A Deep-Dive Conversation With DUPR CEO And Data Scientist On The Latest Pickleball Algorithm Changes

The DUPR algorithm changed recently.

DUPR

A certain segment of the pickleball playing community will never read this sentence, nor will they even open this article, because they think the concept of a rating in the sport is meaningless to their experience playing and enjoying pickleball. And that’s OK: you can absolutely play a sport and never need to know “how good” you are. You can play golf every Saturday and not need an officially litigated and tracked PGA handicap, you can play tennis and never maintain a UTR rating, and you can play chess online without ever getting an official ELO.

However, the concept of “ratings” in Pickleball is important to a certain segment of the pickleball community, and they’re vital to properly organizing any competitive function, and the leading methodology in the sport for determining your rating is DUPR.

DUPR, (which initially stood for Dreamland Universal Pickleball Rating as an homage to the ancestral home of both DUPR and MLP on Steve Kuhn’s Austin-property) stands for Dynamic Universal Pickleball Rating and is an algorithm and data-driven ratings system designed to universally rate every single pickleball player on the same global scale, independent of gender, age, or geography. Ben Johns and Anna Leigh Waters are at the top of the DUPR tables for Men’s and Women’s Doubles respectively for example, and every player who has a competitive match tracked by the database irrespective of age, gender, or ability will have a corresponding score within the same system. There are some competing ratings systems out there, some with different methodologies, but the industry standard for now is DUPR.

In mid-July, as has been widely discussed in various social media circles. DUPR changed the algorithm it uses to calculate and update player ratings. This change led to criticism and confusion in the player base, not unexpected since DUPR is based on Math (which can be a tough reach for right-brain people) and since it’s a change (which people generally struggle with no matter what the change may be).

After reading a ton of back-and-forth in the Pickle-sphere, I thought I’d reach out to the DUPR team and host a deep-dive Q&A session. I met with DUPR’s CEO Tito Machado and Chief Data Scientist Sarah Carpenter earlier this month and tried to ask the “tough questions” about the change. This transcript (lightly edited for grammar and clarity) is the result of that conversation.

DUPR’s team has done a great job getting the word out with this change. They’ve released a great FAQ online that answers some frequently asked questions about the change, some of which we cover here. Carpenter went on the John Kew Podcast right after the change was released to have a long discussion on this topic. Finally, Pickleball tournaments hosted a roundtable that included Carpenter to do live Q&A on 8/12/25, the rebroadcast of which is available here.

For those of you who like long-form reads though, here’s my Q&A session with the DUPR team. I’ll add some post-interview context in [square brackets] as needed.

Todd B.: Thanks for joining me Tito and Sarah! So, the reason that we’re talking today is because DUPR recently made an algorithm change, and of course any change in our industry is met with resistance and confusion. So, can one of you give me just like a 30 second “State of the Algorithm” or even a “State of DUPR” in general?

Tito M.: Sure, I’ll take the business side. DUPR is the biggest, fastest growing platform in pickleball. We have a million plus users, we are in 158 countries, and we run the biggest college tour in the world. We also run the biggest amateur tour in the world in Minor League Pickleball. We’re expanding into Australia and China and all over the world. We have 8,000 clubs registered, so we’ve truly created this massive ecosystem around the rating. You go to Malaysia, Japan or anywhere, anywhere you go people are gonna ask you, ‘What’s your DUPR?’

Sarah C.: Algorithm-wise, our goal is to give every player a rating based on their match data that is accurate and consistent in every way possible. We do that by providing an update to their rating instantaneously for every single match played. The algorithm looks at their match history as well as their performance in the match, and then compared to the expectation it updates their ratings accordingly.

Todd B.: How much of a lift was that from a technological implementation standpoint, to go from a batch-based process to an instantaneous update process for matches? [The context of the question is, DUPR started out only updating ratings of players on a batch-based cadence, meaning a bulk process ran nightly or weekly. Now, matches update your rating the moment they’re entered].

Sarah C.: It was in June of 2023 that we made that change. and yes it was a lift to get the system to go from batch to instantaneous updating, but it was well worth it, since this is one of the features that our users appreciate the most.

We used to take in every match and spit out a rating for every player as one bulk action, as opposed to now, where we look at each match one by one. But, we also do still do batch-based macro analysis of the entire data set whenever we’re testing the algorithm. We still look at it in the whole, and we test by running every single match through version A of the algorithm versus version B.

Todd B.: So for those that come from Tennis, which is a huge percentage of pickleball players, and who are familiar with UTR; How similar or dissimilar is the way that DUPR works today versus how the UTR calculation works?

Tito M.: So, I am a former UTR employee. DUPR is a little more complex than what people think. Like DUPR, UTR has the concept of “lose but still go up, win but still go down.” The big difference between UTR and DUPR has to do with why they were created in the first place. Initially, UTR was only for competitive players. It was all about rating college kids who wanted to play college tennis, and that was kind of the whole purpose, right?

For us at DUPR, we have recreational players as well. We have people from all ages, all backgrounds playing pickleball and having a DUPR keeping track. That’s the way that we look at our product. We also value transparency; we’re way more visible about the way the algorithm operates. We try to show literally everything that’s happening in the back-end and we are very transparent when we’re explaining everything that we’re doing on the back-end. UTR is way more black boxy.

Sarah C.: Like Tito said, the algorithms have some similarities. They’re both using a Spread-Based Model, meaning that there is an expected score for each match. Then, as an actual score is entered, it is judged based on how you do versus the expected score. In UTR, it’s going to be games-based because nobody tracks points when they’re playing tennis. But for us, the typical pickleball score you’re going to remember is 11-4, 11-6, right? So we’re looking at the score and the amount of points per game.

[Note, we get some more details on the Spread-based Model and its implications later on; this is an important part of the algorithm].

Also, UTR still does batch processing and daily updates as opposed ours is instantaneous. In UTR you have your overall rating, but we do show you how you get there through every match. Lastly, UTR does have movement behind the scenes. If you don’t play, you could wake up one Tuesday, and your rating may be different than on Monday. And for us, that’s typically not going to be true. You’re only going move when you’re playing.

Todd B.: So a follow-up on the “movement behind the scenes” functionality. Do you feel like that might be a future feature that you implement in DUPR? If someone doesn’t play for a while, do you assume their rating “degrades” a bit? Do you think that there’s value in that kind of movement?

Sarah C.: Yes, it’s definitely a possibility. It’s something that we’ve studied for a long time. Before we switched to instantaneous match update, such degradation was part of our algorithm. However players didn’t understand what was going on because they couldn’t see how they were performing relative to expected in each match. So, if we move towards that again, then we’ll try as much as we can to continue to be transparent.

If it comes to it, there’s a couple of ways to do it. You can do it where you just truly update things every night, similar to what UTR does. But you can also kind of have it work completely behind the scenes. That’s what happens in like video games (It’s called MMR). So, there’s lots of different options and we study all of them and kind of see how it goes.

Tito M.: I want to add something there, because we thought about this so much. I think movement without playing is one of the big issues that the players complain about. If I haven’t played, I don’t want my rating to drop, right? So that’s when we added the concept of the Reliability Score.

[Note: the Reliability Score was introduced in May 2024 and indicates how “reliable” DUPR thinks your rating, as affected by the number of matches in the system and the age of those matches. A full FAQ on the implementation is available here].

So now right, the reliability scores is 1 to 100 which that impacts the way you move up move down. So the big thing is that we need to make sure that the players and the community understand that the rating is obviously the most important but a rating without a high reliability doesn’t mean anything.

Todd B.: I think that that underscores a key point in in DUPR or in any rating system. Pros who play every weekend in DUPR-rated competitions will have 99% reliability, versus the casual player who might play a tournament or two a year. I think that’s where a lot of the criticism comes from; it’s from the casual players, who just don’t have enough data. DUPR can’t make an evaluation from someone who has two scores reported in a year and that needs to be understood, right?

Sarah C.: Absolutely. Yeah.

Todd B.: When you implemented the new system, do you go backwards and recalculate past matches? Or do you just say, ‘Okay, as of this date forward, it’s now changed.’

Sarah C.: Yeah, that’s a good question. No, we do not go backwards. We want people to know when they go out on the court how they’re going to be measured. And so we did it forward-looking, but our decisions are all based on past data if that makes sense.

As we are optimizing this algorithm, we are using all of our past data, all of our millions of matches in order to fit these models for the expected score. That is how we make the decision on what our formula is. And then we make it forward looking and continue to monitor it day by day. We evaluate how are things performing relative to the past and continue to make tweaks like that.

Tito M.: 99% of the time we’re dealing with recreational players. They play at the park and they use DUPR to be able to play at their round-robins, you know? A lot of people are new to racquet sports to sports in general. Maybe they haven’t played anything competitive since high school. So, we’re coming with this rating and we’re telling them this is representation of your level. We need it to be accurate, but also, it needs to be understood and people need to feel comfortable with the metric.

Todd B.: Fair enough. So on a high level, you guys started years ago with essentially this same logic, then you pivoted to what I call the zero-sum system and now you’re back. What was the impetus? Was it just as simple as, you ran some simulations. and you thought that you had a better way to do?

[Note, by “zero-sum model” I refer to the DUPR algorithm mechanism where if you won, you went up and if you lost you went down, and the amount of DUPR gained or lost by the winner and loser zeroed each other out.]

Tito M.: No, there was a little more to it. At that time, we were in the middle of the “tour wars” and there was a question about whether the merger would happen. We had way less data back then as well. There was a little more volatility in the industry. It was very early on, it was our infancy, and the criticisms were already very loud. So we kind of like, hit the reset button a little bit and we were like, ‘Listen, we need to simplify to then rebuild.’ The goal was always to come back to this spot right now, but we knew that the sport wasn’t ready for something too complex.

Our approach became, let’s have people understanding where we are now, then let’s add a couple of things and through time, we added point impact. Then we added reliability later on that year. So little by little we were coming back to where are today, but we came back in a smarter and more transparent way, so people could get comfortable to a spread model.

Todd B.: Okay, here’s an opinion question. The ‘zero sum model’ seemed to incentivize players to play down and to continue to play don because every time they won, they would tack on a little more rating points. Whereas the current system seems to incentivize players to basically play up and if I play in a higher division and I get a win, well then I’m gonna be rewarded heavily for that. Is forcing players to ‘play up’ good? Or bad? Was that a factor at all in your algorithm change?

Sarah C.: I think a lot of people think it was actually part of our decision- making. But to be honest, our decision-making really is around creating the most accurate rating as you can.

However, when we made this change, it became very obvious: there’s no incentive to play down to improve your rating. That’s just not gonna happen anymore, and we do see that as a good bonus. Like we recognize that like, Hey, this is actually really good but it’s not necessarily why we made the decision.

With that said, it is a common misconception right now that you can only play up to improve your rating. The way we built this formula is based on all of your past results, meaning that our expectation is just your typical result. So, a 4.0 playing against a 4.5, they have to do better than 11-5 in order to go up. So, it’s not trivial; they’re gonna do that sometimes. There’s gonna be other times where they lose 11-2 and they’re gonna go down, so you have to be performing on a consistent basis to move up.

Even like you said, if you get one win, sure that is going to move the needle. But then in all your other matches, if you’re not performing at that same level then you’re going to, you know, move to your true level. Just like with any change, you have to have time to see how things actually play out. And I think that was everyone’s major reaction before. seeing how it all plays out.

Also, when we were in the previous system when it was “win and go up,” It did feel like a reward system. People did feel that, right? They would win and they’re like, ‘Oh now I’m gonna go look at my DUPR and I’m gonna see it go up because I won. However we are not trying to be a reward system, and I think people are now experiencing that that is not our goal and it really was never our goal but it did appear that way.

What’s important for people to know is that we are just trying to measure them. We’re not trying to reward them or penalize them, but we’re just truly trying to measure them based on how they’re performing in their matches.

Todd B.: You mentioned in one of your previous interviews that you are constantly running simulations on your test data, to test the impact of a little tweak here or a little change there and you mentioned that you have like a baseline where you kind of help determine.

I’m just wondering, Do you have like a specific player who, you know, you use as like the perfect ideal player? For example, if its Ben Johns and if Ben Johns loses six DUPR points as he result of a tweak you know that that change didn’t work. Basically, how do you determine like when you run a simulation if it actually is working and helping versus not?

Sarah C.: That’s a good question and honestly, I have not gone into that with anyone yet so that could be like a unique part of this interview.

[finally! I ask a good question nobody else has asked!]

So, no we’re not necessarily looking at one person. We’re looking at the ratings themselves in a predictive way. We’re trying to be as predictive as possible in future matches either to correctly guess the winner or correctly guess the score. We can create our expectation and then we judge ourselves by how close were we predict the actual scores between version A and version B. We count literally, how many did we get correct or How close were we to the score across all of our matches?

Then we also look at some groups. How do we do at Mixed versus Gender matches? How do we do on high age differential matches? How do we do when we have low reliability players versus high reliability players? So we have all of these different sub-groups and we literally are just looking at which one is more predictive in as many places as possible.

Tito M.: People don’t realize how much data we have. And so it’s way easier to create these models when you have more data. We have all of these API partnerships sending in scores from all kind of games. We also have to consider the gender and age component. 20% of our of our games are non-traditional, meaning two females versus two male, you know, one female versus three males. So, we have a lot of different data points to be able to create this predictability assessment.

Todd B.: Sarah from one of your interviews, you mentioned some of these interesting stats about your data set and you just mentioned another in the 20% non-traditional What are some other fun stats about the data?

[Note, Sarah followed up post-interview with the following data points, to go along with some of the other metadata describing the DUPR data set. I’ve added all the metadata points plus what Sarah sent post-interview at the bottom of the post].

Sarah C.: The one that I that I normally share is that 75% of our matches have a gap of 10 or more years between the youngest and oldest players.

Todd B. So I’m gonna ask you a couple of loaded questions. These are the questions and complaints that we’re constantly seeing in Facebook right now. You know, this is the, ‘Hey I won 11-2, why did my DUPR go down’ question.

Do you feel and or have you tested if there should be a minimum score line that prevents any lowering of DUPR? [In other words, as an example if you win 11-2 or better you can’t ever see your DUPR go down ]

Sarah C.: Yeah, good question. So yes, we have tested it. Another thing that we’ve looked at is the Expected Score, the typical result meaning about 50% of the matches are above and about 50% of the matches are below a score line. Basically you want the favorite to go into the match understanding that they have equal opportunity to improve their rating, or to see their rating go down.

So obviously when you get close to a full point difference between the ratings of two players or two teams, you know, the Expected Score is going to become close to like 11-2. We are experimenting with like trying to be a little bit more forgiving in those places in order to keep that that balance. But at the end of the day, we kind of have to keep holding those higher rated players to that standard of performing at a higher level.

These are the exact type of things that we’re continuing to work on day by day. Behind the scenes, we will continue to make some tweaks. Obviously people know when we make big algorithm changes like the one this month. But I don’t think people know that we continue to make tweaks along the way that we’re not always necessarily going to announce because that’s just what we’re doing, always. Our main goal is to keep making this algorithm as accurate and as good of an experience as possible.

Todd B.: I mean would you ever consider making that particular tweak [a minimum score by which no DUPR loss can occur]? Just I’m so just from an optics perspective to stop the criticism. Even though someone loses 3/1000s of a point, which we know is meaningless in the grand scheme of DUPR, but is it worth it just to end that particular criticism from a particular part of the user base?

Tito M.: Todd. We have this conversation. Almost every hour of the day, man. We literally have this debate constantly. I got a DM yesterday on my Instagram from someone extremely upset because their DUPR went down like 0.04.

In DUPR, the big numbers are the five and the zero. Those are the numbers that we truly care about right, four point nine. That’s truly the representation of your reading. You going down a few thousands of a point, honestly, it doesn’t mean anything. But, people just don’t like to see the red score [which indicates a decline in DUPR] or the little arrow down.

So we are talking about this. We don’t want to lose the engagement because in many ways we know people are engaged to that movement and even if it’s a little bit or more, you know, there’s always a two sides of the story. One team left very happy that game that they won; we don’t get an email from the team that won and saw their DUPR rise. Instead, we get our Instagram post complaining about the DUPR going down a few thousandths of a point.

Todd B.: Okay, so here I got a question for you. That affects me personally because this has happened to me, you ready?

Sarah C.: Yeah, let’s go.

Todd B.: Should there be a minimum or a maximum amount of movement that an individual match can impact your DUPR score?

[For context, I played a tournament in late 2023 during a period of time where DUPR was hyper-adjusting scores and I had a low reliability rating, and I lost 6/10ths of a DUPR point in two matches in a 4.5 doubles draw]

Sarah C.: That’s a good question—and yes, we can do that. In fact, we already have caps in place on both the minimum and maximum movement per match for a player.

One issue we fixed in the past couple of years was that really active players weren’t moving enough. So now, we make sure there’s a minimum amount of movement to reflect growth or change over time.

On the flip side, you also have players who are just starting out, or coming back after a long break, and their ratings can swing more dramatically. That’s intentional. It’s kind of “damage control.” If we only have one match to go on, we want to move their rating as much as possible based on that limited information. But if we know they’re going to play ten matches in the same day, we don’t need to overreact to just one result. We can spread that adjustment out over several matches.

So, to be honest, that’s the kind of decision we make by listening to player feedback, and by trusting that the rating system will balance itself out over time.

Todd B.: I understand that. I mean, in my particular case I had a low reliability score. I might have had an inflated DUPR and so then a loss impacted my score much more than if I had like a 99 or 100% reliability.

Tito M.: The most viral example we saw during all the craziness was a post from someone who said, “I played one game and I lost 0.4, my rating dropped to 0.433.”

When we looked at that account, they had only one game in the system, and it was from the Philippines with no real connection to any other player. It was an unrealistic situation—the rating dropped simply because there was no data, just that single game.

There was another case with a player whose DUPR was badly inflated. They were competing in 4.5-level events but were listed as a 6.1—a rating you’d expect from a PPA-level pro. In reality, their matches were very tight, losing some and winning some against 4.8-level players. So the system worked to adjust them back down to the appropriate level, because they clearly weren’t performing at a true 6.1 level.

Todd B.: Have you have you ever thought about trying to account for fluke results? A scenario like, I sprain my ankle halfway through a match, I finish it, but we lose to a team that we normally would have beaten.

Sarah C.: It’s an algorithm approach option, but the only way we’d really consider it is based on how surprising a result is. For example, if the expected score was 11–2 with Team A winning, but instead Team B won 11–2, that would be a very surprising outcome. We could throw that out—but then one team would be really unhappy that we did.

So, the approach we take is to trust the law of averages. Some days you’ll have a fluke in the positive direction, some days a fluke in the negative direction—that’s just the nature of the game. Like Tito mentioned earlier, there’s always someone who’s thrilled because they went up 0.003, while someone else is upset about the exact same match.

That’s why we try to stay as agnostic as possible: take in all the data, use everyone’s scores, and avoid giving too much weight to any single match. Especially for reliable players, those fluke results won’t move the rating much either way—they’ll just be one more data point. And because every match provides data for four different players, it’s important to keep those results in the system.

Tito M.: Todd, I think this is super important when it comes to adding data. You asked earlier about playing down, and if we really want accurate—not inflated—ratings, the best approach is to encourage a good mix of matches.

That means playing with partners who are slightly below your level, at your level, and above your level. It also means playing with different partners—sometimes where you’re the stronger player, and sometimes where you’re the weaker one. Same with playing both the right and left sides.

When we collect data from all of those different situations, we can average it out to get a true picture of who you are as a player. I know people say, “Well, that game wasn’t fair—they targeted my partner, I barely touched the ball.” And sure, that happens. But then you might play another match where you’re the weaker one and you get targeted. That gives us another data point.

By looking across all those games, the system can measure your performance more accurately and give you a fair rating.

Todd B.: So, that leads me nicely to my next question. I think I know the answer to this question, but I’d like to get an official one from you. One of the criticisms about DUPR made by a certain unnamed person on Facebook (who then deleted the post) was asking whether there should be different DUPRs for gender, or for age groups. So my question for you is this: Is a 4.5 60-year old woman comparable to a 4.5 25 year old man. Why or why not?

Tito M.: Yeah, so I actually love this question. I recently moved to Boca Raton, and I play with some incredible 60- and 65-year-olds who consistently beat me—and I’m 32. They’re just better than me, and they compete at a very high level.

That’s why this concept can get a little complex. If we start dividing ratings into too many categories—an indoor rating, an outdoor rating, a right-side rating, a left-side rating—we risk overcomplicating things. Those can all be useful analytical tools, but the reality is simple: 65-year-olds who play at a 5.0 or 5.5 level exist. They’re out there, and they’re really good.

So, if you’re 60 years old, you’re rated a 5.0, and you’re beating 30-year-olds, then you are a 5.0. And that’s okay. The key is understanding that pickleball used to be structured around age categories—you’d be a “5.0, 55+.” But what we’re doing now is just looking at all of your games. We can see where you’re competing and how you perform, and that tells the story.

And honestly, that’s important. It’s okay if a 65-year-old isn’t a 5.5 on the overall DUPR scale, where you’re being compared to Ben Johns and the pros. Maybe you’re a 4.5 there. That’s perfectly fine. This isn’t about clinging to the “I’m a 5.0” label—it’s about building a system that reflects reality and creates fair, level-based play.

Todd B.: That’s okay except for the ego of the guy who still thinks he’s a 5.5.

Sarah C.: Yeah, it’s definitely an ego thing. One thing I’ll add is that another reason we don’t divide DUPR into categories is because of how fluid the sport is in every regard.

We mentioned some stats earlier—like how 75% of matches are played within a 10-year age gap, and how 20% of matches don’t fall into “traditional” categories. That shows just how diverse the matchups are.

And it’s not just about age—it’s about levels too. Pickleball is unique in how fluid the pathway is from amateur to pro. You can have someone who plays in club matches, league matches, and even rec “Moneyball” games, but that same person might also enter a PPA qualifier. That kind of overlap doesn’t really exist in many other sports.

That’s why it’s so important to keep everyone on the same scale—because, at the end of the day, everyone’s playing each other.

Todd B.: Do you feel that there is there has been “level creep” in the sport over the past few years. We all play and we all generally improve over time, so it has the entire sport kind of creeped up in terms of like the ratings?

Tito M.: Sarah before you jump into that one. I want to give a very interesting example about this when I was at UTR. There, with the modified ELO-based rating that we had, we actually had the OPPOSITE problem. Levels were shifting down through time.

We had situations where someone started with a UTR 16 rating, but by the end they had dropped down to a 15, even into the high 14s. College players who were 14 UTRs ended up dropping to 12 UTRs. We had to work through that.

But the nature of what happened here is different. In this case, because of the connectivity—and I’ll let Sarah expand on this—we’ve actually seen a shift upward.

Sarah C.: Well, it really depends on who we’re talking about. If we’re talking about the number one player, Ben Johns—he’s actually gone up.

Here’s something interesting, though. Someone on Facebook recently told me, “Ben Johns is only going to go down from here.” But the truth is, Ben is the exact type of player who’s been somewhat inflated over time. He just kept winning and winning, so his rating kept climbing. Now, since ratings are judged more on performance relative to expectations, his number may not stay quite as high.

Will his rating change a lot? No. But one thing we do have is the ability to make slight adjustments to control distribution based on what’s happening in pickleball overall.

And Todd, going back to your question—I think it’s fascinating to compare a 3.0 player from two or three years ago to a 3.0 today. I’d guess they’re not the same at all. Today’s 3.0 is almost certainly stronger, because the rating system is relative, and the sport itself has advanced so much.

If you watch a pro gold medal match from three years ago and compare it to now—it’s night and day. Yes, technology has evolved, but the level of play has completely transformed. It’s not the same game anymore, and the change has been across the board.

Todd B.: Oh, completely agree. I mean, if you look at the finals of the US, Open back in 2015-2016, it’s like a 4.0 match right now.

Sarah C.: Exactly. I think that’s the key. One other point I’d add is why we don’t actually see the overall mean rating grow over time. A lot of people assume it should, because most players who stick with the game do tend to improve. For active players, their DUPR graph is usually an upward trajectory.

Of course, there are exceptions—some players get older, deal with injuries, or simply play less, and their ratings may decline. But in general, active players trend upward.

The reason the mean doesn’t rise is because we constantly have tons of new players entering the system every single day. Those new players start on the lower end, and they essentially fill in the bottom of the distribution as existing players improve and move up

Todd B.: So I have a corollary to the Ben Johns question, and it relates to the top-ranked woman in the game, Anna Leigh Waters. Is she capped as to how far she can grow, just by virtue of the fact that she’s number one now and never plays anyone who’s ranked anywhere close to her on the female side?

Sarah C.: So, again, this really goes back to the idea of putting everyone on the same scale. And it’s hard—it’s a very difficult problem, especially in areas where we don’t have a lot of data between groups.

What we do is teach the algorithm how to behave in certain situations, and then let it run. Once it’s been trained to handle specific cases, it can place players and groups appropriately on its own.

So in one sense, yes—but in another, no. For example, right now her rating is moving just like everyone else’s. If you look at her rating history over time, it’s gone up steadily, and she’s only become more and more dominant.

Would we love to have more matches between her and male players? Absolutely. And I think that’s something we should actively pursue. If anyone in the sport is going to make that happen, it should be DUPR.

Until then, though, we just rely on the training data we do have to keep the system balanced and fair.

Todd B.: In doubles, where she can she can and does play against men, I think she can continue to gain but there’s just no way that she’d ever have a DUPR-rated match against a top male, right?

Tito M.: Todd. We actually, we actually tried this year to make that happen. We tried to do the battle of the sexes, and we were actually gonna have and Ryan Sherry versus Anna Leigh Waters. If Ryan is reading this: I love you. You are the biggest character on tour and you’re great, but I think Anna Leigh was gonna kick your butt.

I actually am a big believer that she would be extremely competitive in any singles match against a top male; she’s so smart on the court, she’s so good, she’s such a great talent. I think she would make a lot of guys suffer on tour.

Todd B.: Hey money talks. You never know what might incentivize either Ryan or Anna Leigh to do that?

Todd B.: A couple more questions: from a data perspective what is something that you find interesting about the DUPR data as a whole? You mentioned some cool stats already, but what’s your favorite stat that you can throw out there from your observations of the eight million rows that you have at your at your disposal right now?

Sarah C.: Goodness, I might have to think on it and come back to it. That’s a good question. That’s like a nerd-to-nerd question right there.

[ After our interview, Sarah sent me some excellent stats about the data, that I’ve put at the bottom of the article ]

Todd B.: Who’s the most famous person in the DUPR database that has a rating?

Tito M.: Oh, we have a bunch of very famous people. We made a post on social media about this recently. We have some NFL stars, we have actors, athletes, basketball players, and a bunch of former tennis stars like Andre Agassi and John McEnroe. Ivan Lendl in Florida is obsessed with his DUPR.

Todd B.: Ok this has been great. Is there anything we have not covered?

Tito M.: I think we’ve addressed a lot of the questions, but one of the most important things I want people to know is that we’re approaching this in a very thoughtful, educated way.

Sarah is extremely smart—obviously a pickleball junkie who really understands the sport—and the people behind this project care deeply. We’re part of the industry ourselves. We’re not making decisions to upset anyone; in fact, we want people to like us—ideally, to love us. The choices we’re making are based on data, and we truly believe they’re what’s best for the sport moving forward.

As I’ve said before, my DMs are open. People can message or email me anytime. We care so much about this that we want the chance to explain our reasoning and show why this is the right approach.

To this day, I’ve had thousands of conversations, and honestly, there’s only been one person we couldn’t convince. Everyone else, once I walked them through the logic, the details, and the reasoning, said, “Okay, that makes sense.”

Todd B.: Thank you guys. I think this is great. Thank you for your time today.

[Post-interview, Sarah sent some great data points about the DUPR database; here they are. Some of these are duplicated from the conversation above, others are newly mentioned herein]

– The DUPR database has around 8 million games and is growing daily.

– 20% of the matches are non-traditional doubles matches, meaning a mixed-gender match not exactly following a traditional Mixed Doubles match.

– 75% of our matches have a gap of 10 or more years between the youngest and oldest players.

– In best two-out-of-three matches, the deciding third game is nearly a coin flip. Roughly half of these matches are won by the team that took Game 1, and the other half by the team that won Game 2. There’s a slight edge for Game 2 winners, but the margin is minimal.

– Nearly half of our new DUPR users each week are international.

– 85% of DUPR matches are played between teams with a rating difference of less than 0.5.