00:00:00
Welcome, welcome to Wharton Moneyball. Welcome to a full hour of sports analytics
00:00:06
here on the Wharton Podcast Network. This is Cade Massey hosting this week with my two longtime collaborators and friends,
00:00:12
Shane Jensen and Eric Bradlow, both of those fellows in Philadelphia. I am remote as I typically am these days.
00:00:18
We are recording on Tuesday afternoon. This show will go up on Wednesday. We're gonna follow our usual schedule,
00:00:25
the schedule for the last couple of years anyway. Gonna go with a guest in the first half hour,
00:00:29
open it up for broader topics in the second half hour. We have a brand new guest this time.
00:00:36
We're often welcoming people back and we're always happy to have people back,
00:00:39
but we need to have new guests as well. We have Tyrell Stokes joining us this week.
00:00:44
Tyrell is from the hockey world. I happened to be hanging out with some hockey types
00:00:50
a few weeks ago and they said, you gotta get Tyrell Stokes on the show. And so, okay, here we go.
00:00:55
I listened to these guys. Tyrell Stokes supposed to be about as sharp as it gets
00:00:58
analytically in the world of hockey. Tyrell's coming to us from Teamworks. Before that, Zealous.
00:01:06
You've heard us have lots of guests through here on Zealous before they were acquired by Teamworks.
00:01:11
They work with teams, both professional and college level. Hockey's gonna be much more professional level.
00:01:18
Tyrell's background, he's from the Canadian Prairies. We just learned he's from outside of Edmonton.
00:01:23
So he and Shane are gonna be going at it about their Alberta loyalties here before the day's over.
00:01:31
Tyrell's a statistician by training. He has a PhD in causal inference from McGill,
00:01:35
another connection to Shane. Postdoc at NYU, he's done academic research, hospital network data, that kind of thing.
00:01:42
But he's been doing hockey for these last few years. He's up there in Brooklyn, New York.
00:01:48
Tyrell, we're delighted to have you. Thanks for making time for us. Yeah, really excited to be here.
00:01:52
Why don't we start off by just asking your reflections on the Stanley Cup playoffs so far?
00:01:58
We are almost two rounds through. One of the four series is done. We've seen some hockey, having some winnowing.
00:02:06
Just what's top of mind? You as a hockey fan, you probably think analytically anyway,
00:02:10
but anything specific, just like as a fan of the sport, what have you noticed? What are you paying attention to
00:02:15
over the last couple weeks, last few weeks? Yeah, you know, good question. Yeah, we're definitely in this kind of,
00:02:22
yeah, it's interesting the timing of the series. Some are still going, some already finished,
00:02:26
like some really big breaks. That's obviously interesting. I think like, you know, NHL playoffs
00:02:32
is really one of the cruelest of like the playoffs series across sports. Like just being, you can be really good
00:02:41
and you might not win, you know, like whatever, like maybe 20% of the time, the best team is gonna win or something like that.
00:02:48
Well, and so that's been really tough to see. You can see some series where, well, you got outplayed
00:02:55
and you still won or vice versa. Excited to see, like I know you guys had Eric Tolsky on.
00:03:01
Exciting to see that team breakthrough. I think they've been one of the, you know,
00:03:05
better teams in the league recently. And obviously exciting as an analytics person.
00:03:09
I think that's really cool to see a team that's clearly like stewarded by those analytical minds.
00:03:15
And I think they've just had a couple tough, I think Eric was talking about it last time,
00:03:18
they've had a couple of tough playoff series in the last few years, very good teams.
00:03:22
And so cool to see them get that kind of success. But, you know, also knowing the playoffs,
00:03:27
who knows how long, you know, that luck will go. You guys are playing really well.
00:03:32
Tyrell, let's stop there for a second. Tyrell's talking about the Carolina Hurricanes
00:03:36
and they are already through their second series. They haven't dropped a game in the tournament so far.
00:03:40
They're eight, no, I don't think there's only like five teams or something like that have opened that strong.
00:03:46
So they're looking good. The Avs were almost looking as strong. They still are really, they've dropped one game, I think.
00:03:52
They're up three, one now on the wild. They started out behind last night. I just, I was getting on the bus.
00:03:57
I was going up the steps to get on the Minnesota wild bus. And then they came roaring back and ended that
00:04:02
and that series is probably over. But the Avs, what, Hurricanes look very strong
00:04:06
out of the east. How are you feeling about the Avs? I mean, all season long, we kind of thought they were the top, right?
00:04:12
Yeah, I think, I mean, I think they're a really solid team. Yeah, like five on five, it's hard to,
00:04:18
hard to come up with a better team. I guess like the power play is one of those ones
00:04:22
where you like look at the power play and like, it feels like it should be even better than it is.
00:04:27
Just when you think of those, like the talent that they have, we obviously really like that team.
00:04:33
Really like a lot of like exceptional players, a lot of really good puck movers,
00:04:37
a lot of good puck carriers, good skaters. Tyrell, when you watch a game, after the game,
00:04:46
do you have a go-to set of analytics that you review to kind of understand what you saw?
00:04:52
I often want that. I don't know exactly where to go. Like I sat through the game six of the Penns Flyers,
00:04:58
where the Penns seemed like they just beat the hell out of them. And then they lost every period that went along,
00:05:04
including overtime. They were just really outplaying them more and more. And then they lost on a heck of a shot
00:05:08
from the blue line players. And I wanted to go somewhere and say, someone tell me how bad that was
00:05:14
that that team just lost that game. Do you have that set of analytics that you can lean on after a game
00:05:19
to kind of better understand? And what goes into that set of analytics for you? Yeah, no, that's a great question.
00:05:24
That definitely, Penns Flyers was one of those series that I was following closely.
00:05:30
We've actually, like, we have a number of like game reports that I like look at,
00:05:34
like, and that we generate sort of automatically in some cases. And, you know, a lot of it is pretty similar
00:05:40
to what teams probably use of like, you're looking at kind of overall like XG,
00:05:45
kind of like who's dominating in XG, who's dominating in shots. We have some like lower level fancier stuff
00:05:50
that's a little bit more tracking based, like who won in terms of like pressure
00:05:54
as measured by something like an expected shot. We looking at like matchups, something we've been playing with more recently,
00:06:02
and it's like a new kind of game report that I've been looking at that I've found helpful
00:06:05
for looking at games. We have like a system, we call it our like on-puck value system.
00:06:12
If you're familiar with other flow sports, it's basically like an EPV model.
00:06:16
We built this one out to be, to work on all event level games. So just so we can run it on multiple leagues,
00:06:25
not NHL exclusive. And we found that kind of system is really helpful for getting a quick snapshot of like which players
00:06:32
and which teams are dominating how. So like you kind of imagine the really important events in a game,
00:06:37
like takeaways, passes, carries, loose puck recoveries, shots, finishing, about like six or eight of these like important
00:06:45
like actions on the ice. Every one of them, we have sort of a value in terms of how much like XG did that add by this event.
00:06:53
And you can do a number of things. So one, I just look at a bar chart of like which players,
00:06:58
like put their positive contributions on one side, their negative on another side,
00:07:02
order them like among forwards, see where they ranked, each player ranked among all forwards in the game,
00:07:08
among all defensements in the game. And then you can kind of take those same categories
00:07:13
and you can kind of say like which team won face-offs, which team won shots. And then these kinds of things to kind of get a quick sense
00:07:22
of which team is doing better than others. And I found that really helpful. And in the case of the pens, a lot of those games,
00:07:29
it looked like from the underlying sense, the pens were probably outplaying the players.
00:07:32
Okay. Let's give Shane the first question. Yeah, I'd like a quick question of clarification.
00:07:38
In these calculations, are you building in actual kind of positional data, like where into that calculation
00:07:45
or is it more just based on sort of the event data of like this person had possession
00:07:50
and then this person had possession? Yeah, so we have like multiple versions of those kinds of models.
00:07:55
The one I was describing is mostly, yeah, it's meant to work on like as many leagues
00:08:02
as possible. So it's a little bit cruder where we don't have the location of all players.
00:08:05
We have the location of like the puck and a little bit of context around it. We have similar models that we're sort of building out
00:08:14
in the NHL space that is like specifically tracking based that mirror some of the stuff that we've done
00:08:19
in other sports, like say basketball, American football, international football, soccer.
00:08:25
Eric was gonna ask a question earlier on as well. Well, no, I was gonna ask Terrell how much,
00:08:31
so let's say Teamworks presents, let's say it was even a team playing the,
00:08:36
who's playing the avalanche right now? Is it the Wild, you said? Let's say the Wild are playing the avalanche
00:08:40
and Teamworks provides an analysis to the Wild that says, Colorado is killing you in the following four areas.
00:08:48
How much can a team actually adapt its style to information given to it about the strengths of the other team?
00:08:58
Or in some sense, after 82 games, after seven, eight games of the playoffs, you are what you are.
00:09:05
Or do teams actually have, I'll call it multiple versions of themselves, like we can play the fast style,
00:09:10
we can play the style that draws penalties because we know we can't beat them five on five.
00:09:14
How much do you think about that when you're designing stuff for Teamworks because you want people to actually be able to use it
00:09:20
in some efficacious way? Yeah, that's a good point. I kind of tend to think about
00:09:26
in flow sports kind of generally, there tends to be this kind of like pyramid of like,
00:09:34
you know, the lowest hanging fruit is just who's good. And then you try to ask like, why are they good?
00:09:38
And then I think if you do a pretty good job of that, then you can start going to these like more
00:09:42
like tactical realms. I would say like hockey is pretty hard, I think to be super useful tactically.
00:09:51
And there's also a bit of a complication, I think on our end of like, you know, I'm not embedded with a team.
00:09:58
So like, typically, our approach has been to try to just give teams like tools that they can do some of that more specific stuff.
00:10:07
Because a lot of them are not going to want to tell us like, oh, we were running this kind of thing on these games.
00:10:12
And like, therefore, like you're going to change it. And some of those like tactical stuff are tough to,
00:10:17
I think if we were to try to recognize, if our goal is to recognize tactics, like that would be the only thing we do.
00:10:23
So are you guys more of a licensed platform than you are a team contacts you and say,
00:10:28
please do this analysis for us? Yeah, like we, the bulk of what we have is like,
00:10:36
I think of it as like we create like foundational models that we give the teams that allow them to do more analysis.
00:10:44
We also build metrics on top of those and we provide the teams with those. But you know, our really sophisticated users
00:10:50
are using like lower level outputs from like really big foundational models of ours.
00:10:55
And they're building their own like analyses on top of that, metrics on that.
00:10:59
Okay, hold on. Foundational models and lower level inputs from foundational models.
00:11:04
So tell us, give us examples of what those are. Yeah, so like two of them, you know,
00:11:10
I guess like the simplest like suite of like that listeners might be familiar of like,
00:11:15
let's say like an XG model. So everyone is familiar with an XG model. So we tend to call what we have.
00:11:21
Expected goals, expected goals in case there's a listener that doesn't have that, okay.
00:11:24
Expected goals. We think of our, what our XG offering as a framework. So it's really probably 25 or more models
00:11:33
that are all strung together to make multiple evaluations. So if you're like most of the public XGs
00:11:40
will have like using all the information that they have in the public data set at the moment of shot,
00:11:45
this is the value of it. Ours would look like using all of the tracking information
00:11:51
at the moment of shot, this is evaluation. Now, as that trajectory evolves, there's might be deflections,
00:11:58
there might be multiple deflections and so on. And, or just before and after the puck is released,
00:12:04
those are different information states and they're gonna create like multiple XG values.
00:12:10
And then additionally, in order to make something like that, we need, let's say the probability
00:12:15
of every player on the ice redirecting the puck or blocking the puck or... So some of our clients are using,
00:12:23
let's say you can use what we'd call like a post-shot XG. This is a concept we stole from soccer.
00:12:28
So in soccer, where they have tracking data, you take before the shot, you take after the shot,
00:12:33
you use the trajectory information. The difference between those says something about usually the shooter,
00:12:39
like how talented is the shooter? How are they imparting something on the ball? We have the same concept in hockey.
00:12:46
We obviously have like metrics that we've developed to say, like who's really good at imparting
00:12:52
really good trajectories on the puck, which is like certainly an aspect of shooting.
00:12:57
So we make those metrics, but we also just give you the raw values. And some of our teams are much more interested.
00:13:02
They're like, look, like we understand hockey better. Our coaches are interested in like very particular cases
00:13:08
of like this red redirection, that redirection, this shot, that shot. We can't possibly like anticipate all of the analyses
00:13:17
or all of the things that a team is gonna want. So like our approach is just like, I'll give you that.
00:13:21
Like, I'll give you the shot level redirection probabilities. I'll give you all of these different XG values
00:13:28
as they develop across the shot. And sure, I also give you metrics, but like I want to empower like the coaches,
00:13:36
the internal R&D staff to be able to do some of those more like tactical stuff themselves
00:13:40
a little bit better. Tyrell, that speaks to a question that comes up on occasion.
00:13:46
In fact, even on this show, I think in the recent weeks we've asked who's gonna benefit the most or some show,
00:13:51
who's gonna benefit the most from AI and sports? Who benefits the most from advanced analytics?
00:13:55
Forget AI, just like zealous teamworks level analytics in hockey. It sounds like perhaps that kind of,
00:14:02
it's not the entry level folks who don't have their own staff. And it might not be the totally built out folks
00:14:09
who have huge teams doing it all themselves. Is it kind of the middle where they've got sophistication enough
00:14:14
to already be there conceptually, but they want your input so they can process, they can further refine internally?
00:14:21
Or how would you say, who benefits the most? That might be right. Like, I think like, you know,
00:14:27
it's probably like from, if I was thinking of like the, if I was trying to put like an economist hat on,
00:14:32
it's probably is like the bottom, to be honest. Like if you have basically, if you basically have nothing built off,
00:14:39
we have a bunch of like high level statistics and like, you know, things for like drafting
00:14:43
and stuff like that, that you could just plug and play tomorrow. Like going from zero to that, I think is like huge.
00:14:51
The middle, like I think does get a lot obviously, because yeah, like the, those middle groups,
00:14:58
like some of the metrics we create, like just fully plug holes for them, but then also they have like new like data sets
00:15:04
to play around with. And then kind of at the top level, like I said, like they're probably more interested
00:15:09
in these like lower level, like features or like model outputs to like throw in their own stuff
00:15:14
or as like auditing purposes, like in a sport like baseball, where we work with teams that have 40 people,
00:15:21
like internal, like sometimes they're hiring us because they want to just like make sure
00:15:25
their internal stuff's are good, right? Yeah. Right, right. How would you characterize the distribution
00:15:30
of analytics sophistication in the NHL right now? Like compared to other sports or just like across like the-
00:15:36
Just like these rough tiers that you just categorized, like roughly how many folks,
00:15:40
how many teams in each tier? Yeah, it's tough to say, like we just had, there was like a Halo conference
00:15:48
that we helped organize at the beginning of April. And- Tyrell, was that the one that the ABS hosted in Colorado?
00:15:56
Yeah. That seemed really, really cool. They worked, yeah, they worked together with SportLogic,
00:16:00
which is now Teamworks too. I was involved a little bit on the, reading some of the hackathon projects
00:16:07
and things like that. And I think like something like 30 teams like sent somebody
00:16:12
and it seemed like every team had at least a couple of people just there. So nowadays, like there's probably like five or 10 teams
00:16:23
that are kind of like two to three people. And then maybe another like five to 10 that are in that,
00:16:29
like maybe five to 15 that are in that, like three to five. And then you have a couple like really like strong groups
00:16:36
that are in that like five to 10, maybe 10 plus in a couple of cases. Wow, wow, wow.
00:16:42
Okay, they're really pacing, they're lapping everybody else on that case.
00:16:48
Let me ask you a challenging question, Tyrell. Say we're walking, we're at some event
00:16:53
and I happen to be with a couple of NHL execs, senior guys. They run a club, they're really deep on hockey.
00:17:02
They've accomplished a lot on the ice and in the executive suite, but they're not analytics forward.
00:17:10
And we happen to bump into you and we share an elevator going up or down and you've got like a minute
00:17:16
and I introduced you. I said, Tyrell, why don't you pitch these guys not on teamwork per se, but on analytics?
00:17:22
Because they're curious, but they're skeptical. And so you got a minute to pitch these senior guys
00:17:30
on analytics for hockey. Yeah, I think there's kind of like two routes. And like, I think the first one is probably like
00:17:37
the safest is like player evaluation of like, hockey is a little bit like, let's say soccer,
00:17:43
where like there really is a lot of different leagues that are relevant to the NHL.
00:17:49
Like there's a lot of players, like there's what, 600 players in the league.
00:17:53
And then like of guys that are kind of marginal that are existing, you know, and they're in Swiss,
00:17:58
they're in the NCAA, they're like a top junior. Like it's really like a large pool of players
00:18:03
that are like possible value adds, like either today or tomorrow in a way that like,
00:18:08
I think sometimes we have a hard time with like, let's say some basketball teams are like,
00:18:12
yeah, maybe your models are good, but like, there's only 90 people that matter.
00:18:15
And like, I can watch their videotape. You know, I think like hockey is one of these sports
00:18:19
where like, there really is like a pool of a couple thousand and like, you probably can't watch all of them.
00:18:26
But what we can do is we can provide like a digital scouting report. And that's just going to be a first pass filter
00:18:32
to like, look like ultimately you guys are the hockey guys. There's some guys that our model is going to say,
00:18:37
like he's a hotshot in NCAA and maybe he sucks. And your scout, if he's good enough, maybe he can say that.
00:18:44
But we don't want your scout wasting time, like looking at the wrong set of people.
00:18:50
And then I think with like tracking data, like nowadays, like you can get metrics
00:18:54
that are a little bit more in the language of a hockey player. Like you can clue in on something,
00:18:59
like I can show a pass value. I can say like, I value this pass like this much. It's like this percentile.
00:19:05
And I'm taking in enough context that like it passes the eye test in a way that like,
00:19:09
I think lower level stuff, even though I think in the aggregate, working off lower level data is often really powerful.
00:19:17
But we've always had a hard time with like coaches or more traditional hockey people,
00:19:20
because they look at the video and their eyes are seeing information that your model isn't.
00:19:24
And they're like, well, this is wrong. And I think with like the kind of tracking stuff
00:19:28
we have now, there's more and more metrics that we can like point to and just say,
00:19:32
well, like, look at that. Like, does this list not look like the 10 best passers?
00:19:37
Like, do these passes not look like, you know, really good passes or really good carries?
00:19:41
Like the more we're able to zoom in on these kinds of micro things, like the closer it tends to get
00:19:48
to the hockey language, I think. Super interesting. It makes a lot of sense. The way you just talked about that
00:19:55
raises a new question in my mind, especially as a person, I've watched hockey for a long time,
00:20:01
but I've never played hockey. And I'm definitely handicapped in that way.
00:20:05
As I'm watching games, how should I think about where a player adds value? You just talked about through passing, for example.
00:20:13
And some of the passing, good passes are so fantastic in hockey, but let's just stay with like forwards.
00:20:21
So how much of the value a guy creates on the ice during a game, next year, whatever,
00:20:26
comes from shots, comes from carry-up versus off-puck activities? Off-puck or good passes,
00:20:34
receiving passes well, recovering loose pucks. Like where, if you just give me some activities
00:20:39
that maybe as an unsophisticated viewer, I miss that this guy's actually, he has more chances to pass well than he does to shoot.
00:20:45
So give me some sense of that. Yeah, I love that question. I think like, especially when we've looked at this
00:20:54
in kind of our broader models that are trying to encompass as many leaps as possible,
00:20:59
it often ends up getting like, the things that you do like with and near the puck
00:21:04
is probably about half, like as best as we can tell. And, you know, that number is probably a little bit off.
00:21:11
And like, as you sort of add, you know, as soon as we get like, let's say like stick location values
00:21:16
and like these kinds of things from like Hawkeye coming in, maybe those percentages will change.
00:21:21
I think like you're passing, shooting, loose puck recovering, carrying, shot blocking, takeaways.
00:21:28
Like it's probably close to half. It might be a little bit more for forwards than defenders.
00:21:36
I think defenders are a little bit tougher, but that's typically when we try to like blend
00:21:41
like a micro and a macro stat to get something overall, it's often closer to 50-50 than you probably think.
00:21:47
Okay, but that's like shockingly high percentage for off-puck behavior to the uninitiated anyway.
00:21:53
And so give me one off-puck behavior that I could try to monitor a little bit that's high value
00:22:00
and give me one on-puck behavior that's not obvious. Like among all those things you named,
00:22:05
what's something that's especially high value or maybe it's high frequency or whatever
00:22:10
that I might need to pay more attention to? Yeah, so one of the like, maybe an off-puck kind of thing,
00:22:20
like there's some defenders that are really good at just like reducing like pass threat
00:22:27
in whether or not they're like on the defender or not. There's like pretty good evidence of like a lot of those guys
00:22:33
that have that defensive reputation, like a Jacob Slavin for Carolina, somebody I'm being in New York,
00:22:40
I've watched like a decent number of Islanders games this year. And like Adam Pelek is one of these like defenders
00:22:46
that like no matter how I slice it on the defensive side of the puck, he's really incredible.
00:22:50
And like one of those things, like when I'm watching an Islanders game, I'm always like, what are you doing, Adam Pelek?
00:22:55
Like, what exactly are you doing? And it's hard to see. And like, I think I'm surprised
00:23:02
that like I'm watching his positioning and I'm like, whoa, like he feels out of position to me.
00:23:07
And then all of a sudden he's like there to stop a play. So I think that's one of the ones like
00:23:12
of an off-puck behavior. Maybe on-puck, like something that surprised me, it's maybe obvious in hindsight,
00:23:20
but you being the person that is like recovering loose pucks like is extremely predictive.
00:23:28
It's one of those things that like, one of the things that I always try to look at
00:23:31
when I'm thinking about how predictive is a statistic, I really care about how does this translate across teams
00:23:38
and how does this translate if I was to take it, those players that get like swapped
00:23:43
or they're in completely new context, like how predictive is it? Something like loose puck recovery,
00:23:48
like if you're on a good team, if you're on a bad team, if you're in NCAA, like whatever, you're in the NHL,
00:23:55
like the guys that get a lot of value out of that, like it's extremely predictive.
00:23:59
Somebody that's like credible, like Celebrini, I was watching him in the whole like Olympics
00:24:04
and like, I don't know how he does it. Like, you know, he's not that much bigger than I am.
00:24:08
And like, he's just able to grab pucks and they're just on his stick. And like, it's so valuable.
00:24:15
Yeah, I was about to ask, like what behaviors lead a guy to be able to get the puck like that?
00:24:21
I would think that some of that is effort, but you just made it sound it's also a stick.
00:24:25
There's like some really like impressive anticipation stuff. Like, yeah, if you watch like, you know,
00:24:30
I feel like I didn't play the game that that long ago, but like sometimes when I watch the NHL,
00:24:35
it's really in the future, especially with like things like shootouts and then like things with like takeaways,
00:24:39
there's these really subtle like lifts and like people are faking each other out,
00:24:44
like making it look like they're going to take the puck one way and they're grabbing it the other,
00:24:48
or they're out anticipating like really impressive stuff these days in terms of the skill level.
00:24:55
All right, well- Can I ask Tyrell just a quick question? I was thinking about this when Kate asked you about,
00:25:02
you know, what you would say to, you know, whether it's a CEO or someone that works for a hockey team,
00:25:07
would you hire more data engineers or data scientists? Ooh, that's a good question.
00:25:14
Because it sounds like there's extraordinarily rich data that's in even the Teamworks pipeline,
00:25:20
but without a set of data engineers, I'm not sure what we as data scientists
00:25:24
would do with that data. But on the other hand, they can do a lot of description,
00:25:28
but if you want to build predictive models or think about something you studied, causal inference,
00:25:33
then you might need a statistician. So I'm just interested how you think about that.
00:25:37
Yeah, like, I mean, definitely one of the theses of like how we set up our product is like,
00:25:42
we take care of as much of the data engineering as possible to like make that, like, make the data like accessible.
00:25:50
We do a lot of mapping, cleaning, re-representing, like if you want to work with it at a smaller level,
00:25:55
a higher level, join in features. We try to do a lot of that for teams because I think you're right.
00:26:01
Like, if you want to be at the bleeding edge of statistics at this point, like we're about to get Hawkeye.
00:26:07
Hawkeye data is, you know, it's something like 36, like join angles, you know, like a hundred times a second.
00:26:12
We got it in basketball about a year ago. The size of like just one game is massive.
00:26:18
And like, if you really want to be able to use that properly, you need biomechanists,
00:26:23
you need data engineers, but it's kind of true. Like once you have like some kind of a pipeline in place,
00:26:29
like there's just so many more analyses that are possible. But I think like probably the data engineering,
00:26:35
like I think a lot of us that are like, you know, decent data scientists in this space
00:26:39
have had to be kind of scrappy and had to learn. That's something that like Doug Farring likes to say a lot,
00:26:45
the former CEO of Zealous, now at Teamworks, is like, you know, like the best way
00:26:50
to become a better data scientist is become a better data engineer, especially in the sports world
00:26:54
where like you have to be pretty scrappy and you're going to have to do some stuff
00:26:58
like a little bit outside of your world. Has AI helped? Because it helps me be a better data engineer.
00:27:04
Yeah, I think that's true. Like I think there are ways in which, you know, this kind of technology,
00:27:11
if you're like strong technical in one field, that you can sometimes, it can help with that like domain transfer problem
00:27:17
a little bit. You still have to learn stuff, but I think there's definitely ways in which
00:27:24
that can be true. Tyrell, you mentioned that Hockey's about to get Hawkeye data
00:27:29
and that will open up all kinds of biomechanical as well as finer positioning. What's an example of a question you're curious to tackle
00:27:37
with these new data? Or what's something you think we're gonna be talking about
00:27:40
two years from now because of the data? Like, you know, like it's probably not as exciting
00:27:45
as you might want, but like, you know, we just don't know where the sticks are.
00:27:50
And that's really, it turns out, it's probably a pretty big part of hockey.
00:27:53
And like, we have these spatial locations and we can say two players are close, but we don't have,
00:27:58
one of the things that is kind of a bummer is like we only have one RFID chip in the NHL
00:28:04
on the right shoulder. NFL has two, so we don't have body orientation. I don't even know when a player is skating backwards.
00:28:11
So like, I really wanna know like, who's good at skating backwards. I just wanna know where the sticks are.
00:28:17
Like just starting at major events, like where are the sticks and like, just that is going to give you like much better,
00:28:22
like expected goals models. It's gonna give you a lot of stuff for free a little bit.
00:28:27
You could infer that maybe Tyrell, right? But I mean, in the sense of like, you know,
00:28:32
if you had a supervised machine learning model where you human coded people skating backwards,
00:28:37
I would imagine the movement of the RFID chip would have to be, even though it's only on the right shoulder,
00:28:42
would have to be different than forward movement. And so in theory, you could infer this, right?
00:28:46
Absolutely, you must have been talking to Doug cause we were very close to going down that pipeline
00:28:54
and like, kind of like realizing that Hawkeye was probably closer than we originally thought.
00:28:59
Like it was one of those things that made sense to defer. But we sort of thought about one of these
00:29:04
like hit and mark off like type models where it's like, kind of what you're saying
00:29:08
is like, there's probably moments in time where I like, I know for sure a player is going forwards or backwards,
00:29:12
plus some of these like movement stuff, I should be able to like chain it together some way
00:29:16
and get like a probability. And that would probably help quite a lot, to be honest in some of these models.
00:29:22
Tyrell, we're gonna have to let you go. Thanks for making time. I assume you're an Oilers fan, you grew up in Edmonton.
00:29:28
Does that make you an Oilers fan? Yeah, I grew up as an Oilers fan. What's the prospect for McDavid the next few years?
00:29:35
I got on the Oilers train the last few years, which is just a cheap, easy thing to do as a,
00:29:39
let's watch the most exciting team. What's the future for McDavid? Did they miss the playoffs?
00:29:46
No, no, they took a step back this year, I think it would be fair to say, but. And McDavid, like it seems like he,
00:29:53
had like some kind of injury during the first round, like with like the foot, which I think was fairly apparent
00:30:01
in some of the reports that I was like looking at. I was like, oh, these don't quite look
00:30:05
like the normal McDavid numbers. I was definitely thinking that was quite likely.
00:30:12
Yeah, I don't know. Like, I mean, I like the team, like I think they have, you know, a lot of, some of their depth players,
00:30:20
I think are a lot better than people give credit to. I like Evan Bouchard, I know that's like something,
00:30:26
but I think he's like a really excellent, he's like one of the best stretch passers in the league
00:30:31
and has been for probably three or four years. I think like, you know, they're a team that like,
00:30:37
if McDavid and Dreisaitl are playing at their absolute limits, like it's really hard to count them out.
00:30:42
And yeah, like, I think like, yeah, I'm interested to see where that team goes anyway.
00:30:49
Okay, okay, well, we should let you go. All right, Tyrell Stokes, senior manager at Teamworks Hockey Analytics.
00:30:56
I think you head the hockey operation at Teamworks. I'm on the technical side anyway, yeah.
00:31:01
Great, and first time guest, we're delighted to have spent some time with you.
00:31:04
Good luck with all you're doing, Tyrell. Thank you, Ariel. Thanks for having me.
00:31:09
That has been the first half of Wharton Moneyball. We still have a half to go. Come back and join us after the break.
00:31:17
Welcome back. Welcome back to Wharton Moneyball. Welcome back to a full hour of sports analytics
00:31:23
here on the Wharton Podcast Network. Cade Massey hosting with Eric Bradlow and Shane Jensen,
00:31:29
our fourth musketeer, Otty Weiner out and about doing Otty Weiner things. He'll be back.
00:31:34
Some combination of us are here almost every week of the year, maybe 50, possibly 50 of the 52 weeks of the year.
00:31:41
I have been for more than 12 years just off the line with Tyrell Stokes. First time guest, Tyrell Stokes.
00:31:47
Highly recommended, referred by experts, Tyrell Stokes, talking hockey analytics.
00:31:53
Great fun, great fun. Yet another great hire by Zealous. Continue to be impressed with those guys
00:31:58
even after being acquired by Teamworks. Shane, missed you last week. Didn't get to talk about hockey with you.
00:32:07
Anything? Yeah, I mean, the Flyers have risen and fallen in the intervening time.
00:32:12
It's very, I mean, it was probably- Inevitable. Kind of somewhat as inevitable as hockey gets,
00:32:19
I suppose, in the playoffs, which is not very inevitable, but yeah, no, the Hurricanes have been
00:32:24
an absolute wagon this postseason. I think I saw that it's the first time there's been two, like,
00:32:31
they've started out with two series sweeps since like the 80s, like since the Oilers in the 80s.
00:32:36
Jeez. So anytime you're invoking a Gretzky- Wait, you're saying no one has had
00:32:41
two series sweeps since the 80s? In a row, to start the playoffs. The first two.
00:32:45
Yeah. Yeah, yeah. Wow. Yeah, and I think actually- That's actually a remarkable statistic
00:32:53
because we have one in basketball this year, obviously, as well, the OKC Thunder.
00:33:00
And I'm sure, I mean, I think, I know for a fact it's happened 11 times in basketball,
00:33:05
I believe. No one's ever finished this. No one has ever gone, you know, that was the famous Moses Malone thing.
00:33:11
We're going to go fo, fo, fo, fo. They didn't. They went fo, fo, fo, three, one, or four, one.
00:33:17
But no one in basketball has ever done it. But I think that talks, I assume you would agree with the randomness in hockey.
00:33:22
Like, you know, you could be the 0.8 team and you know what? You're not winning every one.
00:33:28
That's right. No, and I think, you know, it's carried an interesting dynamic in this playoff too,
00:33:33
because, you know, Caroline has done what they're saying. I think I counted, I think, you know,
00:33:38
if the Buffalo-Montreal series goes long, and it easily could, those teams are really going at it with each other.
00:33:44
I think they could have like a 10 game layoff or something like that. Yeah. Which, you know, I mean, you could argue, of course,
00:33:51
you could argue whether you would want such a long layoff or not, either way. Come on, you got it.
00:33:56
But it's just, I, it's- Isn't that the perfect mid-playoff break? Are you kidding me?
00:34:01
Two series in, two series to go. Let's take 10 days off, get rested. Because it's such a grind, these four playoff series.
00:34:06
It's got to be, maybe it costs them the first game with the Canadians. Maybe it costs them the first game.
00:34:11
I think that's got to be worth it to take this much time off in the middle. It's not necessarily going to be the Canadians, my friend.
00:34:16
Your Sabres are still in it. They've not- I mean, at least until they meet up with Carolina.
00:34:21
Yeah, exactly. Exactly. No, it is, it's an interesting playoffs because I feel like we've had
00:34:26
some incredibly exciting series, but there are more than I think most years, like these two elephants hanging out
00:34:34
in either side of the room, basically, you know, kind of waiting there. It's hard not to script yourself
00:34:40
into an Aves Hurricanes final, even though we've got some amazing hockey along the way.
00:34:44
I forget who does the, this is the Micah McCurdy, right? The database. McCurdy, yeah, yeah.
00:34:49
So I'm looking at what you posted, or I assume it was you that posted here. And just for our listeners,
00:34:55
like, I don't know how to describe it, but to say that most of the probability mass,
00:35:01
I don't want to say 90 plus percent, but at least is on Colorado and Carolina.
00:35:07
Yeah, I think when you say 82 percent- It's hard to squint. Like I have to squint to see the other lines at the end.
00:35:14
And, you know, my middle son's girlfriend is a huge Ducks fan. And I'm like, are they still in it?
00:35:20
Like, I'm trying to, I know they are, but I'm like, I'm trying to find them
00:35:23
at the end of the rainbow here. And like- No, and I mean, he produces these visualizations
00:35:29
every year, they're really cool. And I do feel like this one is kind of, again,
00:35:32
it speaks to the unusual year, I think we have, where, you know, two teams really do kind of soak up
00:35:38
so many, so much of the percentage. And they're both on either side of the bracket.
00:35:43
So it's not like they're kind of soaking it up from each other. Because I think, you know, somehow if you were to kind of,
00:35:50
like if you somehow were to substitute maybe Carolina and Minnesota, like move the Wild over to the East
00:35:55
and get Carolina out of the East. I'm not sure the Wild aren't like, you know,
00:36:00
suddenly one of the top two teams to make it. I think it is, you know, I think it's kind of like,
00:36:06
it's more kind of, yeah, consolidated in a couple of teams than it usually is.
00:36:11
Whether that'll actually realize or not. I mean, you know, again, another test of how random
00:36:17
it can get in hockey playoffs. It'd be cool to, it'd be cool if another team,
00:36:20
though I would be fine, Avalanche Hurricanes would be such an amazing final. I would, I'd sign up for that.
00:36:28
Okay, well, trying to enjoy the process as well along the way, except for the Sabres getting blown out.
00:36:32
Other than that, it's been a delight. Why don't we jump into a, what caught your eye?
00:36:37
Let's do a couple of rounds. Let's see if we can do a couple of rounds of what caught your eye in the world of sports.
00:36:43
Eric, probably the right place to start us off. Well, I'll go back to a slightly different sport,
00:36:49
which is tennis. Obviously, Alcaraz is still injured. Sinner hasn't lost in forever.
00:36:57
And I don't know when the next time he's going to lose. I mean, he's going through top 20 players,
00:37:01
six loves, six one, six one, six two. I'm not talking about beating top 20 players.
00:37:06
I'm talking about beating top 20 players. His odds right now for the French,
00:37:12
which is coming up, not now, he's still playing, they're still playing the Italian Open.
00:37:15
The French is starting after that. He's minus 270, you know, which probably puts him at about 70%.
00:37:21
And I'm wondering if that's too low. Like, I mean, I checked, the highest Nadal ever was,
00:37:28
with just, it's good to benchmark, was minus 400. Now at that time, I don't know if he was
00:37:32
the five-time champion, eight-time champion at that point. But like, let's put that as an upper bound.
00:37:37
Like, it's hard to imagine anyone being better than Nadal in his prime at the French.
00:37:43
Like, there's no legitimate probability model that would put somebody better, except, except,
00:37:49
this gets to Shane's point with hockey. There ain't no Federer, and there ain't no Djokovic sitting there
00:37:56
who could possibly beat him. So it's not unreasonable to say, Sinner could be minus 500.
00:38:02
Why can't he be? Nadal had different competition. I'm not saying he's better.
00:38:07
I'm saying he doesn't have the same level of competition. So that's what caught my eye.
00:38:12
I think minus 270 is too low. I think, unless he gets injured, I think he's got a better chance.
00:38:17
I was about to ask, what's the base rate of injury? Because that's going to put a ceiling on the odds.
00:38:23
I mean, getting injured within a tournament is not as actually as high as you'd think.
00:38:29
Like, players typically, once they start tournaments, it's rare that they lose or pull out due to injury.
00:38:36
So that's not as common. Now, the fact is, Alcaraz is not playing. It's not even clear Djokovic is listed as he's playing,
00:38:42
but he was just beaten in the first round by like number 300 in the world. So he may be playing, but he's not really playing.
00:38:49
I don't see how he's not more, anyway, that caught my eye. I was like, can minus 270 actually be too low?
00:38:56
I got kind of almost a tennis design question for you, Eric. If I wanted it to be less dominant,
00:39:03
like if I wanted to kind of change a rule or two in tennis so that maybe the best player
00:39:10
didn't like, wasn't so kind of high, to basically reduce the probability of the best player,
00:39:15
or maybe even to kind of like, you know, add randomness essentially, so that it's not so, you know,
00:39:20
the best player winning all the time. And I'm not saying that's necessarily a goal you would want,
00:39:24
but if you wanted that as a goal, what would you do? What do you think kind of- I mean, you could do stupid things,
00:39:30
like say every game is one point. You know, I don't mean that. Let's forget stupid things like that.
00:39:35
No, of course. Something that has been talked about, make every round up until the finals, best of three.
00:39:42
Like you're not, I hate to say it, but with the condition he's in, like it's one of those things that I always remember
00:39:48
when, you know, when my middle son was playing squash. You hit the ball about 10,000 times in a squash match.
00:39:55
The other guy's better than you. They're better than you. And you know, you can hit a bunch of winners,
00:40:00
but not after 10,000 shots, you're not. Sinner right now hits the ball by far
00:40:06
more cleanly than anybody else. You make it end big enough, he's going to beat you.
00:40:12
So unless you shorten the match by making it best of three possibly, or change, for example, the rules within a game,
00:40:21
which is unlikely, tennis has been played this way for 70, 80, 100 years, whatever.
00:40:26
But like get rid of tiebreakers, for example, maybe. Well, they've done that.
00:40:29
They've done that now in the fifth set. Like if you happen to get to the fifth set,
00:40:33
it's now the first to 10. It's not like, you remember when Isner, like the 48, they're not doing that anymore.
00:40:39
Yeah, they're not doing that anymore. But no, right now at best of five, I just don't see it.
00:40:46
I don't see how someone is going to beat Sinner in a five game match anymore.
00:40:51
And because not only is he better than you, but he's in better shape than you.
00:40:55
And he hits the ball cleaner than you. And so, and his movement's better than you.
00:41:01
Do we expect Dakaraz back for Wimbledon? It's unclear. I mean, he's got, yes, I would think so,
00:41:07
but he's got an injured wrist. And you know, he says he's not coming back
00:41:11
until he's fully 100%. I just saw a video of him yesterday. He is still wearing a brace on his wrist.
00:41:16
So I don't know how close he is to being ready. You know, and it's always one of those things.
00:41:22
The good news about, well, it's not a good news. Let's say, let's say, I don't say worst case.
00:41:25
Let's suppose Alkaraz is out for the rest of the season. I don't know that he is,
00:41:29
but let's pretend he is for the moment. Or he's not Alkaraz for the rest of the season.
00:41:33
And Sinner wins the next three Grand Slams. So now he catches Alkaraz. So now they're both at seven.
00:41:39
Is there an asterisk next to his career? No. No, no, I think all these careers ought to be normed for their competition.
00:41:46
We've talked about that before. But no, you're not- I mean, maybe true aficionados like you
00:41:52
look at like, you know, Nadal, Federer's kind of list of Grand Slams and puts like, oh, well, these were the ones where,
00:42:00
these were the years where Nadal was injured. And these were the years, like, I mean-
00:42:03
It is what I do, yeah. Yeah. And I mean, I'm not saying it's wrong to do that.
00:42:10
But certainly I think for, you know, kind of their overall like view, maybe outside of like somebody as deep in it as you,
00:42:19
call me the casual fan. I make no decision. It'll just be the number, the totals in the end.
00:42:25
Shane would be interested in a strength rating of a career set of Grand Slams. Oh yeah, yeah.
00:42:31
You should show me the ELO rating, kind of all that type of stuff. That would be great.
00:42:34
And by the way, except for situations like, let's take the rare cases like in baseball,
00:42:38
like, you know, Ted Williams went to war twice. You know, part of being great and having longevity
00:42:44
is being able to be on the field. You can't contribute if you're not on the court.
00:42:47
So if Alcaraz, look, people say this about Nadal. Nadal played way fewer Grand Slams
00:42:55
than Federer or Djokovic. If he hadn't been injured. Okay, well, yeah. But if he didn't play the style he was,
00:43:01
maybe he wouldn't have won 22 of them either. Let's go. And we're talking about Mike Trout
00:43:05
and your favorite sport as well. Okay, Shane, let's go. Okay, that is, I feel like that is a good segue.
00:43:10
Cause I feel like what caught my eye, obviously, you know, Kyle Schwarber hitting all these home runs
00:43:16
is what actually caught my eye. But, you know, it got me thinking about kind of home run totals.
00:43:20
And, you know, just kind of, there's a lot of, there's a lot of players actually,
00:43:24
kind of this season specifically, that are on track to hit, go over 400 home runs.
00:43:30
I just saw this. This is a great, this is great. And it really kind of, I mean, you know,
00:43:34
I wouldn't necessarily use it as an absolute proxy for Hall of Fame worthy, but like, you know, it's certainly,
00:43:41
you get over 400 runs and you're starting to get in the mix of like, where it's about 50-50
00:43:45
that you'd be in the Hall of Fame. How many guys have more than 400 home runs in MLB history?
00:43:51
I'm gonna guess 60, but I'll look it up. Okay. So anyway, the players right now,
00:43:56
so Eric Judge is only 16 home runs away. And that was from a couple of days ago.
00:44:00
So he's probably only 10 home runs away now. Oh, I mean, I was so far off. It's 59.
00:44:04
Ha ha. That's a good guess, Eric. Nicely done. But yeah, but it really is kind of a list
00:44:10
of sort of people, at least on the fringe of Hall of Fame or definitely in the Hall of Fame.
00:44:15
So Eric Judge is only 16 home runs away from 400. Manny Machado is 25 home runs away.
00:44:20
Bryce Harper's 27. Freddie Freeman's 29. So all of these guys, you know,
00:44:26
we could have four or five players hitting 400, you know, breaking that 400 mark all in the same season.
00:44:33
There's a couple other players like Paul Goldschmidt and Nolan Arenado that are also kind of in that range,
00:44:37
but they're more, you know, the four players I mentioned are still enough in their prime
00:44:42
where I feel like they've got a 23rd home run season. Here you go, Shane. This is a fun list.
00:44:46
Here are all the players in MLB history with 400 home runs not in the Hall of Fame.
00:44:53
Yeah. Okay. It's about 12 names. Here we go. Barry Bonds, we know why. Yeah, yeah, yeah, yeah.
00:44:59
Alex Rodriguez, we know why. Pujols, not eligible yet. Sammy Sosa, we know why. McGuire, we know why.
00:45:08
Palmera, we know why. Manny, we know why. Gary Sheffield's an interesting case
00:45:13
because he has 500 home runs, but he's off the ballot now. He's had his 10 years.
00:45:19
Carlos Beltran, he's coming in. So he's coming into the Hall of Fame. Andrew Jones coming into the Hall of Fame.
00:45:24
Jason Giambi. Paul Canerco. Yeah. Juan Gonzalez. The name I remember from my youth, Dave Kingman.
00:45:33
Darrell Evans, Alfonso Soriano, and Mark Teixeira. That's the list. That is a short list.
00:45:41
That is interesting. Yeah, no, and I mean, again, it is a lot of obviously kind of suspected PED guys.
00:45:48
I guess maybe with Soriano, Teixeira, it's also if you hit a bunch of home runs on the Yankees.
00:45:52
I mean, like, you know, whatever. That's not that notable. Those guys are supposed to do that.
00:45:57
So guys, one thing that stands out to me is we're talking about, you know, in the next year or so,
00:46:02
we're gonna have these five active players. Judge Machado, Goldschmidt, Harper, and Freeman.
00:46:08
Over 400. How often have there been, and there may be others who are still playing.
00:46:12
Yeah, I mean, Statton's at over 500. So we've got active players that are already,
00:46:16
have already, are already over. Are we at a relative high point for a number of active players with 400 or more home runs?
00:46:23
Not relative to the early 2000s. Well, forget those guys. Forget those guys. Take that out.
00:46:30
Yeah, I think we are. I mean, I think, you know. We're about to be. Residualizing LPEDs somehow,
00:46:36
I think we're at kind of one of the highest sort of home run times. Well, I guess I would follow, given.
00:46:42
Yeah, it's been a high home run. Although, I will say, there was an era where,
00:46:47
I'm just gonna do this from memory, guys. Willie Mays, Hank Aaron, Stan Musial, Ted Williams,
00:46:55
Willie McCovey, Ernie Banks. So I could keep going. Those aren't all the same era, though.
00:47:03
No, no, no, no, no. They were. A lot of them played in the 50s. I mean, Ted Williams and Hank Aaron,
00:47:08
did those guys really overrun? They did. Ted Williams, of course they did. No, no, no.
00:47:12
Five seasons. Ted Williams' last season was 1960. Willie Mays' first season was 1951.
00:47:21
Hank Aaron's, I think, was 55. Stan Musial played forever. He played in the 40s and 50s.
00:47:27
Ernie Banks started in the 50s. Willie McCovey. So no, these guys all played in the late 50s.
00:47:32
Mickey Mantle, sorry, I should have named the Mick. Duke Snyder. So I could keep going if you want,
00:47:38
but I'm just saying, there were a lot of guys in the late 50s that had 500, 400 plus home runs.
00:47:45
Interesting. Okay, very good. And they never were in the Hall of Fame at the time,
00:47:49
because it just started to start to run. I'm gonna jump to another, I'm gonna jump to the female version of this sport,
00:47:55
the Women's College World Series, has just been seeded on Sunday. They seeded this thing.
00:48:02
The regionals are this weekend. It's a 64-team tournament. There are 16 regionals, each with four teams.
00:48:09
The winners of those will roll up to eight super regionals, and then the winners of the supers
00:48:14
will go to the College World Series. So that's the basic structure. So there are 16 seeded teams.
00:48:21
Those are the teams that host the regionals. They're ordered, and so you know which eight
00:48:26
will host the supers if they win their regional. And what is it, it's not single elimination.
00:48:31
Is it best of three? No, no, this is this wonderful mixed format where they play round robin,
00:48:37
a double elimination for the regional, and then at supers, they play two out of three.
00:48:43
And then for the first round in the College World Series, they play double elimination brackets,
00:48:48
and then the champions play two out of three. And so it's this- Oh, yeah, yeah, they rotate.
00:48:51
Yeah, they rotate back and forth. It's an efficiency thing, I think. But anyway, it's just starting.
00:49:00
It's a fun sport. It's a fast-paced sport. It's probably my favorite women's sport to watch,
00:49:07
and it helps that the University of Texas is good. But notably, guys, six of the top seven teams
00:49:15
are from the SEC, and then three more from the second eight are from the SEC. So nine out of, I mean, I think it's only 16 teams
00:49:25
in the SEC. Nine out of 16 teams are hosting regionals. They're one of the top 16 teams,
00:49:30
but six of the top seven. Nebraska is number four. Nebraska is the only one of the top seven
00:49:35
who's not an SEC team. So good fun. UCLA is also- Well, probably a different SEC representation
00:49:42
in the college hockey version of this tournament. Yeah, that's right. It's, we get, some of these sports are pretty regionalized,
00:49:50
college hockey being one of them. Baseball actually always surprised, or softball, baseball.
00:49:54
That one always surprises me how regional, because the history of it, so much of the history of it is Northeast,
00:50:02
but- But it's weather. The powerhouse is obviously Texas, the southern- It's weather.
00:50:07
It's a weather thing. The guys can't play in the early spring in the Midwest,
00:50:11
in the Northeast. It's just a, it's totally a weather thing. All right, let's do another quick round.
00:50:15
I've got, I'm gonna end with a question, a little application of stats for you guys,
00:50:20
but let's go one more round of what's caught your eye. All right, so I'll just go quickly.
00:50:25
My son, Zach, told me a stat yesterday. Relates to what Cade was talking about, Shane was talking about with the hurricanes being 8-0.
00:50:32
So I mentioned OKC is 8-0. So on Kalshi, which we all know is a betting site, here are the odds.
00:50:42
It's not a betting site, Eric. It's a prediction market. Sorry, it's a-
00:50:46
Economic rationale for needing to place these investment opportunities, predictions.
00:50:53
All right, there's an investment opportunity on Kalshi with the OKC going undefeated.
00:51:00
And anyone wanna guess what the odds are on Kalshi? I put it in the sheets. You probably saw it.
00:51:06
So we're halfway through. We're halfway through, right? We're halfway through.
00:51:08
So they'd have to win eight straight games against whoever the finalist is in the Western Conference
00:51:15
and whoever, of course, goes into the NBA Finals. So eight straight games. So what do you think the odds are according to,
00:51:23
if you didn't look, according to Kalshi, that I can make a prediction-style bet
00:51:28
that they will go undefeated? And I'll give you a benchmark. 0.8, let's pretend they were 0.8 in every game.
00:51:38
0.8 to the eighth is 16%. And remember, they're gonna play half these games on the road.
00:51:46
So it should be less than that because we don't think they're 0.8 to win a given game.
00:51:50
All right, well, let me also tell you, hold on a second here. 0.6 raised to the eighth is 1.6%.
00:52:01
Okay. So somewhere between 16 and 1.6. 10 seems something reasonable. Okay, so you're not far off.
00:52:13
I would have guessed it's totally unreasonable. The number's 7%. I thought that was just way too high.
00:52:20
Well, back out the implied game-level probability for that and ask yourself, you know,
00:52:28
and then get some spreads of games. And what have been the spreads? Of course, they're gonna be tighter in the next two rounds
00:52:34
than they have been the last two. Are they drumming? 72%. 72% game-level probability would justify a 7%,
00:52:42
8-0 run. Including, by the way, there are road games. Again, half those games are roads.
00:52:48
You have a mixture distribution between half home, half away. And the problem is, I'm not saying they would be,
00:52:55
but the minute you go to 50-50, let's say, for any of the road games, you slice that probability way, way down.
00:53:03
People have bad intuition, as you know. Like you say, what's the difference?
00:53:08
70%, 50%? Oh, wow, that's 30% difference in each of the games. That's a ratio scale.
00:53:15
That's absolutely crushes it. Well, does it say something about the expected strength
00:53:20
of the Western Conference? Is that, are they saying the second and third best teams out there just aren't that,
00:53:26
second, third, fourth best streams? It must be saying something about that. Well, San Antonio had a,
00:53:31
I think it's four and one or five and one record against OKC during the regular season,
00:53:34
for whatever that's worth. Now, OKC may not get out of its series, but it's 2-2 with Minnesota.
00:53:38
You have to believe that whoever comes out of the West is a dominant, dominant, against whoever comes out of the East.
00:53:46
And when I say dominant, at least 80%. Like, I don't see how you can get, well, I don't see how an East game,
00:53:53
and that includes the road ones, because I don't see how you can put OKC more than 60
00:53:59
or even two-thirds against any road game, against whether it's San, let's say it's San Antonio.
00:54:06
You can't put them at more than a two-thirds chance to beat San Antonio on the road,
00:54:11
which then I have to take all that other probability is only over six games, which means I'm gonna have to tremendously
00:54:17
cut down that probability. I just don't see it. Either way, it's that, you asked me what caught my eye.
00:54:23
And remember, I don't know that this is worth anything. Let's just remember, this has never happened
00:54:30
in the history of basketball. Never, despite whatever gap between one and two you ever want to say was the maximum.
00:54:40
The Golden State Warriors at its peak, the Lakers in 71 at its peak, the Boston Celtics at its peak,
00:54:46
the Chicago Bulls and Michael Jordan at its peak. No one has ever done this. Yeah, and there really haven't been many that are close,
00:54:54
right? I mean, it's not like there've been many 15, I mean, 16 and ones either.
00:54:59
Correct, I think the Sixers are the only team to have even gone 16 and one. Maybe Golden State did it one year,
00:55:06
but you're right, there's not even been a lot of, that's a great point.
00:55:08
There's not even been a lot of 16 and ones. Right, that gives you some sense of how unlikely it is
00:55:13
that there aren't even those that are not. And let's even remember, to make it even worse,
00:55:17
I think people, when some would say nobody went on 15, remember Kate, the early rounds used to be best of five.
00:55:21
She didn't even have to win 16. It was less and you still didn't do it. Right, okay.
00:55:27
Well, that adds some spice to the playoffs and it would be a heck of a way for the Thunder to repeat.
00:55:33
I mean, to do it that way. I think, and I think you would agree with this, if they happen to do it, and then I'd love to hear it.
00:55:38
I guess you said you had a question at the end. If they do this, that does put them into a special category.
00:55:46
In addition to the fact that they'll have won back-to-back, in addition, we've talked about this,
00:55:51
they've had the greatest point differential per game the last two seasons. From that simple metric, Dave,
00:55:57
this is the top two seasons in the history of the NBA. If they go ahead and also go undefeated in the playoffs,
00:56:06
I'm not trying to present them as the 90 Bulls or the Celtics or Showtime, but this starts to put them in a special category.
00:56:14
Yeah, Eric, if we're gonna start throwing around that margin given how much tanking
00:56:19
and how weak the bottom of the league has been, we probably need to norm that margin for era, right?
00:56:26
In some sense. That's fair. Okay, my last, what caught my eye, comes from a listener, listener question
00:56:32
down here in Central Texas about, you'll like this, Eric, this is completely up your alley.
00:56:38
You're the perfect person on this podcast to ask. You run a golf tournament,
00:56:45
and you don't trust the handicaps that the golfers put in. It's not a big, serious tournament,
00:56:51
so you have discretion to assess handicaps in any way you want. And you believe that it's important to observe
00:56:58
the players playing the course that you are hosting the tournament on. And so you're gonna use data from the tournament,
00:57:06
within the tournament, the tournament they're actually playing to win, you're gonna use some subset of that data
00:57:12
to establish the handicap. So I don't get to use other tournaments, just this, what happened here?
00:57:18
You don't believe any cards that they reported from other tournaments. You're just skeptical.
00:57:23
You're golfer skeptical about handicaps. And so, plus maybe you have a unique course,
00:57:28
so you're just gonna use something about there, you're gonna play one round, 18 holes.
00:57:34
So basically it's a simple, it comes down to a simple question. What subset, how many holes do you wanna use
00:57:41
to establish the handicap, and then apply it to the remaining holes? Yeah, so the answer is, I'm gonna use,
00:57:47
as a matter of fact, this is, I'll give an answer, and maybe this is not what you were thinking
00:57:50
I was gonna say. I'm gonna relate it to your comment on tanking. I'm gonna look at every hole
00:57:56
that they had every incentive to play at their best ability, which means there was a legitimate probability
00:58:03
for them to win the tournament, or they had to play best, but let's say it's match play,
00:58:09
whatever it happens to be, I'm gonna use all of those holes for which they should try to play their peak performance.
00:58:20
That's what I'm going to use. Again, assuming, wait, wait, I'm assuming, but maybe I've read your question wrong,
00:58:26
they're trying to hide their true strength on the other holes. Well, they're trying to win the tournament,
00:58:33
and so you have the challenge of, you're gonna conduct the tournament, you're gonna award a winner,
00:58:38
but you wanna adjust for their skills. You wanna try to put everyone in the same playing field.
00:58:44
So you have to declare a winner at the end of the day based on the 18 holes they played,
00:58:49
but you're gonna use some subset of them. But here's the thing, if someone knew this,
00:58:53
so let's imagine I said 18 holes, let's say it's a 72 hole tournament for a second.
00:58:58
Let's suppose I said 18 holes, and I'm Scottie Scheffler. I might have some incentive
00:59:04
to dump in those first 18 holes, depends the percentage hand you have. Eric, we're only gonna play 18 holes.
00:59:11
So you're conducting a one-day tournament. Oh, all right, you didn't say that.
00:59:15
Well, that's what you got. And also, we're not gonna announce ahead of time which holes,
00:59:20
because of course people would game it, okay? So we get to, after the fact, by some method, choose unannounced holes.
00:59:28
All right, so I'll just, here's what I'm gonna do. I'm gonna give you an answer, I don't, all right.
00:59:33
This is what I'm gonna say, and I'll tell you the principle by which I'm saying it.
00:59:36
50-50, and let me say why. I'm gonna maximize power on both split samples. I'm gonna get the best estimate possible of your handicap,
00:59:45
and I'm gonna get the best estimate possible of your performance, and so I'm gonna do nine and nine.
00:59:51
That's what my gut tells me to do if I only have 18. Well, so that was my instinct as well,
00:59:55
and so I'm glad to hear you say that. I wonder how the golfers would feel about having the tournament
01:00:03
essentially reduced to a nine-hole. You're basically turning the tournament down to a nine-hole tournament.
01:00:09
I know, but I also wanna get fair handicap estimates too. Exactly, exactly. So this person, this comes from Susan Skaggs,
01:00:17
a listener here in Austin, and she runs tournaments, and she does a six-hole handicap, randomly selected.
01:00:23
That would have been my other number. It would have been six, because I want more for the performance.
01:00:28
I'm willing to take a little more error, and by the way, you know what we're gonna do?
01:00:32
I'm gonna tweet at him, this is a Mark Brody question, if there ever was a Mark Brody question.
01:00:37
I'm not saying he has an answer to it, but he seems really smart to me, and he was just on last week.
01:00:42
If we tweet at him, I think he's gonna give us some sort of answer. I'd love to hear his answer.
01:00:47
How about that? But nine and nine, and six and 12 were the two numbers that I would have said.
01:00:51
Well, those also are the most mathematically simple to do as well, which is not nothing.
01:00:58
Well, it's not nothing. We want practical and implementable. All right, why don't we wrap it there then?
01:01:05
On behalf of the whole team, Audie, Weiner, and Absentia, Shane Jensen, who slipped away here at the very end,
01:01:11
Eric Bradlow, who's been here for the whole show, this has been Kate Massey.
01:01:14
Big thanks to Dion Simpkins making this thing happen. This might possibly be Dion's last live show
01:01:21
after God knows 600, 500 plus episodes, which I'm gonna jump out of a window
01:01:27
after this show is over and protest, but Dion Simpkins can't say enough for the man
01:01:32
and what he's done for the show over the years. For Marissa Reyna, our producer,
01:01:36
and Deep Patel, the boss lady, thank you guys. And for you guys, the listeners, thanks for being with us.
01:01:40
Come back and join us next time. Between now and then, enjoy your sports.