about the big beta data bowl and how it's impacted you
00:00:45
and your ability to train our students. But we're honored to be joined by Michael goes by Mike Lopez. Mike's the
00:00:52
senior director of football data and analytics at the National Football League. Longtime friend of ours here at
00:00:57
Wharton Moneyball. I actually think Audi, it's important. A lot of times I don't need to read someone's bio, but I
00:01:02
think his bio is so impressive because of his ability to straddle academia and practice. I think it's important for our
00:01:08
listeners to know you can have your cake and eat it too. So I'll say at the National Football League, his work
00:01:14
centers on how to use data to enhance and better understand the game of football. Academically, his research is
00:01:20
split between causal inference with a specific focus on causal inference methods for multiple exposures and not
00:01:26
surprisingly application of statistics to sports. He's an associate editor of
00:01:30
one of our lead journals, Journal of Quantitative Analysis and Sports. And then in the more practical side, he's
00:01:35
written for 538, Sports Illustrated, Hockey News. Uh from 2014 to 2021, he worked Skidmore College first as an
00:01:42
assistant professor, then as a lecturer and a research associate. And in 2020, he was named the American Statistical
00:01:48
Associ statistics and sports significant contributor award. I think we could name
00:01:53
Mike Lopez that each and every year. So Mike, first of all, welcome back to Wharton Moneyball.
00:01:58
>> Thanks so much for having me. Really appreciate it. And that was totally
00:02:01
unnecessary. Um, and couple of those publications are out of business now. So maybe I didn't do such
00:02:08
a good job, but >> I don't know about that. 538 still in business. Well, sort of. Not really. No,
00:02:13
I think >> not really. Not really. Yeah. Audi, why don't we start with this? Um, it's not
00:02:18
on the in the list of topics that Mike helped to provide us, it's listed as the
00:02:22
last one, but since, as you know, the big data bowl meant so much to my son and my family. Um, could you start Audi
00:02:29
and just talk about what the big data bowl has meant for you and then it would be appropriate for you to ask the
00:02:33
question for Mike about the 2026 one and what's going and the impact it's had and
00:02:37
maybe tell people what it is. >> Okay. So I mean uh I can from our perspective the big data bowl you
00:02:42
started it as when when Eric when your son Zach was a senior it was the first year and we weren't even a a wasabi
00:02:50
hadn't even existed yet um and its kernel existed as a seminar that I was running every Tuesday evening Tuesday
00:02:58
afternoon with just my great students who are interested in statistics and sports and that included Zach uh Andrew
00:03:03
Castle um and some other students and when you had the dig data bowl they just jumped on it without even I didn't even
00:03:09
know about it. I don't think it it it hit my radar. The students found out about it much faster than I did. And I
00:03:14
remember but you the timeline was terrible. It was like right around finals week. I mean, I don't know what
00:03:18
you did, but you you set it up so that it was right during their finals, yet they managed to to put together a
00:03:24
winning entry. Um and uh that and it was a a terrific opportunity and we we've
00:03:29
jumped on it and and not every year. It depends on what the question is. Um and uh and actually for us it really depends
00:03:35
on how statistical it is as opposed to how almost uh maybe AI the question is or how the goal is to learn something
00:03:44
about football from a statistics lens or maybe you're trying to do forecasting.
00:03:46
So we don't always participate but we've had many many teams uh be finalists and
00:03:50
honorable mentions and of course even many students who did Moneyball Academy with me. They went on to do Moneyball I
00:03:57
mean the big data ball with their different schools uh that they went to and we've had many winners who who are
00:04:02
not at Penn but also but did Moneyball Academy. So it's a hugely important um
00:04:06
uh um enterprise and we we're delighted Mike to have you here and and for the
00:04:10
invention of the contest. >> Can Yeah. Can you tell us about it and just tell us what made you guys started
00:04:15
at the NFL? Could you tell us about the 2026 version and we'd love to just hear,
00:04:20
you know, the impact it's had on you personally and the NFL? >> Yeah, I mean in in the Field of Dreams,
00:04:26
it's if you build it, they will come, right? In in football, it's if you have
00:04:30
the data and share it, they will analyze it. And that's been the motto. Like we
00:04:34
when I started in 2018, it was the first year that the teams had the player tracking data. And when you're going
00:04:41
from traditional game stats in play-by-play, which has maybe 160 rows per game to this massive data set that
00:04:48
has 300, 400,000 rows per game, um, you just don't quite know what to do with
00:04:53
it. Uh, and traditional sports hackathons up till that point had been sort of one night, two night sprints.
00:05:00
And I mean, I had the player tracking data for four months on my computer and I couldn't get anything out of it,
00:05:05
right? So, you know, what are we going to learn in a sprint and a sort of a one night or twoight thing? Probably not
00:05:10
much that's usable by teams. Um, so we really needed four, five, six weeks at
00:05:15
least. Um, apologies for putting it in finals week. It was a it was one of those things that came together. I I I'm
00:05:22
still in retrospect um wouldn't say lucky, but it's it's pretty amazing that
00:05:26
we sort of were able to build that um without too much uh sort of um push back from the league. Um the league wanted
00:05:33
innovation. They wanted new ideas. the clubs wanted it. They they sort of valued this experience. Um Jay Reed and
00:05:39
I Jay's currently at MLS. You know, when we started asking the right questions at
00:05:43
the league office, we we didn't get a lot of push back. Um the the team was teams were interested in it and we could
00:05:49
certainly sell this as something that was going to help teams help the league office in terms of thinking about the
00:05:54
rules and sort of modern metrics. Um and then also helping NextGen Stats. Almost
00:05:58
every collaboration we have um for a topic comes with NextGen Stats. And and that sort of leads me to this year's
00:06:03
topic. Uh where we're really thinking about predicting player movement. You think about when a quarterback drops
00:06:08
back to pass. A lot of what we want to be able to do in terms of thinking about who that quarterback should throw to is
00:06:13
thinking about if they throw to a certain receiver, what will happen? You really need to know how the defenders
00:06:17
are going to move if the ball is thrown to a certain location in order to make those assessments. So it started with
00:06:22
that reality and then sort of worked backwards. Okay, maybe we don't focus on
00:06:26
the quarterback. We can do that a lot. Let's just focus on what the defenders
00:06:29
do. What's their movement like when the ball's in the air? Who covers more
00:06:33
space? Um, should you even be going for where the ball's landing or should you
00:06:36
be going to try to figure out where you're going to tackle the player once the ball's caught? Um, so a lot of fun
00:06:41
questions out of issues of competition, which is is not really surprising given the the effort and and the sort of
00:06:46
quality of folks that are participating at this point. Can I ask you um so all these submissions come in I understand
00:06:51
students there are you know winners finalist honorable mention kind of things like who gets to see the output
00:06:57
of this like besides you and maybe some people that you're like there's some
00:07:01
ones that are maybe so extraordinary you want to make sure it's shared throughout
00:07:05
like the National Football League offices but like do the teams get to see the results of these and like if you
00:07:12
could since you know we're also besides you know we are a business school um has
00:07:16
stuff come out of the NFL big data bowl that you think has either changed your thinking, the NFL's thinking, or like,
00:07:23
wow, teams are now doing something different. And who would have thought? It just, you know, these students or I
00:07:28
know there's a student division, an open division, but like these people don't
00:07:31
even work full-time and they did something that kind of moved the needle a little little bit in the sport we all
00:07:35
loved. >> Yeah. Yeah, I mean the NFC championship game back in 2020, I was sitting, you
00:07:40
know, having chili or whatever, just like everybody else, and suddenly on the Leonard Fornette ran for a touchdown,
00:07:46
and they put up a a thing like yards over expectation, Leonard Fornette plus 17 or plus 19, whatever the number was.
00:07:52
And I'm like that that's the big data bowl, right? Like that we turn from the
00:07:57
competition into a new metric for our um you know, our NGS team. And then not only that, they turned into sort of a
00:08:06
real time way to look at running back effectiveness on a given play, which is then shared with our network partners.
00:08:11
Um, we are over 20 metrics from the big data bowl that have been uh sort of used
00:08:17
or sort of put into the NextGen stats ecosystem. Um, that is then shared by clubs. Um, as always in professional
00:08:23
sports, clubs aren't exactly reaching out to say, "Hey, I'm using this and
00:08:26
it's really helped." Um, because they don't want to give any type of edge that
00:08:30
they found. Um but undoubtedly that it's being used on the club side. Um on the
00:08:35
league side, I think with each passing year, we're looking at the tracking data
00:08:38
as being able to replicate a lot of what exists in the football beta lexicon. Um
00:08:42
it's we're able to take the things that scouts have been doing for years um
00:08:46
turned into metrics and do it, you know, sort of at scale uh in ways that can speed up time, too. So we've used a lot
00:08:52
of it for our analysis of rules changes, looking at the space and the speed of the players in the kickoff, for example,
00:08:57
last year. Um and and obviously our job is to make the game better. We're pre-player agnostic, but you know, we
00:09:02
also understand that our teams are using it, too. >> Um I skipped by the entire 2025 season
00:09:08
and one of that's one of the things we wanted to talk to you about. Um we'll
00:09:11
talk about the Super Bowl in just a second, but you you mentioned for example the kickoff changes. You
00:09:17
mentioned player speed is now an issue. um what kind of things from uh whether it's metrics or now using player
00:09:24
tracking data are like what are the big forget what the solutions are yet what are the questions you guys are now
00:09:31
answering at the NFL I know for a large number of years and this will never go away player safety I know is at the top
00:09:37
of everybody's concern and that's I don't imagine that's ever going to
00:09:41
change but could you give us a broad sense of the topics to which you know when they come to your group the data
00:09:47
and analytics group what are the kind of questions they're asking you guys to
00:09:51
think about. >> So I I would say broadly and it's not always just us answering questions too.
00:09:56
A lot of it's sort of understanding what data is out there and what questions
00:09:59
that we know that they're sort of thinking in the back of their heads that we can use data to answer for too. Um,
00:10:04
and that's kind of the fun part about our group is there's a lot of room for
00:10:06
creativity uh to sort of think think a little bit. >> You let people know, by the way, how big
00:10:10
a group are we talking about and what are the backgrounds of people because a lot of people that are listening to our
00:10:15
show might say, "Huh, you know, I wonder when the next time Mike Lopez is going
00:10:19
to post a job at the NFL, like how group are we, how big a group are we talking about? What kind of degrees do people
00:10:24
tend to have, etc." It' be good to hear that, too, and your thoughts about what
00:10:27
kind of creativity and things you're able to do. >> Yeah. So I mean I I work on our football
00:10:32
data and analytics group and at when I started I think I was the first person at the league office to code in R in any
00:10:38
group anywhere. Um when I had to download R like it gave me a hard time because I don't think they knew what it
00:10:43
was. Um we we are now a lot more modern. Um our group right now is somewhere between 8 and 10 depending on the time
00:10:50
of year. Um that's on the football data and analytics side. Uh we're close with
00:10:53
our health and safety group on the analytic side. Uh we're close with our broadcast um and sort of network groups.
00:10:59
Um we are in the same larger data and analytics group as sort of a fan international um sponsorship. There's
00:11:06
sort of other data and analytics groups that we're we're sort of also um uh sort
00:11:10
of collaborating with at different points. Um our job is to make the game better. Um you know some some of our
00:11:17
folks on our team have PhDs. Some have um you know are really good football wizards and can code a little bit in R
00:11:23
Python. Um but largely we're looking at the competitiveness of the games. Uh the
00:11:28
officiating of the games, health and safety, pace of play and replay is a big one. Uh using modern using technology uh
00:11:35
to improve the game. Um so those are the the the areas that we're focused on. Uh
00:11:40
obviously a lot of the increases in accessibility of data, whether it's the player tracking data that we've had now
00:11:46
the last eight years, um inevitably we're going to be getting player skeletal pose data, trying to think
00:11:52
about that, how to use that for either improving the game, um improving the officiating, um thinking about how to
00:11:58
use to recognize certain objective aspects of penalties, for example. Um that's the the areas that falls under
00:12:03
our bucket and with each passing year, our group's doing more and more and having more of an impact. So it's a lot
00:12:08
of fun. Um, >> go ahead, please. >> The, uh, the new kickoff rules, how much
00:12:12
of a role did your team play with that? Did you try to forecast, for example, what the impact would have been on total
00:12:17
scoring, on on strategy? Um, it just seems like a totally different game in some in many aspects. I mean, the the
00:12:25
kickoff and what was your role in that? How do you think it went? Yeah, I mean in fact like a couple weeks ago I looked
00:12:31
at our um projection from last year where we projected a lot of our key metrics and I sort of put a check or a
00:12:37
check minus as to whether or not we got it right or wrong. Um yeah, I mean our group was in charge of projecting the
00:12:43
return rate um and then conditional off that what are all the things that are going to happen based off that return
00:12:48
rate. Um we built a drive simulation model which sort of simulated drives. We wo that up into a game simulation model
00:12:54
which then simulated games. Um trying to estimate the sort of overall change you
00:12:58
know holding all else equal and assuming that the typical trends from 24 we're
00:13:02
going to hold again in 25. You know what would we expect for scoring what we'd
00:13:06
expect for um you know we anticipated that we were going to drop about 100 punts and sure enough we dropped about
00:13:12
100 punts. Um we anticipated we'd have a little bit more scoring than we had. Um
00:13:16
but some other other things I think sort of are responsible for that. um impact on the total uh sort of we'll call them
00:13:24
action plays where it's not just total play count which is typically what the
00:13:28
league looks at but how much action are we returning to the game return yards um
00:13:32
obviously you can take a lot of those things and then turn them into projected injuries and things like that. So um you
00:13:37
know working with the the folks in in in the the various leagues or league office
00:13:41
to to sort of come up with a encompassing reflection of what that impact would be. So, I have two
00:13:47
follow-up questions to Audi's questions, but it's on the same basic topic, but
00:13:50
it's something Mike you just mentioned. Um, you know, since Audi and I are exactly the same age, graduated the same
00:13:55
time, we're, you know, in our days, you had to know what I'll call mass stat,
00:13:59
you had to build, you know, let's say probabilistic models, etc. Now, a lot can be done by simulation. So how much
00:14:08
of the I'm just interested how much of what you do is simulationbased and in some sense you know let's let the
00:14:15
power of computing you know handle the complex in some sense I'm not saying math's not important math's always
00:14:22
important but if we can just simulate because of massive computing power you know bazillions of outcomes let's let
00:14:29
that decide things >> yeah I mean in fact the two most recent rules changes that our group has had an
00:14:36
impact on uh the kickoff being one of them. And then the other one was overtime. I mean, our overtime model,
00:14:41
there's there's no empirical data to measure against. Um so what we did is we
00:14:45
took play data, we simulated drives, we simulated what what the overtime would look like. Um and then we got our
00:14:51
estimated impact of the the the sort of new overtime rule. So, um I wouldn't say
00:14:56
we're always simulating, but it it it's certainly a tool that we'll want and
00:15:00
depending on the question that we're after. Um you know, it's it's it's
00:15:03
certainly important to to be able to use. >> The next question is kind of a selfish
00:15:08
question for my own knowledge and how you think about this only because I was just teaching it today to my MBA
00:15:13
students. I'll call it a and for our listeners it'll become clear what I
00:15:17
mean, which is a multi-attribute objective function. And what I mean by that is you change the kickoff rule.
00:15:24
Okay, so certain things get better, maybe certain things get worse. So there's multiple metrics or attributes
00:15:30
you care about. And of course, any real decision maker has to decide how to trade off these different things. So I'm
00:15:37
not asking you for the secret decision-m of how the NFL does that, but I'm just
00:15:41
asking you more broadly. I would imagine any rule changes you think of, they're
00:15:46
not paro dominant in the sense that they improve every single metric the NFL would ever want or the teams would want.
00:15:53
I'm just interested how you approach those kind of problems from a leadership
00:15:56
perspective and thinking about it. >> Yeah, I mean that is exactly how we approach it, right? Like so when we did
00:16:01
overtime, we had a scale and depending on where you put your thumb on the scale was going to pick your favorite overtime
00:16:07
format, right? Do you want equity of the coin toss? Do you want the overtime to end fast? Do you want it to be
00:16:14
explainable to fans quickly? Um do you want it to have um uh sort of a traditional kickoff and punt play?
00:16:22
Right. So depending on where you were going to put your thumb on the scale there, that was going to pick your
00:16:25
favorite overtime format. Um we changed overtime because we wanted to sort of increase the importance the decrease the
00:16:32
importance of the coin toss. We changed the kickoff because we wanted um you know, we found a play that had an injury
00:16:37
rate that was more in line with runs and passes and as a result we wanted more competitiveness. Um, so yeah, I mean I
00:16:43
think our job is to be objective and to show the sort of full impact of of a potential change um or modification and
00:16:50
ultimately we're after rules that will benefit, you know, all five or six of
00:16:54
the elements that we're after and and those are a little bit unique and and there's not always easy solutions to
00:17:00
those, but that's uh ultimately what we're trying to do for the game. >> Well, let's talk now about the upcoming
00:17:05
game in I guess uh whatever five days, the Super Bowl coming up. I'll just repeat and then I'd love your thoughts
00:17:10
on this. So, we had Aaron Shatz, someone I'm sure you know of and know well, uh,
00:17:14
we had him on during the season and his comment was that Seattle and the Rams, by every, you know, metric that he's
00:17:21
ever created, which is a lot, that these were historically great teams. Like, measuring in the last 50 years, like
00:17:27
Seattle and the Rams the were in the top 10. And then he also commented that of course New England had like the third
00:17:33
easiest schedule since 1978. So, in your mind, like how do you guys think about the Super Bowl upcoming? Um, how do you
00:17:43
think about the teams? And, um, yeah, how does the NFL think about the game in general and how do you think about these
00:17:50
two teams? >> Well, it's it's funny like all the things that Aaron said, we don't do like
00:17:54
we don't exactly care who our best teams are. Um, we don't exactly care. we do
00:17:59
care a lot about strength of schedule and sort of balancing and sort of understanding the impact of our schedule
00:18:04
but within a season um you know there's not much we're going to be able to do
00:18:07
and so as a result there's not a ton of value in it. Um I I do think it's
00:18:12
interesting this year set an NFL record with I have 1,089 quarterback scrambles.
00:18:18
So that means of every NFL season in our history more quarterbacks scrambled this
00:18:22
year than ever before. Um and I have New England number two in scramble rate. Um,
00:18:27
so I think the the interesting part from our angle in terms of the some of the long-term trends of the game is that um,
00:18:33
New England has has typically made its uh, impact. Um, Drake May in the AFC title game was no different. Throwing
00:18:40
the ball, his numbers weren't great, but his impact was running the ball. Um,
00:18:44
curious where that nets out. I know Seattle plays maybe the second highest zone rate in the league. Um, so we're
00:18:49
we're sort of a little bit more agnostic in terms of, you know, caring which team
00:18:53
exactly is going to win. Um, obviously we have to deal at the league office. Sure.
00:18:57
>> Um, but I do think that, uh, the New England quarterback is somewhat reflective of the trends and sort of
00:19:02
where the game is going. An athletic quarterback that can do a lot and when he doesn't see anything downfield, isn't
00:19:07
afraid to run. >> Certainly, before I get to, it's certainly an interesting matchup because
00:19:11
I, you correct me if I'm wrong. I would assume, but maybe I'm I have pretty good
00:19:16
knowledge of football. Um, a zonebased team could be more susceptible to a running quarterback. Is that in general
00:19:23
the thought? Just to make sure I'm clear, I >> I think the idea is that they'll prevent
00:19:28
big scrambles because they're in the zone. If you go man coverage and your players are staying on their players,
00:19:33
conceivably you open up big gaps, >> but you could also assign somebody to
00:19:36
the quarterback, in which case maybe you'll stop them right at the line of scrimmage. Um, so my my sense is that
00:19:41
there'll be three, eight, eight, 10 yard chunks there for Drake, but maybe not
00:19:44
the big ones. Um, but but not I'm not obviously a prognosticator there. >> Please, Audi. Okay. So, um this year's
00:19:51
Super Bowl, um does not feature really any of the top teams that were thought to be the top teams in the preseason.
00:19:58
None of them. I mean, of course, we regret the the Eagles are not there, but none of the top teams. Uh in fact, each
00:20:04
of them had less less than 4% probability of being in the Super Bowl, let alone winning it. Yet, here they
00:20:10
are. So, I just ran a quick simulation. Turns out that um one out of every 20 years you'll get two
00:20:18
long shots in the Super Bowl. at least if you trust the preseason's odds and
00:20:21
let them run. >> Um that there's any of the these two specific teams is actually much smaller,
00:20:26
but that's not the right the right question. So I'm going to ask somewhat
00:20:29
of a causal inference kind of question to you without data. Um is there anything about the way the season played
00:20:35
out or anything about the way the forecasts were made preseason that suggests that we were just not getting
00:20:41
it? Um and that these teams, which in particular Seattle, but certainly New England as well, turned out to be
00:20:47
terrific and we missed that. Is there anything about the way the game has changed, the u you described scrambling,
00:20:52
potentially other outcomes that have made the made the the models kind of off their their calibration early on and
00:21:00
that in fact it took a long time before people decided that New England was any good maybe because they just couldn't
00:21:05
get over Tom Brady was not leading them or some sort sort of narrative. But I'm
00:21:09
very curious to know that people think deeply about the game in terms of structurally what you think of the fact
00:21:14
that we have two preseason long shots playing each other. It's certainly a trend that we at the
00:21:19
league office would be wanting to watch. Um, you know, we've done a lot of stuff
00:21:23
with the equity of our schedule, right? We have three standings based games after the year. Um, where one seeds play
00:21:29
uh, sorry, one division number one division winners play other number one division winners. Um, and so
00:21:34
understanding the impact of that, I believe both teams this year played the NFC South, which wasn't exactly our
00:21:39
strongest division. Um, so that might have played into their records, which then obviously played into Seattle
00:21:45
getting the number one seed, New England getting the number two seed, um, and the
00:21:48
strength of schedule thing that I know Aaron's talked about. Um, we've, and
00:21:52
part of this is stuff I had done before the NFL, too. But we've compared our our
00:21:56
sort of playoff format and structure to other leagues and have a good sense that
00:22:00
it's a single elimination tournament. A lot can happen. Um, these two teams put
00:22:04
themselves in pretty good position based on the regular season. So, I don't think
00:22:08
anything necessarily that happened in the postseason is that surprising. Um, but to get to this point in the regular
00:22:13
season, um, you know, we've we've tracked quarterback injuries and I do think based on the reality that
00:22:18
quarterbacks are scrambling more often and asked to I mean, we had um, our quarterbacks held on to the ball an
00:22:25
average of 2.7 seconds per drop back. Um, that's it doesn't sound like a lot,
00:22:31
but that's 04 more than a year before that. And it's 20 more than it was a decade ago. So
00:22:38
0.20 times 20,000 dropbacks in a season. You can do the math on how much longer quarterbacks are holding on to the ball.
00:22:45
Um I don't think it's as much just the quarterback wanting to hold the ball and
00:22:49
wait, wait, wait. Obviously defenses are in zone. They're further back, different
00:22:52
personnel. There's just not as many people open. Um so conceivably um you know if you put the extra emphasis on
00:22:59
the quarterback position you know with certainly the the trends of their availability is something that we at the
00:23:04
league office would be wanting to follow and adding noise to that right is only going to um sort of increase the
00:23:10
variance in terms of who moves on. I think Audi brings up Audi's point also brings up an interesting question from a
00:23:16
league office point of view which is let's imagine it happens for a significant number of years that one or
00:23:22
both teams in the Super Bowl were you know not preseason favorites. Now one is the NFL could say this is a good thing
00:23:28
you know in some sense this is parody this is lots of you know it's there's no
00:23:31
longer going to be the years of dynasties that could be good. Um, another thing you could say is, and
00:23:37
maybe this is implication, I'd love to hear your thoughts too, Audi, on this.
00:23:40
Maybe our preseason models need updating, and that's not the NFL's business. Like, you know, it could be
00:23:44
just the models aren't great. The other could be the NFL says, you know what,
00:23:48
maybe we red need to redesign the postseason in some way to make it so that, you know, I'll make it up, the
00:23:56
it's not as easy for someone to sneak in as the seven seed, although that's not
00:24:00
what happened this year. and it's not as easy for that team to make its way through and make it all the way to the
00:24:06
Super Bowl. So, let me start with you, Mike, and then Audi. How do you think about, you know, could this be is this
00:24:12
an indictment on preseason models or does it have implications for how the NFL thinks about designing a postseason
00:24:17
tournament? >> I do believe we are the only league that after a season takes the teams that
00:24:24
finish towards the top and gives them a harder schedule than the teams that finish towards the bottom. I mean, most
00:24:28
of the professional leagues at this point reward the teams that finish poorly with the higher draft picks. Um,
00:24:33
we also reward them with easier schedules. Um, and I think the longer that this would play out, I do know this
00:24:39
is something the league cares a lot about and the clubs particularly care a lot about. Last year, we we tried to
00:24:44
push a um we found like there were a lot of positives about a proposal from Detroit that would have seated the
00:24:50
postseason teams based on winning percentage. um the clubs pointed out that it's not necessarily fair to
00:24:56
compare winning percentage alone when the quality of teams that each division finishes uh faces are much different.
00:25:03
And so I know the clubs are keenly aware of the differences in their schedules. Um and my hunch is that if if we have we
00:25:09
all we always celebrate the sort of last to to first teams but the reality is for
00:25:14
every last to first team there's probably a first to last team. Um and being cognizant of sort of what what is
00:25:19
best for the league is something that we would be tracking. AI, do you want to weigh in on this top?
00:25:23
>> Yeah, sure. Well, first of all, we love I mean the the imbalanced schedule, even
00:25:27
though you try to sort of caused by this offers great Simpsons Paradox examples.
00:25:32
We've every year you find a team that's uh that's better against good teams and
00:25:35
better against worse teams, but ends up having a worse record overall. That's
00:25:39
the classic um confounding. And so, by the way, might be the Tampa Bay Buccaneers. My team this year, they beat
00:25:46
Seattle in Seattle this year. Yeah, >> they also beat I think they I forget who
00:25:50
else they they beat the Brams I think they lost to New England by three points and then they couldn't beat the Saints,
00:25:56
Atlanta, the Panthers. They couldn't Miami and that inconsistency is is is a
00:26:01
football issue when I'm talking about teams that that beat the better teams at
00:26:06
at a they have a lower winning percentage against better team because they're good teams and they beat the the
00:26:10
poor team better than another team who's got a yet because the team has a just
00:26:14
imbalanced schedule, their overall record looks a lot better. And that was why people didn't trust New England
00:26:19
because their their record was so built upon beating beating up on weak teams that we felt we had no no substantive ef
00:26:26
estimate of how they would do against strong teams. Um, I will say that u I don't think it's in in MLS, which is
00:26:32
soccer, they don't have a mis in so much of an imbalanced schedule as they have
00:26:36
these other tournaments that the top teams have to play, which brutalizes their bodies and the travel and the
00:26:43
exhaustion of playing soccer and becomes very hard to repeat because you get to you get to do all these extra things
00:26:50
like um the cups they play in in in South America and the and they travel and that the teams that don't do well
00:26:56
don't get to perform in. that there there's uh other other leagues have to
00:27:00
to deal with that. Um but I but I have to say I mean I'm not sure the models
00:27:05
were wrong. Um I think that maybe people just didn't believe that someone like a
00:27:10
rookie essentially a rookie quarterback like Derek May not rookie but is his Drake May. Drake May
00:27:14
>> Drake May is his second year or is is that what he's second year and someone
00:27:18
who had who played for the Jets for so long could actually be good. I mean, may maybe it's our it's our na our story
00:27:24
narrative that's kind of gotten in the way as a uh as a Jets fan who's who's
00:27:29
sort of given up hope um and adopted my Philadelphia Eagles. Um how does Sam Darnold go for so many years at the Jets
00:27:36
doing nothing and now now being two years in a row um a 13 game winner? We got >> 14 14. So, so I I have one of the things
00:27:47
we do a lot of which I think is is underutilized in football is we try to understand the conditions in which teams
00:27:54
play >> and in baseball every time somebody talks about a game in course field
00:27:58
everybody just naturally adjusts for the altitude of course field >> for whatever reason we just don't do it
00:28:03
in football. Um you know and I think I look at Darnold and we look at we treat all quarterbacks the same etc. Um, I
00:28:11
don't if we look at all the reclamation projects, the quarterbacks that have
00:28:14
gone from bad sort of situations team-wise to good situations teamwise. I mean, Baker Mayfield went from Cleveland
00:28:22
to Tampa. Daniel Jones from the Giants to Indianapolis. Gino Smith went from the Jets to Seattle. Darnold went from
00:28:28
the Jets Minnesota. All those teams went from outdoor stadiums, cold conditions,
00:28:33
tough conditions to domes or indoor stadiums and easier conditions. Um, when we look at Drake May's two worst games
00:28:40
of the year, the last two games, one was in basically a blizzard in Denver. The other one was in terrible conditions in
00:28:46
New England. Not an accident that those were poorly poor conditions. Um, I relatedly, I mean, Drew Brees,
00:28:52
unbelievable quarterback. Are we that surprised that the quarterback who set the single season completion record
00:28:57
played in a dome? No, we shouldn't be, right? That's where we would expect it.
00:29:00
So, um, the differences in in sort of quarterback setup, um, Darnold's had, you know, a tremendous, uh, turnaround
00:29:06
of his career, but I think those things play play a small part of it. And, you know, we're constantly trying to think
00:29:12
about the weather and the conditions of the game. More and more teams are having
00:29:16
turf, so what is the impact on those? We average a couple more points per game in
00:29:20
turf games than we do in outdoor games. Um, and so those are the the the natural
00:29:24
sort of league office type things that we get uh pretty nerdy with. Well, how many turf teams are there at this point?
00:29:30
>> Uh, we're pretty I don't know the number off the top of my head. I want to say
00:29:33
we're getting like 18ish. Um, and then obviously like Seattle's a sort of
00:29:37
inbetweener, right? It's an outdoor one, but it's it's partially closed, but
00:29:41
sometimes it's pretty bad weather there, too. So, um, yeah, they're they're
00:29:45
almost like halfway in between. >> So, Mike, for the last question I wanted
00:29:48
to ask you today, um, let's imagine obviously we hope we have you back long before a year from now, but let's
00:29:53
imagine the three of us are sitting here a year from now. um what rule changes are kind of on at least the that you can
00:30:01
talk about that are are on the consideration set that you guys are analyzing? >> Yeah, I mean I my hunch is we are you
00:30:09
know our our priorities for the offseason, you know, continuing to figure out it refining the kickoff play.
00:30:15
Um you know ultimately league office doesn't decide the rules, right? We put
00:30:18
them in you know we have conversations with the clubs and the clubs if they vote on a change you need 24 votes. So,
00:30:25
um I think the kickoff one and maybe some tweaks around the edges, keep it competitive, um but but making sure you
00:30:30
know the injury rates are low enough and things like that. Um my hunch is in where we are going in football is we
00:30:36
will be getting skeletal data. Um we had Hawkeye in all 30 stadiums this year. Their system is called Skeletrack. Um
00:30:43
the completeness of this data is is fairly promising in terms of um being able to show the the the joints of the
00:30:48
the players. Um and trying to think about where we can use that data to improve the game um is a big part of our
00:30:55
offseason. Um so for example, you know, for years you've had to pay a company to
00:31:01
figure out if a player was in a two-point stance or three-point stance at the time of the snap. We don't need
00:31:05
to do that anymore, right? We'll just pretty quickly calculate that when we see the skeletal data. Um you know
00:31:12
relatedly we can calculate a lot of additional football metrics with this. Um we can uh recreate the officials view
00:31:17
of a play um based on the the sort of where their vantage point of uh vantage point of vantage point was when a a sort
00:31:24
of a defensive pass interference occurred things like that. Um so I don't quite know what we're going to do with
00:31:30
it. Um but I know that it's going to be a bigger part of the gun in year. Hm.
00:31:34
Well, we've been fortunate uh for the last half hour here on Wharton Moneyball
00:31:38
on the Wharton podcast network uh to have Mike Lopez. Mike is senior director of football data and analytics at the
00:31:44
NFL. Uh we've talked a lot about and please please I know you will please continue with the big data bowl. I don't
00:31:49
mean just for our students. I just mean it's you know people love the NFL and as
00:31:54
Audi says uh they think they're learning about sports and all this stuff. We're
00:31:58
teaching them statistics and data science through the Trojan horse of the NFL. And there is no better Trojan
00:32:03
horse. So Mike, Audi and I would like to thank you for joining us today on board
00:32:07
Moneyball. >> Always a pleasure. And um tell Shane and Kate I miss him. >> We'll we'll do. Uh thanks Mike and we'll
00:32:13
join you again right after the break. >> Sounds great. >> Welcome back to the Wharton podcast
00:32:19
edition of Wharton Moneyball. This is Eric Bradlo. I'm here today with my friend colleague uh Professor Audi
00:32:24
Winer. some combination of the two of us. Kade Massie and Shane Jensen are here every week on Morton Moneyball. So
00:32:30
Audi, obviously we just finished with uh Mike Lopez talking about the NFL. I want
00:32:34
to talk to you about a few other things. Whenever it's just you and me, I like to
00:32:37
kind of interview you about how you think about various things going on in the sporting world.
00:32:42
>> So I want to provide you some data on tennis given the Australian Open just
00:32:49
happened. I know you know this. Carlos Alcarez won the Australian. Yes, >> he's not got the career grand slam at
00:32:55
the youngest age ever, age 22, two years younger than Nadal was when he won it at
00:32:59
age 24. >> He's also got seven Grand Slam titles, which is also I'll just give you an
00:33:06
example. Feder and Djokovic had one at age 22. He's got seven. Well, you know,
00:33:13
just to interrupt here, uh we knew maybe earlier with Alcarez that he was going to be great earlier in his life than any
00:33:21
other player. I remember. Well, how old was he when we were talking about this phenom coming up? 16 was he? 17.
00:33:27
>> 17. He started he didn't win a major until I think he I know he's won his
00:33:31
first one in 2022. So, he's probably 18/19 when he won his first. >> But the the reviews of the of Algaras as
00:33:39
a kid were just off the charts. The only thing that I remember that's comparable
00:33:44
in terms of previews of stunning greatness. So we as statistitians always regress forecasts to the mean. You have
00:33:52
to that produces the best forecast. Now for our listeners what that means is when someone makes a forecast that's
00:33:58
really extreme there's going to be some regression meaning moving it down towards the average. Um and we see this
00:34:05
in lots of places. So as a as someone who interacts with the media whenever I hear a you know a crazy expectation you
00:34:12
always have to bring it a little bit down to earth. But Alcarez has certainly b born borne its way out to the fullest.
00:34:18
And the other one that I maybe I'll just which I if you look back which superstar
00:34:24
forecast they're probably others um have turned into absolute dead-on predictions.
00:34:32
>> Well, I mean the f the two that come to mind obviously is Tiger Woods and golf
00:34:36
and >> certainly Tiger Woods and golf. >> No no Tiger Woods in golf. LeBron James
00:34:40
in basketball. LeBron was, you know, at age 15, people knew about LeBron James. And he entered the NBA at age 18.
00:34:47
>> Bryce Harper, you think in in baseball? >> I don't think Bryce Harper is at the
00:34:52
level obviously of a Tiger even comparable. >> No, he's not. But he Yeah. So, in some
00:34:56
sense, he's regressed. He's going to be Hall of Fame, right? To show you, but
00:35:01
he's not going to be tier one Hall of Famer. >> No, he's not going to be a tier one Hall
00:35:04
of Famer. I I'd have to think about the other sports in >> there's one in baseball that I that I
00:35:09
think is his career is only halfway over. Um it's not Aaron Judge because he
00:35:12
was never forecasted to be terrific. Um >> I'm just trying to think who you're
00:35:16
thinking about. Was it Juan Sodto? >> Uh no Juan Sto didn't have that.
00:35:21
>> It's it's certainly nobody predicted Mike Trout until he did what he did at
00:35:24
age 19. Um Um the answer is Show Otani. >> Yeah, but he didn't join the N MLB until
00:35:32
what age? What age was he when he >> No, it's not. No, it was one. So,
00:35:36
usually we discount Japanese performance in the Japanese leagues. And so, I remember listening to Neil Payne on on
00:35:43
his show Hot Takedown talking about this insane forecasting and he was of course
00:35:48
gave it a statistical spin saying you no one ever lives up to the hype of being top in two positions. It just doesn't
00:35:56
happen. Um, but he was forecasted to be a star pitcher and a star hitter. And I remember listening to this and thinking,
00:36:02
"No way. he'll end up being one and not the other. Yet here he is being both.
00:36:08
And so here's my example. Uh let me so let me ask you to do a little forecasting here. So we just said Alcarz
00:36:14
has seven majors. >> Okay? >> Let's say he's basically been playing
00:36:18
since the beginning of 2022. He might have started a little earlier. Let's just assume as a real professional since
00:36:23
the age of 18, he's played 17 majors. He's won seven. So let's assume he's
00:36:27
winning at a 40% rate right now. Right now. Right now he's winning at a 40% rate. Although he's won five of the last
00:36:34
nine, it's fine. If you had to project out, I'm not going to ask you for his
00:36:39
career. We go out another five years. Okay? So, 20 majors. Is there any reason why I shouldn't
00:36:48
predict he'll have eight more, which is 15. >> Well, you know, okay. So, uh, he's been
00:36:54
healthy, correct? >> He has been healthy for I'm trying to decide if he's missed any because you're
00:37:00
right. What you're pointing out also is there's no doubt if Rafa Nadal hadn't
00:37:05
been injured in his career, he might well have the most majors. He has 22. Obviously, Jookovic has 24, but Nadal
00:37:11
missed years worth of majors. Djokovic, I don't know he's ever missed one.
00:37:16
Federer wasn't that injured in his career. Nadal was. Let's Yeah, you're
00:37:19
right. So, we have to take into the probability, right? So, miss some, >> right? So, he's uh so that your eight
00:37:25
forecast is just is just sliding over the previous fraction, rolling it forward. No, he's going to he's heading
00:37:30
into his prime, right? So, >> he's not even No, no. Prime typically in
00:37:33
men's tennis is like 27. This is not even his prime years, >> right? So, by that measure, it'll go up.
00:37:39
But nevertheless, we're you got to regress down slide word because he's been so dominant going. It's a tr such a
00:37:45
tricky thing. Also, I what really really changes the forecast is what who emerges
00:37:50
to oppose him. Right. So, right now he's got sinner as his his principal um opponent and and regular challenger.
00:37:57
Jookovic is probably done, right? He's 38 years old. I don't think we're doing
00:38:01
much of that. >> Um, are there any uh any coming up and coming players who can consistently
00:38:08
challenge him? How many are there? >> Great question. So, I'm glad you use the
00:38:11
word consistently because let's even talk about the big three era, right? >> Um, Stan Marinka won three majors in
00:38:19
that period, but he couldn't consistently do it. Andy Murray, some people even call it the big four. Andy
00:38:24
Murray won three majors, but he couldn't consistently do it. Delpatro won one
00:38:30
major. >> Yeah, I remember that. >> So, no, no, I'm just saying. So, are you
00:38:34
asking me right now? I don't see anybody out there on the men's side. >> The better opportunity. People are
00:38:40
talking about Ben Shelton or Taylor Fritz or even Zarev who as you remember was one game away from beating Alcarass
00:38:47
and the Australian, but consistently absolutely not. And so now the challenge is like it's not unreasonable to and I'm
00:38:57
even being I think conservative here. Let's since they've won combined the
00:39:01
last nine majors. It's the longest streak by the way in tennis history. The two men have won nine the last nine
00:39:07
majors. All of 24 all of 25 and now the Australian and 26. Five have been won by
00:39:13
Alcarz. Four have been won by center. Okay. I think it's fair. It's not unreasonable predict and center's 24.
00:39:19
It's not unreasonable predict over the next four or five years combined the two
00:39:23
of them will win 75% of the majors. It does not mean that there isn't one a year on average that somebody else could
00:39:29
win, but right now I don't know how you could predict less than that. And if
00:39:33
that's true, then 12 of the next uh 16 majors are going to be won by one of the
00:39:40
two of them. And even if it's a 66 split, that puts Alcarz at age 26 with 13 majors. Yeah, I'd say that's a
00:39:47
reasonable expectation quite honestly. >> Yeah. So, >> that's definitely I mean what I'm what's
00:39:52
amazing about tennis is that it really does have this uh longtailed distribution. You don't
00:40:00
that you you don't have this pileup at the at the you almost like you think about it is at what is the end of what a
00:40:06
human can accomplish and then you have this pileup at that near that maximum. Where do you see that? You see that in
00:40:12
in sprinting, right? Um you see that in home run rates even. Um, but it seems like in tennis the one or two or even
00:40:22
three bests just leap out way ahead of everyone else. And is that due to the fact that tennis is such a
00:40:30
>> a tennis match has so many >> um repetitions? I think it is. I think
00:40:36
it is. You know, this is what I always say. You know, it's the same thing. Obviously, you know, I have a son that
00:40:40
plays squash. All three play squash, but one was pretty competitive. you hit like I forget the number of
00:40:46
balls, but if there's like 150 points in a squash match, you might hit a thousand
00:40:50
balls in a match. You know, in tennis, you know, think about just how many shots are hit. And if I hit even 2% 3%
00:40:59
better than you, you add that over a match and now I've hit 30, 40 balls better than you, and that's going to
00:41:05
make the difference in a tennis match. And even more so in the majors where it's best to five. If if anything, the
00:41:12
longer the match obviously is favoring the better player and so it's just bigger N. There we go.
00:41:18
>> And that's that's but you'd still would expect to have a little bit more
00:41:21
clumping than we do considering that there should be some intra interournament variation. Like what I
00:41:28
mean by that is that the center that shows up to the Wimbledon might not be the center who shows up to the
00:41:32
Australian Open, shows up to US Open, etc. you'd expect to see enough variation, which is human to do that,
00:41:40
right? Um to see that the the the compression among the top, but you do you don't in in in tennis, the dominant
00:41:47
players are able to just not only win, but crush um say five or 10. >> I only have 20. I have one other pet
00:41:56
topics related to test and I want I have want to go over to MLB in a second but um you know obviously
00:42:03
Jookovic is the most accomplished player of all time you know he's won the most
00:42:07
majors the one the most the Masters 1000 ranked number one in the world the most
00:42:10
number of weeks I still have a problem calling him the best and let me say why or the GOAT so
00:42:17
let me just say why >> so between Federer Nadal and Djokovic this is a metric Maybe I've cherrypicked
00:42:25
it. He has the worst winning percentage in Grand Slam finals by far. Now, he's
00:42:31
got the most, but that's cuz he was never injured and he made the most finals, but his winning percentage is
00:42:36
the worst in Grand Slam finals. >> Yep. Yep. Back in the Okay. >> He has a losing record to Nadal in
00:42:44
majors and in major finals. >> Well, that's because of the Nidal's
00:42:48
dominance in the French, right? >> That's I'm going to get to that.
00:42:51
>> Now, he has a better record. He has a winning record against Nadal 31 and 29
00:42:56
in his career overall. Winning record against Federer 27 and 23. Although you might expect it to be better given he's
00:43:03
6 years younger and when they played in their career. And actually if you end Federer's career ended at age 37 he
00:43:09
would have had a winning record by the way against Jookovic. Just so you know. Um if you had to rank order the four
00:43:16
majors in tennis from most prestigious to least please rank them for me. uh Wimbledon US Open
00:43:26
French Australia. >> I would completely agree and so would everyone else. Do you know who has the
00:43:31
most Wimbledons of all time? >> Roger Federer. >> You know who has the most US Opens of
00:43:36
all time? >> Roger Feder. >> Roger Federer. >> Do you know who has the most Frenches of
00:43:40
all time? >> The I have to go down to the Australian >> to get to Djokovic. And he's got so many
00:43:47
more. He's got 10. Matter of fact, >> that's another thing about Algres.
00:43:51
No, Jookovic was 10 and0 in the Australian finals. Now he's 10-1. But my point is he didn't
00:44:00
even win the M. I mean, >> he wasn't at work. >> So, let's let's be clear here between
00:44:05
you and me and anyone else listening, which hopefully you're thousands. Is Feder the goat?
00:44:14
That's a tough one. It's a tough one. Here's what I'll say. Um, this was
00:44:20
always my claim about Djokovic. Djokovic always plays great. >> Yep. >> That was his great strength.
00:44:28
When he was at his best, Nadal was at his best and Federer was at his best. Who do I think is winning the match?
00:44:36
I'll debate Feder Nadal. It's hard for me to say Djokovic because I've seen
00:44:40
Djokovic get blown out by each by those players and others. Andy Murray, Winka, Delpatro, blew out Djokovic when they
00:44:47
had their A+ game. I say to me, he's the most accomplished tennis player of all
00:44:52
time. Is he the greatest? Was his peak greatness greater than the others? No, I don't think so.
00:44:59
>> That's just my opinion. That's my opinion. By the way, I'm sure we could
00:45:02
look at a flawed EO rate. By if we looked at flawed ELO ratings, I think he does have a higher ELO rating.
00:45:08
>> Yeah, but that ELO doesn't know how to rate um ma matches by with a different
00:45:12
rating system. Right. It doesn't though. >> Yeah. All right. Well, that's that's
00:45:16
some tennis. So, I'm going to I know you haven't looked at this. I didn't put it
00:45:18
in the spreadsheet. There's no way you could see this. I'm going to give you
00:45:21
the 2025 win total for an MLB team and I want you to give me your guess of what the fan graph's 2026
00:45:31
win number is. Okay. You ready? >> Okay. Sure. >> I just picked out seven or eight. Okay.
00:45:35
Let's start with the Dodgers. The Dodgers won 93 games in 2025. And I'm right. They're the two-time repeating
00:45:40
champion Dodgers, right? Y. >> Okay. What do you think Fan Grass has them for uh 2026?
00:45:45
>> Okay, first of all, Fan isn't stupid, so they regress to the mean, but I think
00:45:49
the Dodgers significantly underperformed their their their payroll, their roster
00:45:55
given injuries. So, I'll bet they're at 96. >> Well, you're off by one. 97.
00:46:00
>> Okay. That was my I was about to say either one of those. Yeah. >> All right. The next one.
00:46:04
>> By the way, to our listeners and viewers, I am not cheating. >> HE'S NOT. I have it in front of me. He
00:46:10
cannot say this. >> Bring it up on my own. I'm just >> Yeah, there's no but he's not. He can't
00:46:13
see. Well, we'll see if you're not cheating in a second for the next one.
00:46:16
>> Yeah, >> the Braves won 76 games last year. >> What does Fan Graphs has them for 2026?
00:46:24
>> Uh somewhere in the low 80s, low to mid 80s. I'd probably say 84. >> 91.
00:46:29
>> 91. What do they pick up that they think that they that they're thinking?
00:46:33
>> I don't know. That's Well, that's the question. I had I didn't get that far to
00:46:36
look. Um here's another couple interesting ones. The Phillies won 96 games last year.
00:46:44
>> I would guess they're probably going to regress them down to about 91 92.
00:46:48
>> 85 8 That's an enormous regression. The Yankees won 94. What do you got?
00:46:55
>> Yankees haven't changed their team. Uh they're going to get Garrett Kobach. Uh
00:47:00
we have some pitchers I would say around 93. >> 87 >> 87. So, >> the Brewers 97. Best record in baseball.
00:47:09
What do you think? >> Under 90 for sure. >> 82. >> Yeah, of course. Yeah.
00:47:14
>> Here's the most fascinating one for me. The Rockies won 43 last year.
00:47:18
>> Yeah. >> What do you think they have? >> 65. >> 66. >> Yeah.
00:47:24
>> So, the two that surprised me, I have to admit, look, obviously, >> I mean, I have to say Atlanta's
00:47:28
surprising me. I mean, to predict them that high, that's why I gave them slightly above average, right? Why would
00:47:34
you pick them to be as higher as high as the Phillies? >> Higher than the Phillies unless unless
00:47:41
they gained some players or unless they you just felt like their true strength last year based on metrics was 85 90
00:47:49
wins and they just way underperformed. There'd be no other explanation. Yeah,
00:47:54
it's interesting because one of the one of the one of the examples I'm doing
00:47:56
with my class is to measure organizational advantage in terms of wins above payroll expectation.
00:48:04
>> Oh, that's a great metric. >> And uh and it's actually many people
00:48:07
have done this and they've actually done it incorrectly. Um a standard way to do
00:48:11
that is to build a model that predicts wins as a function of payroll and then residualize each team and then average.
00:48:18
But if you do that, you don't control >> say average over years. >> Uh yes. So in other words, every season.
00:48:23
Yeah. So you take you take 20 years worth of data, you predict the wins as a function of payroll
00:48:29
>> and then you uh and standardize it because money changes its value. >> Um so you do whatever standard.
00:48:35
>> You're telling me that's not right. >> What's that? >> That's not right.
00:48:38
>> Oh no. So yeah. So if you do that, >> that's what I would have done.
00:48:41
>> Yeah. The problem with that is that what happens as we imagine that better
00:48:47
management gets more money. Then you have a confounding. The worst teams, the people, the organizations
00:48:54
that are the worst could end up with very low payrolls >> causally >> and then and if you think causally,
00:49:01
which is essentially what we're trying to get at, you're going to have a
00:49:04
confounding component. And you see this in sports all the time, see this in field goal kicking, right? If you look
00:49:10
if you try to modeling the conditional distribution of the number of expected wins given the payroll. So, so what you
00:49:17
so so the standard way of people have always done this is they predict wins given payroll then then you build that
00:49:23
model then you residualize for every team and you take your average residual. The problem with that is that if there's
00:49:30
confounding and the teams that have a lot of money also have good management you'll absorb the managerial effect with
00:49:38
your estimate of the payroll. I see. So classic example is and by the way this this undermines the Yankees for example
00:49:45
Yankees don't particularly overperform their payroll but if on the other hand
00:49:50
you say to them that that high payroll is given to them because they're run well you can think about it like that.
00:49:57
So an alternative way to do it is you residualize each team's winning percentage and their payroll and
00:50:04
then you get the and then you predict their residual winning percentage versus their residual payroll. So the Yankees
00:50:12
are judged by how much more money they're paying than what they average in a given season. And you compare the
00:50:17
Yankees to how much they and that's that mathematically corresponds to a fixed
00:50:22
effects model. And that's the one that I come up with. And what's interesting is
00:50:26
Oakland A's are still at the top no matter how you calculate it, but the Yankees move into like fourth position
00:50:32
and the Atlanta Braves and the St. Louis Cardinals are the two teams that are in
00:50:37
between them. So the the the actual value of the and I I don't have the the lecture notes. If you give me a second,
00:50:44
I'll probably be able to pull it up. But at the actual value of the of the uh in
00:50:48
terms of wins above expectation after adjusting properly for the team quality for the Oakland A's it's around six for
00:50:56
the Yankees it's around three and a half and for the Cardinals it's around
00:51:00
somewhere and and and Braves it's around 3 to four. >> So this is some real that's some real
00:51:04
action. I mean those are that's a real number of wins. >> Yes. Now of course this was the A's in
00:51:09
the in the Billy Bean era when they had that advantage. I I wouldn't argue that
00:51:13
they have that going forward, but that might produce one of the reasons why you suspect Atlanta to be good because
00:51:18
they're generally good every year. And therefore, when we look back at a a a
00:51:22
poorer season that they had, that is their their forecast is to regress them much much much higher than 82. Um, and
00:51:31
that's maybe and maybe they've seen some some action off the field that we
00:51:34
haven't that our memories don't don't have and that's why they're coming up
00:51:38
with such a high number. But I am a little bit surprised and that's something I want to dig into.
00:51:43
>> So the last topic related to baseball since you and I always like to talk Hall
00:51:46
of Fame. But just for our listeners out there, by the way, one of my greatest honors in a couple of weeks we're going
00:51:51
to have Josh Trowick, the president of the National Baseball Hall of Fame uh on our show, which will be fantastic to
00:51:57
talk to him. Uh one of the questions I'm going to ask him is, and I told him I
00:52:01
was going to ask him this, like when do you see a day where advanced metrics dominate the plaques in the Hall of
00:52:06
Fame? like this p this person had a you know I well on base percentage has been on for a long time. What would be an
00:52:12
advanced metric that you would like love to see on someone's plaque out of in the
00:52:15
Hall of Fame? >> That's a great question which because I um >> uh B
00:52:22
uh I mean yes I mean for a pitcher batting average on balls of play I'm not really sure that um that would be it.
00:52:29
Um, I certainly wouldn't be war because that is still in in the uh although I
00:52:34
will say I do like the pitcher wars much better than the hitter wars. Um, I think
00:52:38
they're much much more >> percentage has been on for a long time. >> Percentage. OPS is a great number. OPS
00:52:45
>> ops is a single counting stat. Um, that makes a lot of sense to me as as
00:52:50
something that should be. In fact, if you go to a stadium now, OPS is on the is on the
00:52:55
>> Oh, it's on the screen. It's absolutely on the screen. >> And uh that that's a great number. Um so
00:53:00
if it you have to make it simple but um I probably pitching war starting pitcher
00:53:04
war uh is a great number >> interesting to see >> and uh it's interesting I went and spoke
00:53:09
at uh a Daniel X class in he was on our show he teaches a class in um in uh in baseball statistics and I I gave a to
00:53:18
talk about my grid war that I wrote with Ryan Bril and right and how that really
00:53:23
teases out in players that for the Hall of Fame that that that aren't um that
00:53:28
aren't uh um noticed by the the conventional metrics. Actually, interestingly enough, it came up two
00:53:34
pitchers rise substantially in in in well, not >> well, you've been talking about Kevin
00:53:40
Brown and Dave Steve forever. It's not the two of them, >> right? It's not the two of them, but
00:53:44
here's two here's two others that are interesting. They're Yankees. Ron Gidry
00:53:48
is underrated as with Hall of Fame credentials. really on and and if people are talking about Pettit, Gidri is way
00:53:55
ahead of Pettit in terms of of dominance during his during his peak and even longevity. Um yet nobody talks about Gry
00:54:04
for the Hall of Fame. I don't think we really should be talking about Pettit. I
00:54:08
mean that's he's a re but he might make it because >> for lots of reasons. Here's another un
00:54:14
unsung hero that was historically beloved but the Saber Matricians don't like him. Whitey Ford.
00:54:23
Whitey Ford is not I mean every he was I mean he has the best winning percentage
00:54:27
of any pitcher with more than you know >> history right in history >> in history it's not even close and of
00:54:33
course everyone discounts that because he played for the great Yankee teams right but if you actually look at what
00:54:38
he performed as a pitcher ignoring what the Yankees were able to do do for him on the batting side and look at his
00:54:45
opponents and look at the the way he was able to get out of difficult jams and really produce his difficult opponents
00:54:53
leave them with not very many um opportunities. He is an unsung player. I mean, he's a genuinely a top 20 pitcher
00:55:00
of all time. Wow. >> And uh most people don't think of him as that. >> So, just in the last one or two minutes,
00:55:06
uh any quick reactions? I don't know if you and I talked about it to, you know,
00:55:11
Jeff Kent, Carlos Beltron, or Andrew Jones. I mean, any excitement for you? As you know, I'll be there in Coopertown
00:55:18
seeing them. any excitement that you have about either of the three of them being in the Hall of Fame or you like
00:55:24
but no real excitement. >> No, Belchan was never much on my radar. He had his great He's really one of the
00:55:29
classic um um you know postseason player, never playing for my team, so I don't never played that much. And Jones,
00:55:38
of course, he finished his career with the Yankees and he was thoroughly mediocre in those seasons. Um, and I and
00:55:44
I always thought of him as, you know, he so much of his great years were in his early part of his career.
00:55:50
>> And it is a saber metric. I mean, frankly, he did over hit over 400 home
00:55:54
runs. Yes. >> Um, and he was a center fielder and he was a damn good one. Um,
00:55:58
>> damn good one. >> Only like there's like three players in history like him, Maize, and Griffy have
00:56:03
I know they they artificially make these cut offs, but like 400 home runs and 10
00:56:07
golden gloves in center field. >> Yeah. Yeah. Exactly. Um, and I remember when I first, this is a point of history
00:56:12
in our work, when we first got the grant from ESPN, it must be 20 years ago now to do uh maybe not quite that far. It
00:56:19
was maybe 20, 2007. Um, we got this not tracking data because it wasn't tracking
00:56:25
data, but it was video recorded. Human beings watched plays and wrote down information. We and we ESPN bought that
00:56:32
data for us and that launched my sports analytics uh research line and Shane's
00:56:37
as well. And in a paper that Shane and I wrote uh called SAFES uh um basically spatially adjusted fielding um metric,
00:56:45
we we discovered two particular center fielders who just stood out from the rest and that was Andrew Jones. The
00:56:51
other was Jim Edmonds. Um >> well that's that's he was certainly known for that.
00:56:56
>> Yeah. Yeah. And uh andrew Jones was particularly good at at at playing the
00:57:01
low ball, the ones that were they were they were short. And he was just an incredible center fielder. and we
00:57:07
actually valued that that um that contribution. >> I don't have a let me just say I don't
00:57:11
have a problem except for the sign stealing stuff or the banging on the trash can stuff with Beltran, but
00:57:16
whatever. Um I don't have a problem with Kent, Beltron, or Jones, you know. Am I
00:57:22
excited that they're in the Hall of Fame? >> Hall of Famers. All three
00:57:27
>> third tier Hall of Fame. >> Yeah, but they're fine. They're in
00:57:32
they're not undeserving. Are they exciting? Uh, who's coming on down the
00:57:36
pike next year? We looking that far ahead yet? >> Any new book at one point? I don't think there's
00:57:43
anybody that exciting, which is why a lot of people are thinking it could be Chase Utley's year next year given where
00:57:48
he's grown. Um, Andy Pettit could make it. Um, you know, there's even discussions, there's been a lot of
00:57:55
discussions where how can you put Utley in without Jimmy Rollins in who's better
00:57:58
than him on every statistical category. But either way, I think Utley and Pettit
00:58:03
may get in next year. >> I think >> interesting thing that it's funny
00:58:07
because uh second baseman traditionally under hits the shortstop and it's not
00:58:13
because it's a more difficult position and therefore you require it has to do
00:58:17
with the fact that the best athlete tends to get put at shortstop. So it's an interesting confounding, right? If
00:58:22
you think about confounding, right? If you look at the production at which position produces the worst batting
00:58:28
production historically at second base. >> Yeah. Because usually, as you said, they
00:58:33
that's where they tend to move people that can't field that well and you know,
00:58:36
they just want their bat in the lineup, but they're not the best athlete. >> Well, no, it's important. It's a middle
00:58:40
infield position, so at the major league level, you need a very good second baseman and that person has to be quick.
00:58:46
But they tend to as a group to under hit or offensively produce far less >> than the the shortstop which has to be
00:58:54
even better fielder. >> So it's an interesting um correlation that creates a causal misunderstanding.
00:59:01
It's a classic you know if you think about it. So Utley I think is far better
00:59:07
is far more deviant as a second baseman than Rollins is as a shortstop. That's a
00:59:11
And and by the way, a lot of people like myself, and we'll wrap up with that, a
00:59:15
lot of people like myself believe that is an important criterion for the Hall of Fame. Like Jeff Kent, put him in
00:59:22
center field. Jeff Kent's not a Hall of Famer. But does he have the most home
00:59:26
runs ever as a second baseman? Yep, he does. That's worth something, Audi. It's
00:59:31
got to be worth something. >> That's the argument right there. >> That's the argument.
00:59:35
>> All right. Well, this has been one hour of Wharton Moneyball. like to thank
00:59:38
again Mike Lopez, head of data science for the NFL for myself, my colleague and friend Winer. Uh in absentia, Kade
00:59:46
Massie and Shane Jensen. Uh it's been a great hour with you here on the Wharton
00:59:50
podcast network and Wharton Moneyball between now and next week. Enjoy the Super Bowl. Enjoy your sports. Enjoy
00:59:55
your statistics. We'll see you next week here on Wharton Moneyball.