00:00:00
Welcome, welcome to Wharton Moneyball. Welcome to another full hour of sports analytics here on the Wharton podcast
00:00:08
network. This is Kade Massie hosting today with my longtime friend, colleague, collaborator, co-host Eric
00:00:16
Bradlo, Audi Winer, we're expecting at some point in the next hour. Shane Jensen, we are not expecting this week.
00:00:22
Our fourth colleague is out and about doing Shane Jensen things as he has want to do this time of year, but he will be
00:00:27
back some combination of us are here almost every week of the year. And by that we mean I don't know 48 49 weeks of
00:00:36
the year and for pushing 11 and a half years now. Delighted to be here on this Tuesday afternoon. Late summer Tuesday
00:00:42
afternoon. In fact, this is the last Tuesday of the summer months. If you wrap it up in August, it is very much
00:00:49
late summer. That means a few things, guys. That means that football's about
00:00:54
to start. We're going to spend some time on that. It means baseball is getting
00:00:58
more interesting. But it also means that some significant tennis is being played
00:01:02
in the New York City area. And for that topic, we wanted to bring in Jeff Sackman. Some of y'all who've been
00:01:10
listening know that we have Jeff on here a time or two a year and have for a long
00:01:13
time. Sackman is as good as it gets in the world of tennis analytics. jumped off the page at us early in our time. I
00:01:22
think we're reading like some of these economist articles without byel lines
00:01:26
and like we're thinking who the heck is this writing all this great stuff on
00:01:29
tennis analytics and we find out that it's Jeff Sackman. Jeff does a lot of stuff. He's the founder of tennis
00:01:34
abstract. It's an online encyclopedia of sorts covering much of the history of
00:01:38
tennis. But within that site he also writes a blog heavy top spin. Heavy top spin looking at players and trends. and
00:01:46
that's where you see some of the stuff that catches our eye. We're always
00:01:49
delighted for chance to talk with you, Jeff. Thanks for making time for us. >> Absolutely. Glad to be here, guys.
00:01:54
>> Glad to have you. Um, in the meantime, while I rambled around there in the
00:01:58
beginning, Audi Winer showed up. Audi is back. Audi is no longer on sabbatical. Audi's kind of supposed to be here every
00:02:05
week now. It's gonna be it's gonna be fun to have him back around. He is the
00:02:08
best traveler, I think, of the four of us. Audi, great to see you. >> Be here,
00:02:13
>> Jeff. In a minute. uh you and Eric can dive into all things US Open, but maybe
00:02:18
just as a starting place, you can tell us as the tournament starts, as you pull your head out of whatever data you've
00:02:25
been looking at lately, what is on your mind around tennis right now? How are you thinking about this tournament or
00:02:30
what player are you especially interested? Or is there any analytics project that's especially had you
00:02:35
consumed recently? Well, one thing that I was looking at this week is the big story in men's
00:02:42
tennis for I mean more than a year now is that we we can pretty much forget about the big four. We can pretty much
00:02:48
like let NovakJokovic ride into the sunset. It's the Yanuk Center and Carlos
00:02:52
Ocar show and the numbers are starting to put up are starting to be a bit mindboggling even though they are so
00:02:58
young. So, one thing that I do with every Grand Slam is I I I throw the the draw and my ELO ratings into a into a
00:03:05
formula and generate a pre-ournament forecast. And typically over the years, the number one seed, the top player in
00:03:12
the draw, not always the same player, whoever the my favorite is in the in the draw will end up with around a 35% 40%
00:03:19
chance of winning the tournament. And then there'll be maybe a half dozen guys
00:03:23
who are in the 5 to 10% range. Now, coming into this tournament, Yannik Center was at 45%. Not the highest he's
00:03:30
ever been, but that's that's very high for someone to be at 45%. But the
00:03:34
shocking thing is not that. The shocking thing is that Carlos Alcarez is at 35%.
00:03:38
>> So that's that's an 80% chance of these top two guys winning. Just I mean that
00:03:43
that's before a single ball is struck. So your number three normally I've got a
00:03:48
number three at least in double digits. Alexander Zerv is pulling up the number three spot at six whole percent which
00:03:54
you know if you follow Alexander Zerv that might even sound generous but that that's the state of men's tennis right
00:04:00
now. I mean even I don't think Federer and Nadal ever managed to pose numbers
00:04:04
like that partly because like they were half of the big four and the rest of the
00:04:08
big four was pretty good too. But I mean that's how dominant these two guys are
00:04:12
right now. That's astounding and it's a great top line and it's it's exactly the
00:04:16
kind of thing that we often ask about before a big competition. We'll ask it
00:04:20
about a golf tournament. We'll ask it about a tennis tournament and I don't
00:04:23
think we've ever heard anything quite that stacked up. Is >> just ask a followup if I could just ask
00:04:28
you quickly a follow-up to that. We're one point away from sinner having the
00:04:32
sinner slam at the French. So, um, how like is it even really like I'm an Alcarass fan just to get to make sure
00:04:42
everybody knows everybody knows this. Is it really even the big two? I mean, Sinner is one point away from holding
00:04:50
all four majors. >> Yeah, I mean, he's pretty dominant. On the other hand, my overall ELO ratings,
00:04:57
which are not not the surface specific ones, they actually give Alcarez a slight edge right now. They they say
00:05:02
Alcaras is the number one player in the world. not on hard courts. Sinner has the edge on hard courts. But what I
00:05:08
always think about is there there's two ways to be to win a slam. One way is to take down the top
00:05:15
dog, which is what it seems like someone's going to have to do. The other way to win a slam is let somebody else
00:05:20
take down the top dog. And if uh Grigor Deitrov hadn't gotten injured two sets
00:05:25
in his match against Wimbledon, then it would have been an absolute breeze for Alcarez to coast to that title. So
00:05:32
whether it's Alexander Bublick having another big day like he had on grass uh
00:05:36
running up to Wimbledon or whether it's somebody like Dimitro having a career
00:05:40
best day or who knows what could happen. Grand slams mean you got to win seven matches. Sometimes in the late rounds
00:05:46
it's going to be somebody like maybe it's Thafo in New York who's going to
00:05:49
play the best tennis he's ever played. Sinners got to run through those hurdles. So if we do get to the point
00:05:55
where it's a mono ono final in two weeks then yeah the edge goes to center. But
00:06:01
there's there's just too many hurdles in just in the structure of tennis. There's
00:06:05
too many hurdles to to annoy somebody the the big one unless he's unless he's
00:06:09
really ahead of the pack. Uh before his uh his suspension earlier this year, I think he was getting there. Uh he posted
00:06:17
a careerhigh elo rating in the 2300s, which is like top 20 of all time. Um he hasn't gotten back to that level since,
00:06:24
but maybe maybe I I can give you an update after this tournament. We'll find
00:06:28
that he is back there. I mean, he he could he could accomplish that, but just not quite yet.
00:06:33
>> Jeeoff, that was getting close to one of the questions I was going to ask about
00:06:35
the two dominating the probability of winning. How can you decompose that at least at least casually between there
00:06:43
being unusually good versus the rest of the field being a little bit weaker? Can
00:06:48
you give us some way of comparing those two things? >> That's a really, really hard question to
00:06:53
answer and I've never come up with a satisfactory way of answering it. Partly
00:06:57
because my intuition is that you basically never have a weak field in the tennis world because the tennis world is
00:07:04
always getting better. I mean, you have the same coaches working for generations. You have the same
00:07:08
nutritional insights, training insights, all that stuff is is gradually getting better over the
00:07:16
years. uh the the population of players that the tennis world is drawing on, the
00:07:20
number of countries with developmental programs, like every force there is is should be making the whether it's the
00:07:28
top 100, the top 50, or even the top 1,000 should be making it gradually a little bit better. So if if you agree
00:07:36
with me up to that point, I mean I I agree you can argue some of those points, but if you agree with me up to
00:07:40
that point, then if somebody is head and shoulders ahead of the pack, they're
00:07:43
actually head and shoulders ahead of an even better pack than Federer and Nadal etc was or Rod Lever was 50 years ago.
00:07:50
So that that's my my first level take. Okay. >> I don't know how you work out other than
00:07:56
just a little bit from year to year whether the pack is is getting worse. >> So Jeeoff, let me ask you a question.
00:08:01
Let's just quickly if let's imagine everyone gets 30 40 50 ELO points better
00:08:09
um since the ELO model isn't a linear model would that make it so that the other let's call it top 100 the other 98
00:08:16
players in the draw it's really 126 but let's say the other 98 players in the
00:08:20
drawer does that make it less likely then that one person can be as dominant because
00:08:27
it's a nonlinear model the top person's 30 points better now in simply models
00:08:32
just the difference in ELO ratings. So if everyone goes up linearly, the difference cancels out. But what are
00:08:37
your thoughts about kind of the nonlinearity that you know now you have to beat 98 better players and how do you
00:08:43
think about that? >> Well, I think that's exactly what we've seen the last decade or so in the
00:08:48
women's game. Uh before Sabalinka and Fionek have emerged as the their own top
00:08:54
two. I mean, and even EGA struggling this year. Um, we had a run of like 12 slams in a row where different women won
00:09:01
the title. I think I think that's the number. And it Yeah, you have you have
00:09:06
players climbing the rankings when they're number 37 knocking out Fiontech on hardcourts or somebody beating Koko
00:09:12
Goth because Koko Goth has an off day serving and they take advantage of that. Like, every single match is a contest. I
00:09:18
mean, there's no there's no easy matches in in women's tennis anymore. There
00:09:21
aren't very many easy matches in men's tennis. I think the only thing that's
00:09:25
keeping the possibility of a top two or a dominant top player is just the structure of tennis that you don't have
00:09:31
to win 58% of points to be dominant. You got to consistently win like 53 and a half%.
00:09:38
>> So, if you can figure out how to do that um or even 52% and be better in the
00:09:43
clutch. If you can figure out how to do that against a stronger field, then you'll still be dominant. I mean, not to
00:09:48
say that's easy. That's not easy at all. But, uh but that's all you have to do.
00:09:52
If if you if you had to win 58% then I think we would see Yeah, we would see that the that strength at the top start
00:10:00
to weaken. >> Audi, why don't you jump in here? >> Yeah, I mean the observation of course that the field
00:10:08
always gets better is true, but really that's an average observation. The extreme values um can have an enormous
00:10:18
amount of variability and that's what we're discussing. And so it's not um so
00:10:23
it's it's so what you're talking about is the difference between one and two
00:10:26
and three, four, five, six, seven, not between what not what the whole average field is doing which can can continue to
00:10:32
elevate yet you can still not have um elevation at the extremes. That's that's
00:10:37
that's a kind of a different mathematical question that it behaves differently mathematically. So the and
00:10:44
tennis is actually an interesting sport because it's one of the few one of the
00:10:47
sports that has huge numbers of what I would call ladders of of of levels. So uh being on two different levels of the
00:10:54
ladder means you you consistently beat one level below it. In in a sport like baseball, there aren't very many
00:11:01
ladders. I mean obviously there are differences, but any even the worst team, we talked about this last week,
00:11:05
even the worst team could be the best team on any given day. It's not even uh
00:11:08
it's not it's not even that lopsided. But in tennis you go from like one to
00:11:13
what what 20 or what rank? I mean typically I mean you right now we have the top two beating three four five with
00:11:19
nearly nearly 90% probability and that continues to shift all the way down. You don't you have these you know a 15 I
00:11:28
mean I it's not an it's really like an average but number the the fifth best
00:11:31
player will beat the 10th place best player pretty handily. Um at least we're
00:11:35
seeing that in >> that's just not that's not true though. I stop you there. You're absolutely
00:11:40
right about one two against number five. >> The number five beats what number with
00:11:45
what consistency? Like the same the same one two beats five at some consistency and five has to play X to have the same
00:11:52
dominance. >> I see. >> Okay, so I'm looking at my ELO ratings now. Alcarz is 2270. Taylor Fritz is
00:11:59
number five at 2030. So there's a 240 point gap between number one and number
00:12:05
five. go down 240 from a 2030. 2030 and you're at 1790. Is my math right on that?
00:12:13
>> Yep. >> So, for the same gap, 1790, scroll down. Scroll down. You get down, you're out of
00:12:21
the top 50 now. Um, you're down to 53 or 54. That's So, anybody in the top 50 has
00:12:29
a better chance of beating number five than number five Taylor Fritz has of beating number one.
00:12:35
And is that the kind of difference we typically see with a with a one, two, three compared to five or this is
00:12:41
extraordinary? I think so. >> Yeah. >> No, I think that's I think that's pretty
00:12:44
typical. I mean, you there have been times where one through five have been more tightly clustered like in the big
00:12:49
four plus Vanka era. Maybe there was a couple years where that was the case, but no, I think that's um that's pretty
00:12:55
close. >> This is an interesting general topic, Audi. We've talked about it here and
00:12:59
there, but it's it's an interaction between the distribution of quality and
00:13:03
the and the and the sport, the game design of the sport itself. >> And that combination gives you these
00:13:08
changes and it really says a lot about what we should expect. Let me ask a different kind of question.
00:13:13
>> There's a lot of points in tennis, right? So, um it's hard to it's what's
00:13:17
actually kind of incredible that I'm surprised about is that five takes 50 to
00:13:21
get to get that gap. By the way, I think Jeff, you could ask Jeff the same question, by the way, Audi, but say not
00:13:26
for the majors, but how would that change potentially for shorter matches like the Masters 1000, which are best of
00:13:34
three? Well, the So, the the numbers I gave you are for best of three, but I think the ELO stratification is the
00:13:42
same. It's just the the uh the probabilities those work out to. So, let's see. 240 should be
00:13:50
>> it'll shift both of them but it'll shift them the same way. It'll sit from the
00:13:53
same amount is what you're suggesting. >> Yeah. >> The conversion of the delta into a
00:13:58
probability is going to have a different constant if it's a shorter match. So
00:14:01
that's what's >> that's my point. Right. Exactly. That was my question is that but I think
00:14:06
audience by design I mean that is you're right or not by that is a design. Longer
00:14:10
matches favor the better player. >> Oh yeah. So Jeff, you've been writing I
00:14:17
think you know pushing 15 years now. Tennis analytics has advanced a lot in that stretch of time. How would what
00:14:24
would you say is the way it's affected the game if it's affected the game? So
00:14:28
if someone watched the US Open this next couple of weeks and they hadn't watched
00:14:33
tennis in 10 or 15 years and they remembered what style of play they saw 10 or 15 years ago, would they see
00:14:40
differences? Do has the analytics changed the game? Do you think in some way? >> Um, yes and no. And I think what it has
00:14:50
what it has changed for players and coaches is the awareness that the average point is a short point. Um,
00:14:57
there were lots of short points before. I mean, if you go back 50 years, you have a tour full of servant volers. Even
00:15:04
a lot of successful women were serving volleying. Um, which you virtually never see now. So, those were very short
00:15:11
points. lots of serves, aces winning the points, returns not coming back, first volleys winning the point. I don't know
00:15:17
what the average point point length is, but it's like three and a half strokes
00:15:20
even back then. Um, still now it's three to four strokes is your typical point.
00:15:26
Now, that's pretty much always been true. If we go back 20 years outside of the time frame you suggest, you get a
00:15:31
lot of big servers. You have like Richard Krychek and Greg Greg Rosski and Mark Filipus and those guys who are who
00:15:37
are giving you short points because that's just the game they played and it worked. Um what you have now instead is
00:15:43
everyone is aware of the fact that even if you don't have like the Mark Philipus
00:15:47
Scut serve that you should be playing with that in mind. You should be setting up your serve for a plus one. You on the
00:15:55
women's side there's a huge shift towards um swinging big on the return just taking your chances hitting a big
00:16:01
return and maybe ending the point there. Uh so what it has done statistically hasn't really shortened points so much
00:16:07
as it h the game has selected for players who will who are always looking to end the point if you're that's one of
00:16:14
the big flaws for Alexander Zerv is he's very good at hanging in long points
00:16:18
maybe as good as anybody else on tour. He can keep himself alive. He can make you a little uncomfortable and he he'll
00:16:23
give you these highlight reel like 4050 shot rallies and you know what a 40 or 50 shot rally gets you? it gets you one
00:16:30
point and then you're gassed after that. Like it's it's not worth it. And anyway,
00:16:35
I I wrote something about this earlier this year that you do that and you still don't have a great chance of winning.
00:16:40
The very best players at long rallies win like maybe twothirds of long rallies. So if you're, you know,
00:16:47
Alexander Verv playing Riley Opelka, monster server, bad groundstroke guy, Riley Opela is still going to win one
00:16:53
out of three of those points. So that's it's not the strategy. It's not a
00:16:57
strategy that wins tennis matches. and it wasn't ever dominant, but it was always present in that you'd find
00:17:03
players who who played with an eye on stretching out points, defense, and that's it's not gone, but especially on
00:17:11
hard courts, it's basically gone. >> Okay, Jeff, I was in I happened to be in
00:17:16
France when Michael Chang had his big French Open blow up in like 1989 or whatever that was, and he's the first
00:17:23
person that comes to mind. It's just kind of the long rally, get the ball back, he's not going to win any big
00:17:28
shots, but he's going to stay on the court forever. But that was 1989, and even that was a kind of a short moment
00:17:34
in time. >> And that was that was clay court tennis. And if if you watch court tennis now,
00:17:39
you'll see some people who are still playing that way. But it's fascinating
00:17:43
to watch even the the grindiest like small underpowered claycourt guys. They are swinging big. like they'll hit these
00:17:51
spin serves out wide to open up the court and then they just wind up for the biggest forehand they can hit. It won't
00:17:57
usually be a winner because you know these guys are 5 foot nine and they don't hit that hard. It's all just spin
00:18:02
spin spin spin spin but they're trying. Like the goal is still the same. Shorten
00:18:06
points, force the guy out wide, create an opening, get it over with. >> Okay. All right. Well, listen, uh Eric,
00:18:13
what what are you thinking about as you go into the US Open and are you going out there this year? You often make it
00:18:18
out to watch a little tennis. Um, I' I'd be curious to hear any specific
00:18:21
storylines you're curious about. >> Well, I am going next week to the US
00:18:26
Open. I'm going on Wednesday, which is Wednesday evening, which is men's and
00:18:29
women's quarterfinal, which should be interesting. Um, but you know, Jeeoff,
00:18:34
I'd love your thoughts on for some reason, at least at these rounds, I find the women's game much more interesting.
00:18:41
I mean, the match, I'm sure you watched it last night. The match between Venus
00:18:45
Williams and Mukova was incredible. I mean, the fact that Venus played that well at 45. Um, I mean, in that second
00:18:55
set, she looked great. How do you think about the women's game right now where I
00:19:00
could literally list certainly the top seven or eight players I could list at least seven or eight players that could
00:19:06
win it on the women's side and it might be even more than that if you include
00:19:10
Andre Evva you know certainly Rabbakina can win certainly the big three of Sabalena Goff and Swante I mean there's
00:19:18
at least five six seven eight 10 players that could win it how it's the women's
00:19:22
game that's got me much more because of the variety and diversity how do you
00:19:26
think about it. >> Yeah, I I think the women's game has a reputation for that variety and
00:19:32
diversity. I'm not sure whether I buy that narrative anymore. I don't disagree
00:19:35
with your main point that it's more interesting game. Uh like I was just uh talking about with the in the broader
00:19:42
trend, points are getting shorter. Someone like Mukaba is super interesting. She can play an allcourt
00:19:46
game. She can grind it out from the baseline. She can do a lot of different things. She is a dying breed. I think uh
00:19:53
more likely you have you're going to see somebody who's sort of a mini Sabalanka
00:19:58
um who will swing big on return, hit big serves. There's a a lot of women like
00:20:03
that. Um I mean one >> Anna Samova as Anna Samova as an example. >> Anna Samova is a great example. Yeah,
00:20:09
huge hitter. I mean she's a little more interesting in the sense that she doesn't if you saw her walking down the
00:20:14
street you wouldn't think, oh she scares me the way that Sabalena might. Uh but
00:20:19
yeah, she hits a Oapeno. I mean there's lot I mean there's lots of these big
00:20:24
hitter on every ball. >> Yep. Absolutely. Um so the so there's variety in the sense that there's some
00:20:30
women like that and there's women who aren't quite as extreme. I mean certainly Shantek balls in that
00:20:34
category. Uh but yeah I mean I think there's it's less likely that the serve
00:20:40
itself is going to end the point. It's it's more likely you're going to have a
00:20:43
player who's who's going to end it on the second ball. You don't see that as
00:20:46
much in men's tennis. Whereas with people like Osta Peno or Anosimova that you mentioned, you are going to see
00:20:51
players who who are really geared to ending the point or forcing the issue on the second ball. Uh and then yeah, you
00:20:57
get you get the the Mukova, someone like Koko Goff can rally with the best of them. Shriant can rally with best of
00:21:02
them. So there's certainly some some diversity there. What you're never going
00:21:06
to see like you'll see in the men's game is you two people who are just ace ace
00:21:11
ace ace ace. And that's always been the knock on the men's game even if it
00:21:14
wasn't true of everybody. uh you won't get a match like that. So, you're spared
00:21:18
that in the quarterfinals. Uh you just might get a lot of return winners, which I'd argue maybe not much more
00:21:25
interesting, but uh but it is a different strategy. >> Let me ask a question. What role do you
00:21:30
think it's a follow on Kate's previous question, what role do you think analytics plays in tennis? And here's
00:21:35
what I mean. Like, I don't need advanced analytics to go maybe I should hit more
00:21:40
balls to Federer's backhand. Like, they knew that back then. And I don't need
00:21:46
advanced analytics and ball tracking to know that. So let's call it, you know,
00:21:50
from a statistical perspective, let's assume that they could assess main effects like, wow, the first serve
00:21:56
really matters. Wow, this person's got a weaker backhand. Wow, let's try to, you
00:22:00
know, hit more slices. This person has trouble getting the low ball. Let's assume all of that was known even in the
00:22:05
McEnroe, Borg, Connor's days and all that. What impact do you think analytics
00:22:10
has had once we take away I'll call it the main effects? Yeah, those that's a good point. I'm I
00:22:16
am quite skeptical over the of the overall influence of analytics. I find it all very interesting from a a
00:22:23
postfacto kind of perspective. But but yeah, if I if I were a coach, there would be a lot of information I would
00:22:30
not pass on to the on to players. And for as long as I've been doing this, I'
00:22:34
I've been struck by how little coaches tell to players when they have the opportunity. There was a span of time in
00:22:39
the women's game where they would do encore coaching by having the the coach
00:22:44
come down and and talk to the player uh on the sideline. So you'd get some insight into that and they'd come down
00:22:50
with, you know, two really basic points and like you say, everybody knows hit to
00:22:53
Feder's backhand, but the coach would come down and say, you know, just focus
00:22:57
on getting your serve deep, hit to their backhand. Like are you serious? Like that I heard that stuff when I was
00:23:03
getting coached when I was what, like 13 years old. Is this really >> what it is? But I mean, one perspective
00:23:09
on that is tennis is hard. I mean, it's all coming at you fast. You don't have a
00:23:13
lot of time on every ball to sit there and think like you're not like a basketball defender who's thinking like
00:23:18
how can I edge this guy, you know, one foot to the left or into this zone or that zone. Like that would be nice if
00:23:24
you could do that. But if you're facing a 120 mph serve in the corner, you're
00:23:28
not thinking how do I edge her back into that corner. You're thinking, how the
00:23:32
heck am I going to put this in play? So I think that's the issue with virtually
00:23:36
ev every question like that. Some of it you can build into training and I think that's what coaches do is they they
00:23:41
structure their training around analytical insights but in match I think you you you fight with the tools you
00:23:48
come in with. >> Uh we want to go to Audi but real quick followup that I was going to go that
00:23:51
direction. The self scouting has to be a part of it right because even if you knew you we've heard we've heard this
00:23:57
across sports actually. Even if you knew that you needed to, you know, rotate your hips a little bit more when you're
00:24:04
swinging a baseball bat, to be able to be told precisely where you are and where the average person is and where
00:24:09
the best people are, that really helps on the coaching side. It's got it must
00:24:12
be the case that in tennis there's that kind of self scouting and say, "Well,
00:24:16
you're much stronger in this part of the court than that, or this stroke is better than that, just for your own
00:24:21
development in the training regimen." And then Audi's going to jump in here.
00:24:26
>> Yeah, absolutely. And and yeah, I think that's where most of it's happening like
00:24:29
I say on the on the coaching side and and more self- scouting because yeah, you have maybe maybe one full day to
00:24:36
prepare for an opponent that your your coach can can scout a player and prepare a game plan. Sometimes not even a full
00:24:42
day, maybe not even a chance to practice with that in mind. So I mean the self scouting is certainly the the biggest
00:24:47
place where there's an opportunity to exploit. >> So sometimes even the most obvious
00:24:52
things are missed. So, for example, like in baseball, analytics told told um most
00:24:59
base dealers to cut it to don't do it. It was negative. Um and and that only
00:25:05
certain players should do it into certain times. We're undoing that slightly because of the way they've
00:25:09
changed the rules. Um it also showed people that they're standing in the wrong spot. For 100 years, they stood in
00:25:14
the wrong spot and and and the fielders are now standing in the wrong spot and the correct spot and it's just just
00:25:19
changed the game. So going back to tennis, I've had I run um research seminars atmies over the summer and
00:25:26
every now and then a student wants to do a tennis project and they they ask questions and it most of those have
00:25:32
fallen down pretty hard because it's hard to get the data and it's and the
00:25:36
analysis is hard. But I I'll I'll throw out one of them which I wonder whether
00:25:39
or not um analytics could potentially answer if you had the right data. So there's obviously a there's some
00:25:44
trade-off between first serve um and second serve max velocity, right? Um and has that been analyzed? Like if you're
00:25:54
if you treat your second serve like it's a first serve, obviously giving up more
00:25:59
double faults, but maybe getting more second serve aces, what's the optimum?
00:26:04
that's some seems to be something that that is a hard question that that I would imagine isn't known or would take
00:26:12
some serious analytics to solve. Um is that a problem that is worth considering because I have some students trying to
00:26:18
do this couldn't actually get the good data but but um that would be like a problem that I would imagine could
00:26:23
potentially have some impact or no? >> Well, there's the the yeah the I mean
00:26:28
the simplest version of that question is should players hit two first serves? Um,
00:26:32
and that question has that that question has an answer and the answer is no. Um,
00:26:37
it's it's it's not a huge difference to to the typical uh typical approach, but
00:26:42
you you'd lose maybe like 1% or one and a half% of the points you win if you
00:26:45
>> Let me ask a question, Jeeoff. Would you add variance, which could be good for
00:26:49
the weaker player? So, you lose mean but gain variance. And I want to gain variance.
00:26:55
>> Um, you gain double faults. That's the problem. I mean, aces. >> Yeah, you gain aces and you gain double
00:27:04
foil. This is the if you're if you're if the if you're the severe underdog, and
00:27:07
we've just learned from from our from our discussion that most players are severe underdogs, um they've got to take
00:27:13
chances, which means they're going to lose badly a lot more often. Um but they
00:27:18
might also be more competitive a little bit more often, too. Yeah, I think I think that's why you see
00:27:24
players swinging away on return because where you want variance is on return. On
00:27:28
the serve, at least in the men's game, especially true in the women's game, but
00:27:31
more so in the men's game, even if you're playing Alcarez or center, if you're a pretty good player, you're
00:27:36
going to hold serve most of the time. >> Um, that's not your concern. You know,
00:27:40
you can have a you can have a good day and you're going to hold serve every single game against Senator Alice.
00:27:46
You'll still probably lose in the tie breaks, but you will you could you'll
00:27:48
hold serve most of the time on the return. That's where you want to add variance. If the classic example is
00:27:53
Dustin Brown knocking out Raphael Nadal I think 10 years ago now. I mean he he just went went crazy on the return. Uh
00:28:00
he I mean Chip in charge came in behind everything he could hit big like and it worked. It probably n out of 10 times.
00:28:08
This is a great this a great analytics point which is you know you focus your efforts and ROI on the dimensions that
00:28:14
matter and even you know it's the problem that Fritz has against Alcarass isn't holding serve it's just that he's
00:28:21
not going to get breakings he's not going to break the serve that often and so you might as well add the variance to
00:28:27
the dimension where you have more to gain that's a great point great general
00:28:30
business point let me just comment >> Jeff you're about to make some point
00:28:34
nine out of 10 times what they're going to be humiliated >> it's not going to work
00:28:37
Oh yeah, but I mean you are going to be but that doesn't matter. You're I mean
00:28:41
you're going to lose anyway if you if you're like Vit Copa just lost to Yannik
00:28:45
Center one and one and one. I think that that was just a couple hours ago. So So
00:28:50
yeah, that's the alternative. If you go in big with a wacky game plan, then yeah, maybe you will knock out Yanik.
00:28:56
Maybe it's one in a thousand for Capria versus Ser. But yeah, the alternative is
00:28:59
you play your game, you are gone in 80 minutes and you won three whole games in in three sets. So yeah, I mean that's
00:29:08
not that great either. You might look stupid, but if you I mean I'd rather I'd
00:29:12
rather win. But to go back to your u your question auding the serves is I think the hardest thing about that
00:29:18
question is knowing how practical it is. So I mean what I think that presupposes
00:29:23
is that a player can have like a first and a half serve. >> Uh and in in theory, yes. So there's a
00:29:32
couple researchers whose names are escaping me at the moment who have looked at this at least from a purely
00:29:36
theoretical perspective. They looked at every match and what they what they basically tried to do was figure out how
00:29:42
effective each player serve was based on how hard they were hitting it in that match. So they would say, you know, when
00:29:49
when they were hitting at 123, they would win 73%. 122 they would win, you know, whatever percent and work out a
00:29:56
curve based on that relationship. So they could say, you know, we don't want
00:30:00
to go all first serves, but if we could go first serve and we could go 1.7 serve, then that's your optimum. So you
00:30:09
can you can work it out mathematically, but what I don't know and what I what I
00:30:13
doubt, frankly, is that you can take that information to Taylor Fritz and say, "Okay, now give me your 1.7 serve."
00:30:21
>> He doesn't know what that is. He's been he's been hitting the same first and
00:30:24
second serves for over a decade of his life. He's perfected those deliveries.
00:30:28
So, so let me let me counter that though. Taking another page from baseball is they have these high uh
00:30:33
speed cameras and immediate feedback. I can imagine and they've used this to to
00:30:38
essentially tailor design pitches for individuals. I can imagine if you sat with the with the player and say, "Okay,
00:30:44
your first serve is 125, right? And your second serve is I don't know what the
00:30:48
right number is. Uh I don't even know what it is. 80. >> Say 105. >> 100 105. I want to develop a 115 and
00:30:56
we're going to sit there with the equipment until you can do it. And and is that is that is that an outrageous
00:31:02
proposition? I mean, or or is it with with the >> Yeah, >> I would say in the state world of
00:31:07
Tennessee that >> I keep thinking of that Andre Agassi. I don't know if it's commercial or show
00:31:12
where it's him versus Feder and he goes, "Now I'm going to give you a 113. Now
00:31:16
I'm going to give you a 109." And Agassi is like hitting the gun on every single
00:31:22
number he just said. Like he knows what a 113 is. He knows how hard to hit. Now maybe that's Agassi just fooling around
00:31:28
or cuz he's Andre Agassi and you know Taylor Fritz ain't Andre Agassy. >> Yeah. I mean there's only one Andre
00:31:35
Agassi for sure. I would doubt even Federer could do that and Federer would be the second guy I'd think of who might
00:31:39
be able to do that. Um it would be a pretty out there thing to suggest to a player. Um,
00:31:46
and >> but it's a it's a neat idea because it adopts probably the biggest revolution
00:31:52
we've seen in baseball in the last five or six years. It's player development
00:31:56
using a lot of technology and it's transformed some players careers. Now, one question might be whether it's
00:32:02
valuable enough to be worth the effort, but it's a neat hypothesis. borrowing
00:32:07
from another sport into into this sport >> would be the >> absolutely I mean I I would love to know
00:32:13
that someone was out there trying it. Uh and maybe maybe it just needs to take a
00:32:19
little more time because I think when players are really developing es especially
00:32:24
well I'm not sure whether I'm going to pick a gender on that but when players
00:32:27
are really developing their adult game there I mean they're they're 12 or 13
00:32:30
years old. I mean there are exceptions to that rule but they're figuring things
00:32:33
out when they're pretty small. So even prospects we're seeing now like they
00:32:39
>> their game has been pretty much in place for a long time. They're not going to
00:32:43
>> their tennis schedule doesn't allow them to take 6 months and you know work out
00:32:46
their 113 serve >> uh the way that a baseball player could or they're certainly not going to take a
00:32:51
year off tour because you know they got cut the way that a baseball player might.
00:32:55
>> So it might and we might see it when you know the developmental period catches up
00:33:01
with the uh with the length of time we've known this stuff. We're gonna need to let Jeff go. Eric is
00:33:07
asking for one more question. Eric, is it gonna be a quick one? Can you jump in here?
00:33:11
>> It is. It's a quick question. So, Jeeoff, I would also imagine that the
00:33:14
answer has to be what was called a mix strategy. And what I mean by that is if I knew your second serve was always 115,
00:33:21
that isn't doing squat either. So now I'd have to have the first serve. And
00:33:25
sometimes I hit it 105 in the corner. Sometimes they hit it 115 cuz if everybody knew Jeff Sackman was the 125
00:33:32
115 guy, they can return 115. So it would have to be I mean there'd have to be a mixture no matter what you do.
00:33:41
>> Yeah. And I mean to some extent there already is. I mean I I was simplifying
00:33:45
to say that you know Taylor Fritz has a first serve and a second serve. Taylor Fritz can hit a whole ton of serves and
00:33:50
they are all pretty consistent and they're all pretty reliable but it's a
00:33:54
it is still a limited number. So I mean it it might be four different first serves and the real magic comes from the
00:33:59
guys like Federer or Ashardy who can completely disguise them. So it's like a
00:34:03
pitcher throwing four different pitchers out of the same arm slot. >> Uh but but that's rare and even even
00:34:11
then that that's enough variety. I mean, and going back to tennis just being
00:34:14
freaking hard if you're facing Taylor Fritz from the line, then if if your opport your options are, you know, 125
00:34:21
that way, 118 that way, 106 with slice that way. I mean, good luck, man. We we don't need much strategy to just to say
00:34:31
you're probably not breaking that guy today, >> right? All right, Jeff. Thanks. Always a
00:34:37
always a pleasure to talk to you. um wish you the best with the work that you're doing and wish you um good fun
00:34:42
for the next couple weeks with the with the US Open. >> Thanks very much. Thanks for having me,
00:34:47
guys. >> Absolutely. Jeff Sackman, the best in the business in tennis analytics. You
00:34:52
can read his stuff. His blog is called Heavy Top Spin and you can find it at Tennis Abstract. Tennis Abstract is an
00:34:58
organization that he founded. Jeff Sackman. That has been the first half of Wharton Rainbow. Come back and join us
00:35:04
after the break for the second half. Welcome back. Welcome back to Wharton Moneyball. Welcome back to the second
00:35:12
half of this week's show. Second half of a full hour of sports analytics here on
00:35:18
the Wharton podcast network. This is Kade Massie hosting this week along with Eric Bradlo and Audi Winer, two
00:35:25
longstanding co-hosts and colleagues of mine. Shane Jensen is out this week. Shane is out doing Shane things here in
00:35:32
late August. We are just off the line with Jeff Sackman. Always a pleasure to talk to Jeff. It's as much of a pleasure
00:35:39
to read him. If you have any interest in sports analytics, you should be looking
00:35:42
at his work on tennis analytics. It's as good as there is and it has been. He was
00:35:46
kind he was one of the first real sharp quants in tennis. Guys, late summer, um,
00:35:53
we've got football dead ahead. We've got college football week one, a full slate.
00:35:57
We'll talk a little bit about that. And then we're going to have an overtime
00:36:00
segment, an extra segment with Ralph Russo to talk in more depth on college football after this regularly scheduled
00:36:08
show. We've got some other big stuff going on though. Golf. We've been talking golf off and on this summer.
00:36:15
It's a writer cup year, which makes it all a little more interesting. And then
00:36:19
we've just come through the I Eric, you have to remind me what they call these
00:36:22
things. The the FedEx or whatever it is these days. >> FedEx Cup. Y >> the the FedEx Cup. But but the tour
00:36:28
championship, it's really it's not a major, but it's one of the bigger non-
00:36:31
majors. And amazingly and finally, the long national nightmare, long international nightmare is over over
00:36:38
Tommy Fleetwood won a PGA event. Eric, were you watching this thing? This was the thing that they cut down to 30
00:36:44
golfers. The last round is the last tournament is just 30 golfers. Fleetwood got the lead, I think, on day three and
00:36:51
then he closed the day on closed the deal on day four. >> Yeah, I mean, I watched almost all of
00:36:55
it. Um, I thought it was great sports, great TV. Um, the good news is, you know, all the golfers and so, you know,
00:37:04
as you said, it is the top 30 and, uh, it was a tough course, an interesting course. Um, and the thing that's
00:37:11
interesting is maybe this is, you know, it would be interesting to talk or, you know, maybe I'll message something
00:37:16
towards Rufus because this is one of those situations where I felt like if you looked at Tommy Fleetwood's, whether
00:37:23
it's an ELO rating or world ranking, he was probably his chances of winning were
00:37:28
probably downgraded too much because, well, he had only come in second seven times in the last year. All right, find
00:37:36
me another golfer other than Sheffler that's had as many top five finishes as
00:37:41
Tommy Fleewood. So, because he didn't get the number one, all of a sudden, we're downgrading him so much. I mean,
00:37:48
to me, it was almost not certain. It's never certain in golf, but I'm not
00:37:52
surprised he was in the top five. And if he's in the top five, why can't he win
00:37:56
it? And so I I almost feel it was like an outcome bias because he hadn't won
00:38:02
even though his world ranking was so high and he was in good form that all of a sudden he's not winning.
00:38:08
>> Well, that's a that's a that's great. That's a great question whether people
00:38:11
overb the the story. The narrative was he can't get it done. But let's be clear
00:38:15
about a couple of the stats right now. Now this is post tournament, but it wouldn't have been that different before
00:38:20
the tournament. Data Golf, which has the most predictive publicly available world
00:38:25
rankings, much better for predicting performance than the world golf rankings. Data Golf has him number three
00:38:31
in the world. All right, so just behind Rory, who's obviously just behind Shuffer, a stroke behind Shuffler. Um,
00:38:37
the other thing is he has 30 career top fives on the PGA without a win before this tournament, which is the most in
00:38:46
100 years. And so that's where the narrative has traction and you're questioning whether okay fine but did it
00:38:52
get overb essentially. And here's my empirical question for y'all. Do y'all
00:38:55
think we could model whether someone plays better with the lead after having won a tournament than
00:39:03
before having won a tournament or do you think there's not enough data to do so?
00:39:07
This is like career level modeling as opposed to you know any given tournament. Do you think do you think we
00:39:12
could observe a difference? Can we test and if there is a difference, find it that they perform better or differently
00:39:19
with a lead after having ever clinched a tournament? >> Yes, I do. I mean, but it depends how
00:39:26
far you're willing to assume stationarity. And I mean by, you know, there's basically almost a golf tourn
00:39:31
let's say 40 weeks of the year and then let's say we take the last 8 to 10 years
00:39:36
of golf tournaments. So now there's 300 and something observations where we've
00:39:40
had a winner. And again, if you only want to model the probability of winning, that's different. We could also
00:39:46
say, does the person perform better, which is, you know, their world golf ranking or their, you know, their data
00:39:52
golf ranking would have been the person should have come in seventh. Like it could be exceedences of their ranking.
00:39:57
We don't have to look just at do they win more often. Do they play better once
00:40:02
they've won? Is it like a change point model kind of thing? But I think the answer is yes. I think there's enough
00:40:07
data. If let me just say though this is probably your all other point and Audi will I'm sure be happy to jump in on
00:40:13
this the effect size is probably smallalish. So your question is it's not can we model it of course we can model
00:40:20
it but is the effect big enough that we'll see it >> right I mean this goes Audi's one of
00:40:26
Audi's great lines early in our show and then we kind of learned it is that he's
00:40:30
not saying that the effect doesn't exist it's just too small to observe. So, he's
00:40:34
talking about a clubhouse effect, some momentum. Like, sometimes these things might be there, but they're just too
00:40:38
small to this. >> I wasn't even thinking this is momentum, but sure. >> Since we're on golf, guys, since we're
00:40:44
on golf, it's time to point out that even something which we know is important like putting is nevertheless
00:40:53
the effect is too small to to I would say know with strong statistical confidence who is better and who's
00:41:00
worse. That's an that's an in golf even something that we can measure is is uh
00:41:06
nevertheless doesn't vary that much among the players. So when you get to Fleetwood in particular and you look at
00:41:12
those 30 tournaments, what did he what did he come short in? I mean what was I mean what aspect of his game because I
00:41:18
can imagine someone not closing because you know putting just didn't come through.
00:41:23
>> I love it. That's great. So this so we I don't know how much we've talked about
00:41:26
this. We need to do a full segment on it, but Audi has some research just over the summer, I think, finally.
00:41:31
>> Yeah. >> How random putting is and therefore how difficult it is to say some guys are
00:41:37
actually better or worse putters. And I think I mean, you could give us a quick empirical observation. It's some very
00:41:43
small number of guys that you can reliably say are better. >> Yeah. So what so what happened with with
00:41:48
putting is that you get we get we did what's called empirical bay shrinkage which doesn't impose um a knowledge of a
00:41:56
standard deviation on talent but doesn't we did impose a normal distribution on
00:42:00
the distribution of talent um which we could we could we could argue about right particularly the tales of that
00:42:05
distribution but what ends up happening is that we do have in any given season we we do have uh players who putt better
00:42:12
and butt worse but when we do our shrinkage or empirical based shrinkage those differences get really small and
00:42:19
among the golfers in any given you know two or 30 hund of them we do what's called an FDR which was uh we
00:42:26
essentially try to control the number of falsely discovered Good.