00:00:00
Welcome, welcome to Wharton Moneyball. Welcome to a full hour of sports analytics here on the Wharton podcast
00:00:09
network. This is Kate Massie hosting this week with my longtime collaborators and friends, co-host, colleagues, Shane
00:00:16
Jensen and Eric Bradlo. Our fourth co-host, Audi Winer, not in today, sadly. Audi out and about. Audi doing
00:00:22
Audi things. Audi doing things the rest of us are jealous of. Audi's is not gonna be with us today, but he will be
00:00:28
back. Some combination of us, y'all know this. Some combination of us are here
00:00:32
almost every week of the year. 48, 49, 50 weeks of the year. We're here doing
00:00:37
the show. Have been for more than 11 years now. Delighted to be back. We're recording on Tuesday afternoon as we
00:00:44
usually do. Gonna do an hour. Gonna do the first half hour with guests. Going to roll into open lines, open topics in
00:00:51
the second half hour. Got to start the show straight away with a guest and one of our regulars, one of our favorites,
00:00:58
one of our long times, a family, practically families, Wharton Moneyball family, Brian Burke. Brian Burke is here
00:01:03
from ESPN. Brian, thank you for making time. >> Yeah, thank you uh for having me. Uh
00:01:09
always enjoy being on and I feel like uh like the fifth beetle kind of like uh I want to be a co-host and I feel like
00:01:19
you watch like a show long enough like Seinfeld, right? And you're like, I could be in the part in the apartment
00:01:24
with Jerry and Elaine and George and we'd all just get along. I Yeah. Here's
00:01:29
what >> I think. I think Hold on. I think I think you're the Eric Clapton of the
00:01:32
Beatles. I think I think that's the right role for you. I'm gonna give you
00:01:35
the >> I was going to say, Brian, the next time I can offer for myself, the next time
00:01:39
it's just me that's available, rather than me talking to myself for an hour on
00:01:43
the radio, why don't you come join me and you can co-host with me and we'll
00:01:46
we'll talk some sports for an hour. >> Yes. Oh, absolutely. I want to be the
00:01:50
designated Audi. Um >> I don't That's tough, man. That's tough.
00:01:56
>> Those are those are some big shoes to fill. Literally. >> I just He asks the toughest questions
00:02:01
and I actually I appreciate that more than anything. Even when he's asking me
00:02:05
the tough questions because you learn you learn that way. >> Yeah, he's uh we all aspire to more
00:02:11
audience in our intellectual work. For sure. Brian, as most of y'all know, is
00:02:15
at ESPN. He's a sports data scientist at ESPN. He's also the founder of the
00:02:20
website advanced football analytics. He was one of the very first to step out um
00:02:25
when football analytics started getting sophisticated. He was a Navy fighter pilot in a previous life. He moved from
00:02:32
this black and white engineering world into the shades of gray probability big transformation midlife transformation
00:02:40
and now he's right on the edge of all things all things football analytics. Brian, so much to talk about. Week one
00:02:45
is in the books. Um, why don't we start there? We've got deeper topics for you,
00:02:50
but you know, you're a football guy. You've been waiting all season for this
00:02:53
to happen. Now that we've seen 16 games played, what's top of mind for you?
00:02:58
>> Yeah. Um, a team that impresses me the most right now is Green Bay. Obviously,
00:03:03
they they picked up a a superstar defender. Um, convincing win over a top uh conference and division opponent. Um,
00:03:13
and so we actually have we we're looking at them with the fourth best chance to
00:03:17
make a Super Bowl. So, um, yeah. >> So, talk real quickly about your fourth
00:03:22
best chance. Is that one of these flat maxima situations? Because this seems like a pretty open year. So, what are
00:03:26
the first three look like? And where are the pack? Where's the pack? >> Yeah. Yeah. So, you're you're absolutely
00:03:31
right. It it pretty much always is is that way. It's a little flatter this year than than normal, but we have uh
00:03:38
Baltimore, Buffalo, Philadelphia, and then Green Bay. Um actually, here's another one I'll throw
00:03:45
out. I'm just looking at this. Uh Chargers have a better shot over Kansas City right now. Obviously, they have a
00:03:53
win in hand over them. Um but I'm still really surprised to see that. Uh you
00:03:58
know, we'll see if that comes to pass. I'm always prepared. One of the last
00:04:02
questions sometimes is like what are you looking for, you know, in this upcoming
00:04:06
season? And I'm I'm waiting for the Chiefs to kind of have have a down year.
00:04:09
It they're way overdue, but this could be this could be that year. >> Brian, what would you what would you
00:04:16
take as an early marker? It's easy to call a down year, you know, threequarters of the way through the
00:04:22
season. If you want to call it early, being a good prognosticator, what would be a sign of this is going to be a bad
00:04:28
year for the Chiefs? Gosh. Um, uh, Chris Jones not playing well. Yeah, just uh early injuries. Um,
00:04:42
they've already got a suspended player. Um, so yeah, just depletion early on.
00:04:48
Depletion early on and and just um, you know, losing a couple games that they they shouldn't. I mean, the the Brazil
00:04:54
game, it's really weird. There's just a lot of travel. It's international. It's
00:04:57
the first game of the year. you know, you can set that aside. But, um, that would be one thing to look at. They just
00:05:03
they've been so good for so long. Um, maybe it is maybe it is time. Maybe it's
00:05:09
just me wish casting as an AFC, you know, rival for him. >> But could be the thing to
00:05:14
>> Exactly, Brian. I I put in my I lit my candles last week on this show, said my
00:05:19
prayers, put put it out there as a long-suffering a not just one AFC team, but two AFC teams beating at the door
00:05:26
behind the Chiefs. I'm ready for them to step aside. Eric's trying to jump in
00:05:29
here. Speaking of uh Yeah, a little a little AFC ascia. It's a little bit. >> No, no, I'm staying with I'm staying
00:05:34
with my question is more >> your answer to Kate's question. So, when Kate asked you like what caught your eye
00:05:41
in week one, I could imagine a lot of things. One could be which teams are at the top. Two is which teams surprised
00:05:48
you the most? Three, it could be which teams foot FBI move the most. um which teams exceeded the line the most which
00:05:57
could be different than which teams m move the most. Um how did you think about answering Kate's question and what
00:06:04
what was surprising to you? And I'm leaving surprising broadly defined here.
00:06:09
>> I mean that's a good way to organize all those all those things. Um the Colts
00:06:14
moved the most uh up um just because they the size of their victory was so big. Um, and the way FBI works is
00:06:23
basically um, EPA uh, per play. Uh, so if you have big chunk EPA gains um, and your defense plays really well on top of
00:06:33
that, then you're you're going to you're going to move up. And the priors are
00:06:37
fairly weak um, on the NFL side. Uh, so it's easy for teams to move. They they
00:06:42
bumped up almost three points on a per game basis. So, if we thought that they were, let's say, a three-point underdog
00:06:50
against the average team on a neutral site, now we think, you know, they're they're probably just about average is
00:06:55
where we are. >> This is great because you've answered my question because I was going to say
00:06:58
you're not focused purely on the difference in their rank because they could be very bunched up and so they
00:07:04
moved a little bit, but they jumped. You're actually focused on their underlying latent strength parameter,
00:07:10
which is something that's continuous, which seems to me to be a better measure
00:07:13
of strength. Uh yeah. Yeah. I mean you're you're you're absolutely correct. So in the NFL
00:07:19
in particular, the the middle so like the middle 20 teams, let's say, um are
00:07:25
bunched up together almost indistinguishable in in a certain way. Like yeah, we might have, you know, the
00:07:30
the number 12 best team ranked as like a plus one over the average team, one point per game over the average team.
00:07:37
And you know, the the 20th or 22nd best team might be, you know, minus one or 1.5 or something like that. But then
00:07:44
there's this uncertainty around all of them and they're just kind of in this
00:07:46
this muddied middle. Um so you can shoot up a bunch of rank point, you know, rank
00:07:52
numbers pretty easily, especially early in the year. Uh the Bengals are one, they struggled against the Browns. Uh so
00:07:58
they're they got moved down. So they're the second biggest mover. Um looks like
00:08:03
Houston as well. They were they they dropped a few point few uh ranked points, but yeah, the Bengals just did
00:08:08
not did not impress. Brian, can you talk a little bit more about FBI uh in general? How much do
00:08:15
y'all revise that year-over-year? My sense is that some years it's minor
00:08:18
tweaks, some years it's major tweaks. Where are we in the cycle? And what has
00:08:22
been your role? How confident do you feel about that power ranking system? >> Um it's a really good question,
00:08:28
especially in light of the college rankings last week. Uh we took some arrows. Um we can get into that if you
00:08:35
like but uh the I I rebuilt this FBI model from the ground up uh starting last season. So last season was the
00:08:44
first year with a new uh completely new revised model. It is it's not only rebuilt every year though as far as like
00:08:52
the parameters go. It's rebuilt every every time we run it. it it re um relearns all the different parameters
00:09:00
and how heavy the the prior should be weighted and all the other kind of hyperparameters within within the model.
00:09:06
It's a basian regression. Um so everything is kind of relearned. It's always sort of up to date. It goes back
00:09:12
about you uses about the past 10 seasons to kind of calibrate things. Um so I'm
00:09:17
I'm really confident it did like embarrassingly well last year. uh so well that I I know we're just due for a
00:09:26
horrible reversion to the mean uh this season. But yeah, really confident. Part part of the reason we're so confident is
00:09:33
the priors are very solid. We we kind of cheat and we reverse engineer the Vegas
00:09:39
overunder like win totals um as kind of our starting point for the prior. So we're always pretty chalk in the
00:09:46
beginning of the year. Um >> Okay. Well, let's stay with that for a second because you you said is basian
00:09:51
and of course this is a quite a basian podcast if only implicitly but the the hallmark I would say is as the only you
00:09:58
know probably most outside basian person on the podcast I would say the hallmark
00:10:03
is this whole idea that you have prior and that you update with as new information comes in that's kind of the
00:10:09
hallmark of a basian system and so trick it's one thing to it's one thing to
00:10:13
backwards engineer prior from the betting market that's fine lots of folks can do that But now you've got to decide
00:10:19
how much weight to give those things. >> Yeah. >> And so can can you talk to us about what
00:10:23
you find? What's the optimal weight? And we often talk about it in terms of I I
00:10:27
think I I think I got this term somewhere fellas fictitious sample size. Like how many games worth of information
00:10:34
is in the prior? Meaning after after how many games of the season is the rating that you have on the team half prior and
00:10:42
half games? Shane, I'm sorry for staying in there for just a second. Yeah, and if
00:10:46
I could just add a specific kind of thing, too. You you you talked earlier about how, you know, these teams after
00:10:51
week one are moving quite a bit because the the these carefully designed priors,
00:10:56
it sounds like you don't actually put much weight on to start the season. I'm
00:10:59
a little surprised by that. >> Well, we don't decide the weight of the
00:11:03
prior. The the model decides itself. So, we're looking at the past 10 seasons or
00:11:08
so, and we're letting the model find the the way I term it is like the least
00:11:12
unlikely. They're all unlikely. All these parameters are extremely unlikely.
00:11:17
You know, we have the Ravens number one right now, like plus 6.1 points per game. The the the chance that they
00:11:24
actually end up that way or that they are like, you know, sort of um you know, ground truth truly 6.1 are is extremely
00:11:33
unlikely. It's just the least unlikely of all the different uh and then the weights of the prior basically the the
00:11:39
width of the the the prior distributions the uncertainty around our estimates is
00:11:46
what is basically what the weight is when we say weight of the priors and that is left for the model to learn on
00:11:53
its own. Yeah, I was going to say this is a this is a specific term as you know Brian in basian inference that's why it
00:11:59
also pays to be basian besides it's the only coherent form of inference it's
00:12:03
called self-norming you don't have to decide the weight it decides the weight
00:12:08
it's the ratio of of in some sense what K described the effective number of games in the prior in this case the
00:12:14
effective number of games in the likelihood or the data but the data decides that and so
00:12:21
>> with an extra loop where you cross valid I mean the model doesn't inherently a
00:12:25
basing model does not decide that >> you're adding an extra loop over kind of
00:12:30
like the you know cross validation kind of loop over over possible waitings of the prior and the likelihood I guess
00:12:38
I'll rephrase my question I'm surprised that the priors are given by this
00:12:42
process a small weight >> given he doesn't know he >> you might be exaggerating you might be
00:12:50
overemphasizing his comment that things move >> it's smaller than college. I'll say that
00:12:53
it's small. >> Well, the fact that you would move say the Chiefs, the Chargers over the Chiefs
00:12:57
based on one game, I think shows that the data is maybe >> ranking that was the forecast. I was
00:13:06
>> right. So, we we we think the Chiefs are slightly better than the Chargers still,
00:13:10
but uh because of that win in hand and and the chance that they'll have the tiebreaker uh at the end of the year as
00:13:18
well, just those those two things are enough to give them a better chance uh at the Super Bowl given at the current
00:13:26
snapshot. I'm asking my question because I'm always trying to develop my own
00:13:30
intuition, but also helping our listeners develop their intuition for how much weight we should have on prior
00:13:36
because that's that's kind of the whole game is how much we're reacting to new
00:13:39
information when it comes in. And there are lots of people who aren't doing this
00:13:44
empirically and they're saying, "Well, at this point in the season, we shouldn't have any old information, so
00:13:49
we're just kind of writing it down to zero." which we would oppose. But I
00:13:53
think that is so wrong because I think if you let the model tell you what's optimal and what we're talk what we're
00:13:59
saying is what's optimal for prediction. So when you say I let the model tell me
00:14:03
what the weight should be that's because the model's looking historically over 10
00:14:06
years if I wanted to predict as effectively as possible I would keep this weight. It would move like this.
00:14:11
Great. Do you have any sense? I know that you didn't you weren't prepared for
00:14:14
this question, but I'm asking you, you know, either h at the halfway point or
00:14:17
at the end of the season, at the end of the season, what's the weight of the
00:14:21
prior still? Because I think most people don't have good intuition. It >> I don't know. I I can exactly the way I
00:14:28
can't give you a number. I know it's still there. It never goes away. Um and
00:14:32
it probably shouldn't uh in in the vast majority of cases. One of the things I
00:14:38
have been thinking of is that um before I get into that I will say one of the things I have studied in the past is
00:14:46
how much luck is involved in observed outcomes. And I know that you get in the NFL I think you have to get past maybe
00:14:54
11 games before sort of the observed signal is as as strong as the the random um share the variance in in game
00:15:04
outcome. So you can you can it's it's a pretty easy exercise to do with a
00:15:08
binomial distribution. Um you know the variance of a a purely random binomial and then you have the variance of the
00:15:15
actual game outcomes and you can just subtract the variances and you and so you you I think you have to get past the
00:15:21
11th game before you are just even just break even with the luck. So um everybody take a breath um take a couple
00:15:31
breaths long. >> Yeah. And and the the thing I mentioned with the college um our college FBI is
00:15:38
at the other end of its kind of life cycle. It is at the end of its life cycle. It is is not as modern as as the
00:15:46
NFL model is. I'm not exactly sure. I don't work directly on that project. Um
00:15:51
but we yeah we were criticized because uh Texas lost Ohio State um and they remained number one in in FBI and a lot
00:16:00
of people don't quite uh fully understand what FBI is supposed doing combined with a lot of conspiracy
00:16:08
thoughts and uh and and and other things and we we we took a lot of arrows for that but um I think the the prior are
00:16:17
stronger on the college side and that that's probably why they stop. >> No, no, no, no. It wasn't even it wasn't
00:16:22
even that the postgame expectancy was not only pro Texas, but it was like wildly pro Texas. Like the the signal in
00:16:30
the game to the extent that there was signal in the game favored Texas like like I think Bill Connley said it was
00:16:35
one of the biggest discrepancies. It was the biggest discrepancy of week one. It
00:16:38
was one of the bigger ones you'll see. Yeah, we we took um I think we took solace. I think Bill had Texas like 15th
00:16:46
or something after that. So, we're like, "Yeah, we know we maybe we're wrong, but
00:16:51
um we're not alone." Uh but yeah, saying Bill dropped Bill dropped Texas to 15
00:16:58
after >> I don't think they started very high. I don't think they had nearly as high. We
00:17:02
had them as a solid number one >> going into the season. I don't think he
00:17:06
was as high, but I don't know. I don't >> Is Bill just putting his thumb on this
00:17:11
scale? He He doesn't He doesn't like Texas. He's He's a Oklahoma guy at
00:17:15
heart. I'm I'm joking, of course, but I I would take whether they're one or 15.
00:17:19
I think that's interesting. Who you you moved too little, he moved too much, it
00:17:23
seems. >> Yeah, we did move them down. I mean, they lost by one score on the road, you
00:17:26
know, against a great team. And so, they're not going to move very much. So,
00:17:29
it's just a a question of how tightly were they bunched kind of going in uh to
00:17:34
that game. And and we we had, you know, our priors are unlike the NF on the NFL side, they're kind of purely um we don't
00:17:42
cheat with them. They are built on top of how well you did last year. Uh how many returning starters do you have? Did
00:17:49
your coach change? I think there's some special attention placed on a returning
00:17:53
quarterback. Um and that's what drives Oh, and how many, you know, four or five
00:17:58
star recruits do you have in the pipeline over the last I don't know, like two or three years or so. Um, and
00:18:04
that used to work really, really well. And over the last, you know, two years or three years, the sport has completely
00:18:11
changed. And so that no longer works as well. And we should probably have a lot less confidence in our in our preseason
00:18:17
estimates of these teams. >> Well, it'll be fun. When the FBI, I first got to know it on the college
00:18:24
scene, and when it first came out, I thought it was one of the best public models available. And it and we we found
00:18:30
it really tracked massive body, which impressed us. And this is just speaking to how hard it is
00:18:36
to keep up. I mean, the world evolves, you know, not only the world evolves, but other other people jump in,
00:18:42
competition evolves, and you got to stay on top of things if you want to be a cutting edge power ranking system. And
00:18:49
let me just say I mean this sounds it is of course the realm of the geeks and we
00:18:54
and we are all geeks in this podcast but the claim is you you don't understand the game very
00:19:01
well or you don't understand what you're looking at very well if you can't
00:19:04
predict what's going to happen. It's almost the hallmark of how well you
00:19:06
understand it if you can predict what happens next. And so I take it this is kind of performance evaluation. You know
00:19:12
it's like do you understand? I mean can you really tell us who's good and why
00:19:16
they're good? And if you could, then you could predict what's going to happen
00:19:19
next. And that's really hard to do. So, let's all just compete on who can
00:19:22
predict because it says something about how well we understand the game. >> Absolutely. Yeah. I mean, just and
00:19:27
that's true of science in general. You have a theory and theories make predictions and you test the theory
00:19:32
based on, you know, how how closely the observations uh track with with their predictions. And that um you know that
00:19:40
basically what we're talking about is like a a science of football just applying the scientific method to to
00:19:46
something. just yeah it's wildly you know hard to pin down of course there's so much that
00:19:51
we don't understand yet and there's so much that we don't understand because
00:19:54
we're not as deep in the game as the people who do it for a living and let's
00:19:57
talk about a detail of the game another thing that you've done in your football
00:20:01
life is build fourth down models and so I know you've you continually tweaked those as well because we think
00:20:09
they're really good and then we find holes and we make them better we find some more holes we make them better but
00:20:14
there was a a questionable fourth down call. The high-profile game, one of the highest profile games of the
00:20:20
weekend was Baltimore going up to Buffalo. Lost a 15-point lead in the last four minutes of the game, including
00:20:27
uh a decision to punt the ball away um late. Do you have a take on that decision? And how could how should we
00:20:34
think about it? How are you thinking about it? >> Yeah, the model the model had I think
00:20:40
80% for go for it. The initial uh play by play I think had it a fourth and two and I think later they change it
00:20:49
to a fourth and three which isn't going to change things very much because we
00:20:52
had the punt at 70%. So I'm approximating. So I think it might have been like 81 versus 72 or something like
00:20:58
that. So the order of magnitude of the error is like 10% like win probability which Brian you're you're going a little
00:21:05
shortand some who may not be familiar including possibly me. 81 versus 72. >> Oh chance to win. So, if I go for it,
00:21:12
your chance of winning in that situation would be about 81%. If you uh punted in that situation, we
00:21:20
would estimate it's about 70 to 71, something like that. It was about 80 to 70. Um, which is a huge difference. Most
00:21:28
even big sort of enormous obvious fourth down errors can be maybe like two percentage points of win or three
00:21:36
percentage points of win probability. So when you get you get somebody like John
00:21:39
Harbaugh with a 10% error, you know something something is up. Something something is off. Um and after the game,
00:21:49
I think there there's reporting that uh Lamar Jackson had was cramping in that
00:21:55
final series. So and he's integral obviously to any kind of short yardage. Um and they elected to punt. So I think
00:22:02
that that played a part. I watched the play before. or if you watch very closely, Ed Oliver absolutely
00:22:07
obliterates Lamar Jackson on the play before and I think that was probably there's some some you know that would
00:22:14
cause some quote unquote cramping for sure. I think that must have played into it because the Ravens have really been
00:22:19
the best at this for a long time. And in fact, there was almost an identical situation. I think it was 2021 week two
00:22:27
against the Chiefs and they had a small lead, you know, within a field goal lead. And there's about two minutes left
00:22:34
to play. They were faced with a fourth and two uh on their own side of the field about their own, you know, 39 38
00:22:40
yard line, exact same place spot on the field. They went for it, converted and kept kept the win. So, uh, very
00:22:46
surprising, but I think the the answer lies with, uh, just a banged up quarterback.
00:22:52
>> Something we don't know. Interesting. Really interesting. Shane, were you
00:22:55
trying to get in? >> No, I mean, I I I I guess it is interesting to kind of hear that maybe
00:23:00
that context uh that Brian just brought up just because, you know, I was going to kind of jump in and say like, you
00:23:05
know, of all the teams to go forward on fourth and three, I think maybe only Philadelphia would be a higher chance of
00:23:11
converting. I think in general the personnel that the Ravens have. But now you say that, you know, if Lamar was,
00:23:17
you know, kind of banged up and very kind of locally right around that play that that that would that would help
00:23:23
explain it. That would be the only thing in my mind that would >> Do the Bills know that though?
00:23:27
>> Do the Bills know that? And and do do you have any other players on your team
00:23:31
who can pick up three yards on the Ravens? Like so I And is and it's 10%. you know, and like so I I still would
00:23:40
have probably gone for it if that were me. I'm I'm not trying to make excuses.
00:23:45
I think they still should have gone for it and I'm on the record and I you know
00:23:48
said so on on on Twitter but uh I mean that that probably explains and it probably you know we we when we look at
00:23:56
the end of the season and we look back and do a complete sort of meta analysis of like fourth down trends and who are
00:24:02
what which decisions were the biggest errors. Oftentimes we find the biggest errors
00:24:08
are on the like the go for it side like which is very unexpected and the reason is like injured kickers or injured
00:24:15
punters >> long snapper you is is in the locker room getting x-rays >> there's always there's usually some sort
00:24:24
of context that we're the models aren't picking up. Isn't this isn't this
00:24:27
literally the broken leg problem that people talk about with when you go with models and when you don't go with
00:24:32
models? Like you you the one of the few times if you have a good algorithm, one of the few times that you don't go with
00:24:38
the algorithm is if you know it's missing something. And the example people give is like if you're supposed
00:24:41
to model who's going to win a race and you know the runner has a broken leg. So
00:24:45
then you don't use the model and you're literally giving those examples from
00:24:49
your from your review of Sports. >> So this is an interesting example. So I
00:24:53
would assume Brian whether Baltimore won the game lost the game. Your estimate of
00:24:59
their FBI or strength parameter wouldn't actually change that much. But what does
00:25:04
change a lot is now their probability of let's say winning the AFC or going to
00:25:11
the Super Bowl. So can you give us a sense of how big a magnitude of an effect you had there? It's not like if I
00:25:17
if Baltimore I don't even know who they're playing next week. Whether they
00:25:20
m went for it and won 38-37 or lost 40 to 38. The probability of win ain't going to change Buckus as we say now. I
00:25:30
mean there's only I hate to say it there's only 16 games left and they got
00:25:34
to catch up two games essentially on the Bills and that's not going to be easy.
00:25:37
So how much is their let's call it go to the Super Bowl probability change? >> Yeah, I honestly Eric you asked the
00:25:44
exact same question I was going to anyway. It's like can we kind of think about because you talked about this 10%
00:25:48
on a game level being dramatic like can can we talk about season level kind of errors given the sort of the the
00:25:55
tiebreaking consequences of this particular victory and stuff like that? >> Yeah, we don't I'm I'm trying to look at
00:26:01
it right now. Um so we would say I think we had them 19.6 to make the Super Bowl
00:26:09
percent. Right now it's gone up >> 24 down. probably lost too. >> They weren't, you know, that was a 50-50
00:26:20
game going into the season against Buffalo. >> So, we weren't banking on them winning
00:26:26
as far as the model goes. So, um you know, all things considered, uh their percentage actually went up. I think
00:26:33
let's be honest, >> their offense looked really good and offense right sticky and defense is like
00:26:39
kind of like more about who you're playing off often times than uh and so given that the Buffalo Bills has have a
00:26:47
great offense and a great quarterback, the the Ravens defense just doesn't take
00:26:52
as big a hit as uh you might think. >> This lets me ask this is a perfect lead
00:26:56
in Brian to the question I was going to ask just a few minutes ago, but now that
00:26:59
we moved on to this game, I can talk about it here. Is FBI built to do well at predicting who's going to
00:27:09
go to the Super Bowl or pred? No. That's what I assumed. The answer was no. Like
00:27:13
you would build an entirely different model if you were trying to predict who's going to the Super Bowl versus
00:27:20
strength parameters on a given week A is playing B, how much we expect them to win by. So, I just want to I was going
00:27:26
to ask that earlier, but now I have a perfect leadin for that. um it wasn't built for super. So the fact that their
00:27:34
if you'd like their um FBI didn't change very much, but their probability of
00:27:40
winning the Super Bowl actually went up, that's not that surprising, right? It's
00:27:44
not inconsistent. >> It's a bit surprising. I I'm personally surprised. Um
00:27:51
>> but the yeah it's a combination of things that offense is gets a you know
00:27:57
there's more confidence in an offense. There's it's um if you were given the
00:28:02
choice between being the best team on offense and the best team in defense on the in the league. You would choose
00:28:07
offense just because the distribution of offenses is is wider. The be the number
00:28:12
one offense is always better than the number one defense. Um so it's it there
00:28:17
there are some sort of intricacies uh whereas you know if you if it's like a basically a 40 to 40 game um and both
00:28:27
the Bills and the Ravens offenses are um kind of bumped up more than their defenses are bumped down
00:28:36
and we again we're looking for the differential of that particular play right so so you know really what we
00:28:42
would be talking about is their change probability if they had won versus Yeah, I I I think for that particular uh
00:28:50
situation, you know, you could imagine that yes, Baltimore still went up in Super Bowl odds even after losing, but
00:28:56
they probably would have gone up even more if they had executed won and, you know, gone for it on fourth
00:29:03
down, executed and won. >> Well, yes, if they had the win in hand, absolutely. just they're just sitting,
00:29:12
you know, in a better perch. Um, but the FBI model doesn't take the the win isn't
00:29:18
anything special. There's no bonus for actually winning. I I I've built models
00:29:24
that that are hybrid that do both sort of EPA per play like an efficiency. >> But Brian, the Super Bowl prediction
00:29:30
comes from a simulation, right? Like a forward simulation. So there it would definitely weight heavily the fact that
00:29:36
they have a win when trying to do the project. and you presumably condition on the actual outcomes in that sim.
00:29:42
>> Yeah. >> Yeah. They're completely separate enterprises. You're you're estimating
00:29:46
power models and then you're putting them through a simulation and these are
00:29:49
completely >> that's what you guys have talked about Kate for a long time about using the
00:29:52
Massie Pbody. I forget the name of the platform you've talked about but uh you
00:29:56
know >> yeah you can you can we we we there some sims are better than others and we've
00:30:01
always massive people body always had I think an advantage of having a good sim because we we put a lot of uncertainty
00:30:06
into it. Um, yes, and that's the key. That's the key, right? So, Brian Bryant
00:30:11
does that, but he's saying within his power ranking model, it doesn't care if
00:30:14
you actually put in more uncertainty than your It's kind of weird like if you
00:30:21
I should say it's surprising because if you put in like the mathematically proper amount of uncertainty,
00:30:27
then you get a result that's still overconfident. It's still under It's
00:30:32
still overcalibrated to >> No. Yeah, that's right. You have to >> you actually have to fudge it to to
00:30:38
you're kind of like where is this uncertainty extra uncertainty coming from?
00:30:42
>> But your whole but your whole game is to match the historical spread, right? So
00:30:46
you have to parameterize whatever you need. You got to get in there to match the historical spread. Hey, by the way,
00:30:51
let me let's let's take a moment since we're talking about uncertainty and
00:30:54
we're at the very end of this conversation with Brian. I want to I want to take a moment to to celebrate a
00:30:59
few examples of c of of of of communicating uncertainty because it's so important and so hard and there just
00:31:07
aren't good examples of it. So, we had um Russo on our show um last was it last
00:31:16
week talking college football? Three, no, two weeks ago talking college football.
00:31:19
>> And he did his um his his playoff picks a very particular way. And then I
00:31:25
noticed that Barnwell did his playoff picks the same way, >> which was not like expected value
00:31:30
maximizing, but rather it was representative. And so both of them, so Russo, when he went to his playoff
00:31:36
picks, he said, "Look, I looked historically. I said, "In the preseason
00:31:39
polls, how many how many of the top five usually make the playoffs?" He said,
00:31:43
"Four out of five." So, I got to kick one of those guys out when I make my
00:31:47
playoff picks. And then of the next five, how many make the playoffs? He said, "Two." So, I picked two. And then
00:31:51
he goes on down. He says, "How many who are outside the top 25 historically make
00:31:55
the playoff?" Three. So, I got to pick three teams. He says he has to just to
00:32:00
be entertaining. He's This again is not an expected value maximizing bracket.
00:32:05
This is a representative bracket which I want to celebrate because what he's
00:32:10
doing is communicating to the reader how much uncertainty there is. It turns out
00:32:15
Bill Barnwwell does his playoff picks exactly the same way I just read him. Before Brian answers that, what I'd
00:32:21
rather him do, by the way, as being a pure basian here, is sample from that historical distribution, compute some
00:32:28
sort of bracket, do it many, many times and average over that as opposed to just
00:32:33
saying the mode is three or so. And I'm just saying what I would like him to do.
00:32:37
Who cares what I want? I'm just telling you. >> I I'm just trying to celebrate the
00:32:42
communication of uncertainty because his because it's really hard to do. And so
00:32:46
anytime it happens, well, it's notable to me. I want to give you one other example. We talked about Connley. Bill
00:32:52
Connley is doing a new thing this year which I think is freaking fantastic. We do a version of it on our show
00:32:57
periodically, but he's nailed it. He's saying, "Look, let's pick four long shot
00:33:01
games and construct the forsome such that in expectation one of the long shots is going to win." And so I think
00:33:11
each week he's going to in his weekly column say here's four, here's the
00:33:14
quartet or whatever he calls it. and he picks four. So that the expected probability of a long shot winning a
00:33:21
game is 0.5. And it's fantastic because he says look you know any one of these
00:33:25
you think no way but probability suggests that on I'm going to build it so that it's half and half that we'll
00:33:32
actually get one. And again I think this is terrific communication of uncertainty
00:33:36
which is a very hard thing to do. >> Yeah. No chalk is boring. Um uh you you you you know that you know
00:33:47
one out of these top five teams is not going to make the playoffs on average. You just don't know which one and you
00:33:54
don't know which team to replace them with. So if if you really want to kind
00:33:58
of maximize Yeah. you you keep the the top five in there and that's no fun. That's not That's not interesting. As as
00:34:05
dry as I am with all these FBIs and fourth down computations and just pure numbers and everything, I'm I'm a fan,
00:34:17
too. And I completely endorse completely endorse that. It reminds me of a strategy you need when with uh
00:34:25
brackets for the NCAA basketball tournament. If you you can maximize you can do just you can have the best model
00:34:33
in the world like you know an FPI or BPI rather you know but but times a million
00:34:39
with the smartest AI and all that stuff and you have the best bracket ever and you're you're going to come in second
00:34:45
place because you know the the rando in the cubicle down the hall from you um at
00:34:53
at the office is going to have a lucky one. there's just going to be one if your league is big enough there's going
00:34:59
to be somebody out there that's just kind of a little bit lucky and so we used to enter the you know like the
00:35:03
Kaggle basketball competition stuff the way we would do it we would have our BPI
00:35:08
kind of chalk bracket and then we would pick a couple and maybe we would enter multiple of
00:35:16
these brackets each time picking a different kind of upset and that was the way actually to win those competitions
00:35:22
is but you got to be lucky too >> what Kade said is really interesting to me because I remember um both in a
00:35:31
sports context and in in my dissertation case and a survey context, we did exactly what Kade did. We said when
00:35:38
simulating data, what should I condition on? Like in other words, should I condition on
00:35:44
there's going to be at least three upsets? Should I condition on there's
00:35:47
going to be at least four out of the top five teams, but no more than that? Like
00:35:50
it's a really interesting exercise to think about what features of the history
00:35:56
do you want to bring in to your simulation. Those are priors, but they're priors of a different kind. And
00:36:03
I really think that's a fascinating and I'm glad K brought that up. It brought
00:36:07
back some fond memories. But it's an interesting question about what aspects
00:36:11
of a picking from a as you said NCA pool or what if you're simulating future NFL
00:36:16
outcomes. You know, we've always talked on this show, roughly half the teams in
00:36:20
the NFL playoffs don't make it. Do you want to condition on that? What do you
00:36:23
do? I I think that's fascinating. >> It's It's so hard. It's so It's It's
00:36:27
something we come back to every football season this time of year. It's like what
00:36:31
the that especially like consider, you know, we're going to have random teams
00:36:36
in the college football playoffs. We're way too sure we know what teams are going to be in there. And we have to
00:36:40
somehow we have to keep learning that lesson over and over. And hopefully if you're paying attention, you do learn a
00:36:46
little bit over time. Hopefully at our best, we're walking away from these conversations a little bit more of the
00:36:51
uncertainty in the world as a result of them. Hopefully. Brian, we're going to
00:36:55
have to let you go, man. We've kept you longer than we expected to. Always a
00:36:58
pleasure. Please let us get you back in here before the season goes away. We always enjoy talking to you.
00:37:04
>> Yeah, thanks for having me. >> Absolutely. Brian Burke, longtime friend
00:37:09
of the show. He is with ESPN. You can see his work there. He's behind the scenes on a lot of important work that
00:37:16
they do. And every now and then we get him out in front get him front of house as well because he's always fun to talk
00:37:21
to. Brian Burke, welcome back to Wharton Moneyball. Welcome to the second half of
00:37:26
this week's show. An open lines segment. What we used to call open lines, open
00:37:31
topics. Just off the phone with Brian Bert from ESPN. Always catch Brian early in the season. Find out what his
00:37:38
offseason project has been. find out how he's taken in the NFL. He's a diehard.
00:37:43
He didn't he didn't come clear as much, but he's from the Northern Virginia,
00:37:48
Annapolis area, and he's a diehard Ravens fan. Um, and uh, so he was talking dispassionately about something
00:37:54
that is really kind of killing him, guys. Anything else coming out of that conversation or about the NFL in
00:38:00
general, week one? Um, you know, we had we had the Eagles kicking us off with the Well, heck, Jaylen Carter kicked us
00:38:07
off. That was a hell of a beginning to the season. Um then we had games like all weekend
00:38:13
Chargers took down the Chiefs on Friday night. Um we had a big great Monday night game last night with Minnesota
00:38:19
coming back. I mean what a debut from the QB out of Michigan for those guys. They go down big pick six even and then
00:38:26
21 points in the fourth quarter. I mean I completely everybody had given up on those guys. So exciting for Minnesota.
00:38:34
Um what else jumps out to you about NFL week one? >> Just for me just a couple other things.
00:38:39
um you know, under the right scheme and under look, Aaron Rogers can still play,
00:38:44
right? I mean, threw four touchdown passes. I watched a lot of that game. He looked pretty good. Now, again, he's
00:38:50
still immobile. If you rush him, he's still going to make some bad decisions.
00:38:54
He's made that his whole career, even when he was more mobile than he is. But,
00:38:58
you know, um I could see now why the I'll call it the conservatism of Mike Tomlin together with when you really
00:39:08
need it from Aaron Rodgers. I'm upgrading my belief on the Steelers this year. The problem is they're in the AFC
00:39:15
and I there's no chance to put them above Buffalo or Baltimore or really or Kansas City right now. But I can see now
00:39:23
why the combination of Tomlin's coaching style and Rogers could work effectively.
00:39:28
So that impressed me. I think I Oh, sorry. Let's stay with that one. Shane, you want to go ahead and
00:39:33
>> Yeah. I just I was going to kind of ask you both whether you feel like, you
00:39:36
know, because we're fresh off of Brian Burke talking about how offense is stickier maybe or more predictive
00:39:42
retweet than defense. would concern me about the, you know, the Steelers kind of, I guess the Steelers like long-term
00:39:49
chances is more that their defense led up that many, you know, this is supposed to be an elite defense and, you know,
00:39:55
maybe Justin Fields just had a great game, etc. But, you know, >> if if if that, you know, if that was
00:40:00
just kind of a a oneoff kind of performance from the Steelers defense, I think they really are a contender.
00:40:07
If they've taken if they've taken a step back, even like a marginal improvement
00:40:12
that Rogers brings is not going to be enough to really Yeah. as you said, push them into the conversation of real AFC
00:40:17
contenders. >> No, I think you would agree, Shane, if it if I told you right now at the end of
00:40:21
the season the Steelers end up with a top 10 offense, you'd be like, "Whoa,
00:40:26
that team could do something." >> Cuz I'd be assuming a top 10 defense
00:40:30
just cuz the Steelers Exactly. So, that was one team. Yeah, look, obviously, um,
00:40:35
you know, with the Colts, wasn't it Danny Dimes, right? Played for the Colts
00:40:39
and, you know, the thought was I didn't see a lot of the game. I saw some of it. I was
00:40:45
watching on a lot of screens. Um, and so, um, he looked good. Now, it could be the team they were playing
00:40:54
was bad, but all I'm commenting on is um you know, this idea that he can't play
00:41:01
good football is just not true. Now, that doesn't mean he won't have 16 other
00:41:06
bad games, but I'm a believer, you know, same way with academics when I evaluate
00:41:10
people for tenure. Shane, if someone can run write one great paper, maybe you could do it again. So there's no reason
00:41:18
for me to believe that Daniel Jones under the right circumstances can't have, you know, maybe he's a Sam
00:41:25
Darnold like season. Why not Sam Darnold? >> I I don't think we can write off any
00:41:31
quarterback based on kind of mediocre performance in a single system, especially if that system is a New York
00:41:39
based system, it seems like, you know. So no, I agree. Sam Darnold, Eugino Smith, and we have countless examples of
00:41:46
these of of quarterbacks kind of, you know, looking mid or worse, not really kind of being legitimate starters in
00:41:52
their first goaround and and, you know, they get in a different system with different personnel and an offensive
00:41:58
line. Um, and and and watch them do something. I I don't know, that's not
00:42:03
really a statement about whether the Danny Danny Dimes and Indie is going to be a sustainable thing. Uh, or whether
00:42:08
they're, you know, going to be at all a good team, but >> find out a little bit. Yeah, we'll find
00:42:13
out. >> He's definitely behind a better offensive line. >> Just quickly, we'll find out a little
00:42:16
bit this week. It's Broncos at Colts and the Broncos are at least a good team if
00:42:21
nothing else. So, we'll find out something. >> A couple of the things about next
00:42:25
weekend. Um the Philly KC game. I mean, that's exciting. Uh give us a Super Bowl
00:42:30
rematch in case. >> Can't wait for the highlights. The highlights are going to be so fun to
00:42:34
watch. >> Can we say if the Eagles beat the Chiefs that the Chiefs are in quotes whatever
00:42:39
this means, trouble? Well, two games down. I mean, even if you're still a good team, that's a hole.
00:42:45
>> Yeah, team teams dig out of it. I mean, I think Baltimore started 0 and2 last
00:42:49
year and dug out of it. But, >> but still, that's you don't you it's not
00:42:53
a great spot to be in. Other ones that jump out to me, um I'm really curious
00:42:57
about Washington and Green Bay. I mean, Green Bay looked so good, but with Jaylen Daniels, I mean, it's a good test
00:43:02
for Washington. And then here's a sneaky one, guys. I just noticed it. I was
00:43:05
going through the schedule. Atlanta goes to Minnesota. Atlanta lost a heartbreaker to to to Eric's team last
00:43:11
weekend. >> I was so upset about that. >> Yeah. Here's the thing about Atlanta
00:43:14
Minnesota. The quarterbacks in that game are the quarterbacks from the NCAA final
00:43:19
two years ago. So, Michael Penn was the Washington quarterback when Michigan's
00:43:24
JJ McCarthy won the national championship in McCarthy's senior year. And those guys will be facing each other
00:43:30
again in Minnesota Monday night. >> Monday night. >> I watched all >> I'm sorry. Sunday night.
00:43:35
>> Sunday night. Sunday night. I watched all of the Bucks Falcons game just to
00:43:39
let you know. And I'm going to say this right now. I'm going to say the same
00:43:42
thing I said about Jaden Daniels last year when I watched Washington against the Eagles in week one. I'm sorry, the
00:43:48
Bucks again, sorry, Washington against the Bucks in week one. You're not going
00:43:52
to want to play Michael Pennock Jr. very soon cuz I'm going to tell you something, that guy can play football.
00:43:59
He's got great decision making. He's got a great arm. I mean, he made some throws
00:44:04
no more than five quarterbacks in the NFL can make. Michael Pennix Jr. can really play. Falcons were a good team.
00:44:12
They were right there with the Bucks, man. I mean, it was a great football game.
00:44:15
>> Well, Bjon scored on that first uh drive as well, which is like Oh, yeah.
00:44:19
Remember, they've got Bjon Robinson, which is a good place to be as well. >> Don't they have the highest paid? They
00:44:24
must have the highest paid backup in all of football, >> too, as well, right? I mean they, you
00:44:30
know, >> right? >> Okay. So, um, real quickly, I just a note that ESPN put up today that is
00:44:37
striking is how much kickoff returns are up. And it's fun. It really is fun. You
00:44:41
forgot how much fun we lost with. So, guys, did you realize that we had dropped down at the low in 2023? The
00:44:49
season was the average 22% of kicks were returned in 2023. Last year when they changed the rules, which is great, but
00:44:57
they had the the the the you could get the ball at the 30 if you put it in the end zone. We only went from 22% to 33%.
00:45:05
So, they got a lift, but only up to 33% this weekend when the penalty was bringing the ball out to the 35. We saw
00:45:12
70 76% of the kickoffs were >> Kate, I'm not You're not making me reflect. I watched 10 different games
00:45:19
this weekend, maybe more. I don't remember what I'm not saying what didn't
00:45:23
happen. I would have told you was 98%. I don't remember one ball kicked into the
00:45:27
end zone and I watched every a as many games as I could. I don't remember one.
00:45:32
>> Yeah. And they're they're even higher variance I think than they than they
00:45:35
have been in the past. Um it's it's a it's an major props to the NFL for
00:45:39
playing with the rules like that. really gives it gives kickers kind of I mean we're already seeing I feel like elite
00:45:45
like like like kickers kicking 60 field yard field goals routinely and stuff like gives kickers another thing to like
00:45:51
kind of become elite at to to kind of work up and land right on the one yard line
00:45:56
>> drop it right at the five. >> Okay, one last note on football before
00:46:00
we change football. We do have college football. It was a bit of a quiet weekend, fun weekend. It would no huge
00:46:06
games except for Florida going down to South Florida. Um, but next weekend, got a couple notables just to kind of sell
00:46:12
you guys. This is my weekly note to sell y'all on some college football. Clemson's going into Atlanta to play
00:46:18
Georgia Tech. And they're only four-point favorites. And this is kind of a test on Clemson because they lost
00:46:25
game one. And they looked bad last week against Troy. They were down like 16 nothing at halftime. They won.
00:46:31
>> People are skeptical on Clemson. Georgia Tech's solid. They're only four-point
00:46:35
favorites. That's going to be interesting. the big one. Um, Florida's going to LSU. People were more
00:46:42
excited about that before Florida lost their nine-point dogs. But there's a big
00:46:46
SEC game. Let's see if Florida can bounce back. Texas A&M going up to Notre
00:46:51
Dame. A night game in South Bend. 6 and a half point underdog. >> You just skipped on. You put these on
00:46:57
here. You just skipped the most interesting game in my mind. You know what it is. You know, USF at Miami, man.
00:47:04
>> Well, it is USF Miami. Wait, wait, wait. Let's be clear. >> This is because USF beats Miami, you got
00:47:10
to put them in the playoffs. >> No, they've beaten two ranked teams already, Shane. If they beat Miami, come
00:47:15
on. They're in. >> No, I forgot about two or week three or whatever. >> Favorite saw is is Group of Five. Group
00:47:22
of Five. I forgot about >> usually doesn't kick in until like November, but no, I I I like it. I like
00:47:26
it. It can really jump. You're always ahead of the game, Eric. Yes. Well, so
00:47:30
so undoubtedly USF is one of the stories of the season so far, but they are 16point underdogs against Miami. So
00:47:37
let's let's keep our powder dry on that one. But it would be fun >> if they win.
00:47:41
>> Yeah, it would be fun. Okay, that's it for football for the week. That's a lot
00:47:45
of football. Of course, that's where we are early season working some things
00:47:48
out. We had some other major sports in particular the fourth and final tennis championship, major championship, US
00:47:56
Open, of course. um both sides of the bracket, but I I I didn't watch, but I
00:48:00
have to say I was surprised. Not only did Alcarez beat center in the final, but he beat beat him in four sets, and
00:48:06
they weren't even particularly close sets, at least some of them. So, Eric, give us the update here if he's how is
00:48:11
it that center is the best player in the world, but Alcarz has beat him like six
00:48:15
of the last seven or something. >> Seven of the last eight, actually. >> Seven of eight now. Okay.
00:48:19
>> Yeah. So, in the last two years. So, well, first, Alcarez now is number one
00:48:24
again in the world. So, let's just be clear about that. He took over the points lead. He is number one again. Um
00:48:30
look >> well to be clear the way we talked about center all summer was that he was the
00:48:33
next thing to freaking Jesus Christ. And so this is a little bit surprising to see to see this to see this go down.
00:48:40
>> Yeah. Or I guess to kind of more contextualize it, how often do you see
00:48:43
like two kind of players at their kind of elite Pete showing such an like an eight eight out of nine kind of
00:48:49
imbalance in sort of matchup? >> Good. Yeah. So there is there's an imbalance there. Look, I think most
00:48:56
people tennis experts, you know, I know uh one week we're trying to get them on.
00:49:00
We had him on. We had to switch him off. We're trying to get Paul Anacone on.
00:49:03
Would be great to talk about this. I think what's been shown over the last two years. For now, tennis is not is
00:49:10
non-stationary, but for right now, Garez's top end game is better than S's
00:49:17
top end game. It does not mean center can't beat him. It does not mean center
00:49:21
doesn't have a better record against common opponents. He does because his game has much lower variance. But
00:49:30
Alcarez knows the importance of every single match against Sinner and he's going to bring his top end game against
00:49:39
S. And look, he's seven and one against S. That is an imbalance. Look, when we
00:49:45
look back on their career, Shane, this is what we do right now. Even though their ears were a little different.
00:49:50
Djokovic has a slightly winning record against Nadal. Djokovic has a slightly winning record against Federer. Djokovic
00:49:57
has a winning record against the other big two. It matters when we look back on Sinner and Alcarez. I'm not saying S
00:50:06
can't get there. It's now I it's either 10 and five or 11 and five. I don't
00:50:09
remember which one it is. Alcarez has a 10 to5 lead I think it is over S. and he's got six majors and center has four.
00:50:20
And if Alcarez wins the Australian, which he says is his number one goal for next year, he'll be 22 years old with
00:50:26
the career grand slam. Do you uh when you think about kind of in in your kind of personal sort of rankings of kind of
00:50:34
great going up against greats within a particular era, do you think more about total grand slam titles or do you think
00:50:41
more about the personal head-to-head? I I think about I not I think about the peak. I think about when the person was
00:50:47
at his best because here's the problem with comparing Federer to Nadal. >> It's they're not aligned exactly but not
00:50:53
really. Federer is five six years older than Djokovic and so they're headto-head. I'm actually surprised I
00:50:58
happen to know Federer's record against Djokovic. I'm surprised he's even he
00:51:03
didn't have a winning record, but Feder Jookovic was 27 and 23 against Federer.
00:51:08
And so that's a winning record, but it's not that winning a record. And for five
00:51:12
of those years, Federra was definitely not at his prime. Um, I think of the I if you want to be and we know ELO has
00:51:20
its problems. I tend to look at peak rating when the person was at their best. Could you beat them? And I say the
00:51:29
same about center that I say about Djokovic. Djokovic was is not the greatest player I've ever seen. He's the
00:51:35
GOAT. He's the most accomplished player. There's no debating it. But he's not the
00:51:40
greatest player I've ever seen. If Feder has his best day, he won. If Nadal had
00:51:46
his best day, he won. I saw Stan Roinka blow out Djokovic on his best day. But those best days come very very
00:51:54
infrequently. And if you're not at your best day, you had no chance against Djokovic.
00:51:59
>> Okay. So, I just want to note that you're you're you're making you're
00:52:03
there's a distinction here between their peak EO ELO rating and what you're
00:52:08
talking about their best day because you're saying even within an era when you'd have an ELO rating on a guy, you
00:52:13
still have a distribution of performance and you're talking about that right tail
00:52:18
that when the guy's at his best, how good is that? And we don't have a number
00:52:22
around that. And it'd be nice to have it'd be nice to have that assessed in
00:52:25
some way. But I I was struck in the I was recalling our summerlong conversations about center and Alcarez
00:52:31
and this characterization that some quants some quants in tennis have put numbers around it and have validated
00:52:36
this this there is a higher floor and a lower ceiling for center. But what but the way you talked about it Eric is that
00:52:43
it's almost as if Alcarez he isn't taking a random draw from his distribution when he gets in in these
00:52:48
matches against center. He's somehow able to shift his draws to the right side of the district.
00:52:54
>> It's not symmetry isn't symmetric. He's somehow strategic in terms of when
00:53:01
also the the I don't call it the rumors. Um >> Alcarazz's coach, Juan Carlos Ferrer,
00:53:07
former number one in the world by the way, said that they worked the entire summer on making Alcarz more consistent
00:53:15
and raising his floor. Alcarz lost only one set that was to sinner in the finals
00:53:22
the entire US Open. Had he gone under had he won all three sets, he would have been the first person in the open era of
00:53:29
men's tennis win the US Open without >> incredible that he lost one set to
00:53:34
center and that happened. So he didn't go on. But just to let you know, I think
00:53:38
we're going to see this is what's scary for men's tennis. I think we're going to
00:53:43
see the same peak from Alcarez but a higher floor. And that's even scary. That's really scary. And by the way, the
00:53:51
other stat that I found interesting is the last two years um Yannik Center against everybody else in tennis but
00:53:59
Alcarz is 169 and4 and he's 1-7 against Alcare. >> I mean that is >> that's interesting. That's that's a
00:54:11
separation of one two big separation. But this but and let me just say I think we'd also want is you know we're now at
00:54:19
eight and what I mean by eight we're at eight straight majors >> where it's been Alcarz or Sinner. Let me
00:54:25
just give them both credit. It's four to four. They both have four of them. Now
00:54:30
how many in a row are we going to go where one of them wins? That's I mean could it be another I'm
00:54:39
making it up is another five years. >> What's No, we won't go five more years.
00:54:43
But what's the longest run we've seen with two guys taking all the slams?
00:54:47
>> I think this is it now. >> Eight. >> Cuz remember there was the big three.
00:54:52
>> Yeah. And how many? That was >> No, there was the big three and there
00:54:56
was Snampis and Agassy and you know Jim Courier won one every now and then and there was Michael Chang won one and you
00:55:03
know and even during the you know the Borg Mack andro Connor there was always you know more than two and so I think
00:55:10
this is the longest. It would not surprise me if this grows at least to double digits like why next year would
00:55:16
you predict anybody but one of the two of them? >> This is a good this is a good overunder.
00:55:21
Eric, what do you give us an overhead or me and Shane will take? We we'll bet.
00:55:24
Give us a number and we'll take >> Okay, we're at eight right now.
00:55:26
>> Yeah, >> I am going to go 13 and a half. >> What do you want, Shane? So, that's
00:55:36
another year and a half basically. >> Not basically. That's it. Next year and
00:55:40
a half, >> I guess. >> Before you pick before you pick, let me say why I picked that number,
00:55:46
>> Eric. 14 and a half would be a year and a half. No. >> Oh, yeah. I'm sorry. You're right. Six.
00:55:51
14. I meant four. Then I'll go with 14 and a half. Let me say why I'm going
00:55:55
with that number. Like a lot of times you could say, well, you don't even know the set of potential
00:56:03
players. I do. Like no one's coming up from the bottom. That's just like I
00:56:09
don't know about now. There's no 17-year-old. Like it's gonna have to be
00:56:12
someone like a Ben Shelton or it would have to be someone like a Felix Aliim. It's gonna have to be or is vera finally
00:56:21
gets his act together for seven matches or Medvadev has some resurgence. Like they're so good. It can't be some random
00:56:30
17year-old that's now eight. It's just not it's not going. So it's it's within
00:56:34
the set of players we have. So I'll go with 14 and a half. Thank you. 14 and a
00:56:38
half. That's my overunder. That's my rationale. >> Shane, what do you got? What do you got,
00:56:42
buddy? I guess I'll I'll I'll I'll >> I'll I'll take the under just to be
00:56:46
interesting and because we just had a conversation with Brian Burke about uncertainty.
00:56:51
Um good for you. >> And I guess and I guess you know what I what what what would make it happen? I
00:56:56
my mechanism I guess is maybe you know all takes a one or two injuries and all of a sudden it's you know I mean that
00:57:03
that's probably that's probably one of the biggest factors in this is both of
00:57:07
these players or at least one of them staying healthy long enough to kind of keep this streak going. My viewers, they
00:57:12
better both be injured because if you're playing one of them right now, you know,
00:57:16
you it does sort of sound like you almost you need both an injury and an impressive right by
00:57:23
I'll get this stat and I'll post it on W Moneyball. >> I think this is the largest gap in the
00:57:29
history of men's tennis between number two and number three. S has like a 4,950 almost 5,000 point
00:57:37
lead on the number three player, which I think is still Zerv. And just to let you
00:57:42
know, you get 2,000 points for winning a major. And he has a 5,000 point lead on
00:57:47
number three. So just to let you know how big the gap is to norm it, it's big.
00:57:51
And Shane, which Kate, which one are you taking? >> I'm going to go under also. I think it's
00:57:56
the only reasonable way to go. I think we're supposed to go that way. It's
00:58:00
almost like, you know, in golf, Shane, I don't know if you play enough golf, but
00:58:02
Eric, I think, does. There's the amateur side of the cup and the pro side of the
00:58:06
cup. Amateurs tend to play too little break. They're often below the cup at the end of the putt. I feel like this,
00:58:13
you know, we're supposed to be professional statisticians. Some of y'all actually are professional
00:58:17
statisticians. I think you got to take uncertainty. The things that we're outside the model. And so, I'm I'm going
00:58:23
to go I don't love it and it's not fun, but I'll go under as well. >> I'll tell you what, I don't This is a
00:58:26
great discussion for statistical topic. I don't think we have as good enough
00:58:30
intuition as to what I'll call the distribution of the maximum of multiple players cuz remember let's assume that
00:58:38
let's eliminate changes for one second injuries. Now we have to talk about we
00:58:43
have two people who have this much distance to number three. You're right there's variability but now I'm taking
00:58:49
the maximum of the two cuz you know that I don't think we have really good intuition for. As a matter of fact, you
00:58:56
could argue that's even a farther exceedence than we have intuition for, which was why I'm comfortable with my 14
00:59:03
and a half. >> You can imagine, Eric, to take that further, like if you put three in the
00:59:07
set or four in the set, we probably have really bad intuition because you probably
00:59:11
more of 68 majors going to the big three over a 17year period. >> Exactly. That kind of thing. So, it's a
00:59:18
fun it's a fun little problem, actually. All right, guys. We're down to just a
00:59:22
couple of minutes before we should wrap this thing up. But let's talk baseball.
00:59:25
I mean, it's been a quiet week in my mind mentally because there's been so
00:59:29
much football, but there have been baseball games being played. What's going on over there? What do we need to
00:59:33
be paying attention to? >> Well, we had a couple uh It's worth noting that this is uh continued to be
00:59:39
We haven't had a no hitter yet this year, but >> after average, we we've come so very
00:59:44
close over the last couple weeks. This is I'm so I'm so glad you put this in
00:59:48
the rundown because like I there have been no no hitters, but >> can we actually just basically then say
00:59:56
that even if this year ends with none I mean we had what one to eight and two/3 innings was were they both to eight and
01:00:02
two/3s innings? We should have like a no hitters in expectation or something extra stat that kind of like gives us
01:00:07
like 0.98 of a no hitter or something >> or a distance metric like suppose I told
01:00:12
you we have none this year but we had 10 no hitters through 8 and 2/3 you'd be
01:00:18
like oh all right that seems reasonable. So I like this hard cut right at zero hits like
01:00:27
>> it. I understand what a no hitter is, but I'm just saying >> I know it's like weird it's a
01:00:33
particularly weird discretetized >> consequence of what we're generally
01:00:37
seeing across baseball is is increased offense um a little bit, you know, and and and so another kind of example of
01:00:46
that kind of discretetized, you know, consequence of that is we still got uh four players now on track to hit over 50
01:00:52
home runs. And that's only happened twice before in MLB history. So, you know, if you know, Judge unfortunately
01:01:00
has dropped off a bit because he's, you know, he hasn't been playing. Uh, but
01:01:04
Rally, Schwarber, Otani, and Suarez are still all on pace for over 50 home runs.
01:01:08
And if they hit that, that's something that also has not happened in a long time.
01:01:12
>> When has that happened before? You said twice before. What era? >> It was uh n I looked up 2001. So, both
01:01:17
were steroid era. >> That's what I figured. Okay. >> 1998 and 2001. The last time it
01:01:22
happened, just to name the guys in 2001, it was Alex Rodriguez, Juan Gonzalez, Sosa, and Barry Bonds.
01:01:28
>> That's all steroid. And how about 98? >> 98. Well, okay. So 98. Uh
01:01:34
Boon, Ken Griffy Jr., Sosa, and Magguire. >> Steroids. >> Half Well, nobody nobody's Ken Griffy
01:01:42
Jr. steroids here. >> Alleged. In case anyone's listening, alleged. >> But yes, this would be the this would be
01:01:48
the first time outside of the steroid era. We could call that the peak that >> I prefer to think about that. That was
01:01:55
that was my grad school Chicago grad school era. It was a good time to be a Cubs fan. It was a good time to be in
01:01:59
Chicago. >> Oh, people people slam it down, but it was an exciting time to watch baseball.
01:02:04
I mean, we got, you know, we got what we kind of like. >> It's another one of those interesting
01:02:08
stats, by the way, cuz we've never had more than four, right? obviously and um
01:02:12
you know there must have been many years in the 20s where I'll make it up Babe
01:02:15
Ruth and Hack Wilson or Babe Ruth and this person but like Garrick never hit 50 in a season. People don't know this.
01:02:22
Aaron never hit 50 in a season. >> That's what people criticize Aaron for.
01:02:27
You know I consider him the home run leader but whatever. Um he never hit 50. May hit 50 in a season once in his
01:02:33
career. So you know let's not make it seem like hitting 50 is that easy. It's
01:02:39
not. I think it's notable. I think what we're seeing here is kind of the
01:02:42
combination of in some rule changes that have increased offense and and and you know helped hitters over pitchers in
01:02:48
general plus the fact that you know we've got like a generation of like kind
01:02:53
of orientation towards home run hitting maybe you know the three outcomes whatever that I I think is kind of
01:03:00
driving a lot of this. We have a lot more home run hitters now than we you know had in the early years of baseball.
01:03:07
>> Yep. I'm I'm always All right, guys. Well, it's only going to get more interesting in baseball over
01:03:12
the next couple of weeks as we come down to the wire, playoff races, etc. We'll
01:03:17
look forward to it. All right. Uh, why don't we wrap it there? Been a full hour
01:03:21
here on Wharton Moneyball for the whole crew. Shane Jensen been in here the whole time. Eric Bradlo been in here the
01:03:28
whole time. Audi Winer in Absentia. This has been Kade Massie. Many thanks to Dion Simkins who I think we named the
01:03:34
fifth Beatle today. D Patel, Boss Lady, and our producer, Marissa Rain. I appreciate all y'all do and thank you
01:03:42
guys for listening. Come back and join us next time. Between now and then, enjoy your sports.
01:03:50
[Music]