Alter Everything Podcast

MaddieJ · ‎08-30-2021

What's the best way to practice predictive analytics? Build a model to predict something you're passionate about! Alteryx teammates Will Davis and Luke Minors, share how they used Alteryx Assisted Modeling to build a predictive model for the 2020 Euros.

Panelists

Will Davis - @wdavis, LinkedIn
Luke Minors - @LukeM, LinkedIn, Twitter
Maddie Johannsen - @MaddieJ, LinkedIn, Twitter

Topics

Euro 2020: Predictions and Review webinars
Creating your own xG Model
Alteryx Assisted Modeling
- Blog: Is Assisted Modeling Right for Me?
- Alteryx Intelligence Suite free trial
Analytics in Sport - Predicting the Euros
Alter Everything podcast episode 34: Efficiency vs. athleticism in basketball with Stan Van Gundy
Book recommendation: Football Hackers: The Science and Art of a Data Revolution

Analytics in sport LI.png

Transcript

Episode Transcription

MADDIE: 00:00	Welcome to Alter Everything, a podcast about data science and analytics culture. I'm Maddie Johannsen, and today, we're going to [music] talk about sports. But as a quick disclaimer, I'm not the most avid sports fan. I mean, I know quite a bit. I usually know the rules of most sports, can recognize and tell you why certain plays are a big deal, but I don't always know who the best team is, who the top players are, or any of the drama or politics that goes on behind the scenes. But I do love the magic of sports. To cheer on a team, see an uplifting story, or watch as the underdog prevails, there's a certain something to it that I think we all tend to gravitate towards. But what about analytics in sports? From actual sports analysts to dedicated fans, data-minded folks everywhere are using analytics to unpack performance of their favorite teams and to predict outcomes. So in this episode, I sat down with two of my coworkers from the UK, Will Davis and Luke Minors, to talk about the 2020 Euros, the football tournament that took place in the summer of 2021. Luke and Will are big football fans, enough so that they put together a two-part webinar series about the Euros where they built a predictive model attempting to figure out who would win the tournament. We'll unpack what went into the predictive model, how they collaborated and tweaked the model for optimized performance, and we'll get into a discussion about how sports data and analytics can really add to the magic of the sport. Let's get started. [music]
WILL: 01:29	Thank for having us, Maddie. Hi, everyone. I'm Will Davis, and I've been pretty upset for the last four days. I thought I'd start this off on a nice, cheery topic as we decided well in advance to film this or record this four days after England losing in the final. [laughter] But onto, I guess, happier grounds. So I work as a sales engineer for Alteryx. Essentially, my role is to help people help businesses address their challenges and problems. In my day-to-day, though, I love all sports, anything that I can get involved with. This does mainly involve football as I'm sure we're going to come on to talk a lot more about. But it's been great for me over the last couple of weeks of having the opportunity to combine my two passions, which are sport and football alongside data and using Alteryx to bring these together and to do some predictions with my friend, Luke.
LUKE: 02:24	Yeah. Well, I guess [laughter] that leaves me to introduce myself. I would actually, yeah, tell you my job title, but I'm actually not quite sure what it is at the moment. [laughter] I also work for Alteryx. [laughter] But my job title has changed three or four times in the last six months. I think it's currently solutions consultant. But essentially, I help customers, both internally and externally, solve their problems with Alteryx, or more strategically, around conversations about digital transformation and things like that. Will and I have been working on this analytics and sport sort of series that we've been doing webinars and things about, and it's been really good. And with that, my love for sport comes from slightly different angle. I played rugby since the age of five. Obviously, that's not the only thing I know about. I'm very passionate about most sports as Will said, but combining those interests and that sort of background with what I learn in my day-to-day, speaking to customers and things, has been really, really interesting.
MADDIE: 03:27	Yeah. And during the webinar, Will, did you say that you played cricket or that was kind of your favorite sport to play?
WILL: 03:35	Yeah. So growing up, it was all cricket. So I'm from a tiny, little village in England where there's about 200 people live, and then cricket is pretty much the only thing that anybody does. So [laughter] it was less of a decision to go into it, but more a rite of passage. So I'm sure, for the American listeners, it's effectively baseball but takes a bit longer. [laughter]
MADDIE: 03:56	Don't some of the games go for days?
WILL: 03:59	Five days, yeah. Five days of pure entertainment, [laughter] I could add.
MADDIE: 04:04	It sounds fun. Well then, actually, yeah, I was going to say, we don't really have rugby here, but I think that might just be my perception. Because I went for-- I live next to a lake, and I went for a run around the lake the other day. And I saw people playing rugby, and I was like, "Oh, that's interesting. I never really see this." But maybe it's like a-- maybe it's a thing here, and I just never noticed, so.
LUKE: 04:23	I think it is. I think the rugby-playing population of the US is growing significantly or has done over the last couple of years.
MADDIE: 04:30	Good to know.
LUKE: 04:31	It's an interesting one because there's always a comparison of rugby and American football. It's purely probably on the shape of the ball and not very much else [laughter] because the games, they're similar. I think that maybe the destruction of the human body element of it as well comes in.
MADDIE: 04:50	Yeah. It seems really brutal from what I've watched. But no, that's cool. Yeah. And I also remember last year, or I guess, two years ago when England was in the rugby finals of some tournament, I think, a big deal.
WILL: 05:03	Yes, mentioning more finals [that?] England [crosstalk]. [laughter]
LUKE: 05:04	The World Cup, yes. [laughter]
MADDIE: 05:06	The World Cup, yeah, some big tournament. [laughter]
LUKE: 05:10	Are we 10 minutes in, and [laughter] we've already, yeah, both our recent losses in finals of major tournaments. I'm so sick.
MADDIE: 05:18	Well, I remember I went to-- I was in England at the time. I was in London for Inspire. And I went to a pub. I woke up early and went to a pub because the game was at 8 AM or something. And that was such a fun experience, too. Everybody was going crazy, but.
LUKE: 05:32	Well, yes. That's the other classy British thing of going to the pub at 8 AM, or people were actually queuing at 6 AM to get into a pub for 8 AM and everyone was--
MADDIE: 05:40	Yeah. I felt really classy. [laughter]
LUKE: 05:43	--yeah. [laughter] Yeah. It led to a pretty interesting day afterwards, especially because everyone's mood was absolutely through the floor. [laughter]
MADDIE: 05:51	Yeah. Well, this is a good segue because I'd like to focus our conversation talking about the use of analytics in sports as you guys mentioned. And you recently completed a two-part webinar series talking about this and used the 2020 Euros as your examples. And to prepare for the webinar series, you guys did a lot of research. And I believe you spoke with sports analysts, correct?
WILL: 06:15	Yep.
MADDIE: 06:16	And can you share what you learned about their use of analytics and how that compares to how Alteryx users are using their data?
WILL: 06:24	Yeah. It was really interesting to sort of see into the minds of everything that's happening behind the scenes. I think for people like Luke and I who are really interested in sport, and your main focus is just sort of on game day of watching what happens on the field. You sort of don't take into account anything that's happening behind the scenes. So to get an insight into the work they're doing, the data they have access to, and who they speak to and who they interact with was really interesting. I think one of the key things we sort of found and discovered is that, actually, the approach that they all take is incredibly similar to all of the customers that we speak to, day-to-day. Fundamentally, these people are all analysts. They've all got a challenge or problem that's presented to them whether it be by a manager who you see on the TV or just somebody kind of in your office, they're presented with it. They've got to go and crunch the data and provide some insight back in an actionable manner. And this ranged across kind of pretty much all the topics that-- ultimately, the way that they say they address these problems and approach them was incredibly similar.
LUKE: 07:36	And actually, the thing that was interesting was-- our main aim for these webinars in the series was to take the passion that people have for sport and the interest and the knowledge that they own and try and leverage that in a really tangible way for them to learn something that's useful for their job. And one of the quotes I really liked from a guy we spoke to was-- and this is completely unprompted, but it fell perfectly in line with the angle that we were taking was, he said that all business intelligence problems are the same. And your three elements are, you need to understand the data you're working with, you need to know what question to ask to try and solve that problem, and then you need to be able to communicate the result or your findings to the stakeholders or whoever it is that you're working for, which basically rang true, and then we took that and leveraged that through our-- that's the key theme for looking at sports analytics and comparing that then to the business and how that affects everyone.
MADDIE: 08:39	Yeah. That reminds me of the weekly challenges in community. And I think you shouted this out on the webinar, too. But just to kind of emphasize it again, there's just so many different topics. But regardless of the theme or the topic of the challenge, the practice can be very much the same. So yeah. Let's talk about how you can use Alteryx to make predictions in sports. And one of the things that you started out with in the webinar was an xG model. Tell me more about this.
LUKE: 09:11	Yeah. So if you haven't heard of it-- had you heard of xG before, Maddie?
MADDIE: 09:17	I hadn't. No. Yeah. I'd love to hear more. Because I got some from the webinar, but I'm sure people out there don't know what that is either.
LUKE: 09:25	Yeah. So xG is a funny one because it is a term that appeared in the media around football probably a couple years ago now but was never really explained to the general public. And it's taken a while for people to kind of understand it. But it became a really good talking point for Will and I on our webinar. xG essentially stands for expected goals. And it's a metric, which essentially determines the likelihood that a particular player would score at a particular point in a football match based on whatever metrics you might want to include in there. So if you think about, on a really basic level, the most simple version of xG, is where they are on the pitch. So how far away from the goal are they when they kick the ball, and what angle are they at? So how much of the goal can they actually see? And that then gives a percentage chance likelihood of that goal, of the ball going in the net on that shot. And then as a model gets more and more complicated for xG - and this is the same with pretty much any model - you can add loads more factors in. So are they using their best foot? So are they using their left foot or their right foot, and which one are they better at shooting with? Are there defenders in the way that they have to avoid kicking the ball at? Are there-- I can't even think of another metric at this point. [laughter] But yeah. You can add all these different things in. And this was like a starting point for us when we tried to talk about doing predictive models because it's easily translatable into a business sense. Because if you were looking at a customer, and they-- if you were doing some analysis around your customers and looking at whether they might buy a product from you or whether they might sign up to a particular promotion or something, there are metrics around that that would persuade them to do these things. And this could be the demographic information or whether they have historical purchases of similar products or things like that that you can then include to help you determine what might happen in the future.
WILL: 11:39	Yeah. I think it's a really good example as well, as Luke described, that when it was first presented, it was sort of-- people just talked expected goals, but they didn't give that explanation and that background. And I think that's why people didn't really take it on initially. There was a lot of kind of your traditional football heads who would go, "Well, that wasn't the actual amount of goals there were. So it doesn't matter." And I think the more it's been explained, the more people are kind of getting on board with it and understanding more about what's happening kind of behind the scenes and getting involved in the analytics. And I think that's what's important kind of business context as well is, you can't just present a number and expect people to go, "Oh, yeah. Well, let's use that. We can now do it." You do need to have that interpretability. You need to be able to explain it to somebody that might not be interested in football or the challenge or the dates that you're using so that they can go, "All right. Not only can I use this, but I can understand how you've got to that number."
LUKE: 12:41	I think there was more reference to this. And I think it adds to the drama around football because you can start comparing it to what the expected goals were and how kind of interesting it is, the disparity between the actual result and the expected goals metric that was calculated. And the application of that expected goal has actually not just changed the way that the media portrays or talks about the sport or the layman in the pub that watches the game, how they think about the match, but some of the big stories about how data has impacted the sport itself around leveraging the data to look at these sorts of things because there's some really famous cases, Brentford being one of them. They're a sort of West London football club who've recently achieved promotion to the Premier League, which is the top tier of football in England. One of their key things was they stopped caring about the actual result of games - it wasn't like, "Oh, we won that game, lost that game - that they started using different KPIs, so the expected goals metric. Because if someone kicks a ball slightly to the left or the right, and it goes in the net and doesn't-- at the end of the day, you can't actually judge the whole team's performance on that one shot, whether it went in, and they won, or it didn't, and they drew the match. It's more about if they made enough chances to have a good xG, then that was actually a good performance. And if they continue doing that, then they're going to be a better team and improve results in that sense.
WILL: 14:17	I should point out, before they became famous for that, I think their most notable factor was that they had a pub on every corner of their stadium. [laughter] So I think that the view of Brentford has changed over the last 10 or so years. [laughter]
MADDIE: 14:32	I mean, that's a really good point because I always feel so bad for the goalie if they miss in any sport really, so hockey, soccer, football, whatever. If they miss one goal, and then they lose the game, and then I feel like there's just so much weight on the goalie or the keeper, whoever. But I feel like it shouldn't all come down to that one tiny thing. So yeah. Keep bringing in all those other factors. It definitely isn't--
LUKE: 15:00	I've never thought about that, actually.
MADDIE: 15:02	Really?
LUKE: 15:02	Well, as in, I imagine, now, that a goalkeeper would look at the expected goals metric of that particular shot and be like, "Oh, well, data says that there was only a 1% chance I could ever stop that from going in." And then, therefore, it's not his fault. [crosstalk]--
MADDIE: 15:19	Yeah. Maybe then to a layman sports viewer, maybe then that's just a perception that I have of like, "Oh, I hope they don't feel bad." So [laughter] maybe it's good that they're more data-minded in that way, and it takes some of the pressure off for them.
LUKE: 15:35	Yeah. So data will be making their days better, improving morale of the goalkeeper. [laughter]
MADDIE: 15:41	Well, and this is kind of a quick tangent. This whole conversation reminds me of an interview that we did with Stan Van Gundy a couple years ago now at that point. But Stan Van Gundy was a NBA coach. And one of the questions that we asked him in interview was if analytics and data was taking away from the style of the game. Was it making it rigid, or was it taking away from, I guess - what am I trying to say? - the fluid, fun, charismatic nature of the game? You know what I mean? But I mean, to your point, Luke, I think you were saying that it really enhances how people are watching football, and it's adding to the magic, I guess, in a way.
LUKE: 16:28	Potentially, I think it's very divisive, isn't it?
WILL: 16:32	Yeah.
LUKE: 16:32	Because you've got us three, and I imagine the majority of the listenership and the Alteryx customers who are relatively data-savvy find this sort of data interesting and can understand how it might affect certain things, the same with technology. However, there are people who have a completely different background and opinion around technology in sport and how data is affecting different things. Would you agree to that, Will?
WILL: 17:04	Yeah. It's always really interesting. And I think you pick a lot up when you're-- A, Twitter is a great place for it, and also, B, with just when you're in the pub watching and listening to people of those who are interested in the data aspect and those that just kind of want back to the traditional ways. And I think it's relevant to all things and all change. You'll always get people that are resistant to it and like things to be the same, what they're used to, and they don't like all these new numbers people. And you often hear it with some of the pundits as well of, they want people that are in football - they've been there; they've experienced it; they know it - as opposed to kind of these new data people. And I think that was an interesting part, again, of some of the conversations we had of how the analysts interact with those people that potentially don't have that, almost, future mindset, that are kind of stuck in their ways of how they had to present back an idea to them that was data-driven as opposed to, "I've played football for 10 years professionally. I know better than somebody who's looked at numbers for 10 years and has never kicked a ball."
MADDIE: 18:12	Yeah. Two sides to that argument for sure. It is interesting just to kind of look at the data in that way and to watch sports with that in mind. And kind of going back to the features that you guys were talking about with the model that you demonstrated in the webinar, and this may be a basic question, but how do you know what features to add and when it's maybe more of an asset to keep it simple and only have a couple of features, so only including historical data and data about the opposing team versus expected weather outcomes or player health or anything like that?
WILL: 18:52	I guess the simple answer is you can never be 100% sure without trying. And I think that's the approach we sort of took. So a lot of the work-- I'm going to give Luke his kudos here as the main feature man. I would say I was the sounding board for ideas and [laughter] providing that along. But for a lot of it, and our approach really was to start off to keep it simple with some of the features we fed in. So that was historical results going into that. And was it the world rankings that we brought in as well, Luke?
LUKE: 19:23	Yep, the feature rankings.
WILL: 19:25	Yeah. So in our first sort of iteration of keep it simple with the idea as well of introducing people to the concept and not wanting to overcomplicate it at that stage with lots of features, but as it progressed kind of through the tournament and as part of the webinar, we did want to add more features in. Now, as you said, when you're starting out with your data, some people might come from that position of, "We've only got a couple of features that could possibly go in." Some will have hundreds. I think that's where we were able to use Alteryx to help us out rather than kind of almost guessing in those first instances of what should go in. You can actually use the assisted modeling to do a lot of that for you. So that's going to pull out your feature importance as you feed all of these in. And then you can combine that with, I'm going to say our footballing expertise or where we'd like to consider ourselves as experts of almost, "These are the factors we know will have an impact," and then statistically, "Here's what assisted modeling is presenting back as those other ones that we should be including."
LUKE: 20:28	I think this is where the use of Alteryx in whatever capacity you can allows you to learn these things as you go through the process. Neither Will nor I are data scientists by any stretch of the imagination, but learning through doing and some of the community, we've picked enough to be able to start building these things out and then teaching them to customers as well. One of the elements that we were taking with how we present it in the webinar was around-- the features sound really complex because we're talking about data points in an example, but if you were to have a conversation about the upcoming sports match that you were going to watch, you sat down in the pub with a beer in your hand, you would speculate over what the result might be based on these features. So, oh, we think it's going to be where? And therefore, this team is going to play better. Or that this quarterback is injured, and therefore, they're less likely to have as many passing yards or whatever it might be. Those elements are essentially features, and when you boil them down and put them into a predictive model, you can then understand how they have more of an effect on what you're trying to predict. And it was really interesting because we used the Q&A and the chat function in the webinar to start asking people what they thought might be effective as predicters or things that might determine whether someone would win a football match or not. And when we eventually put all that stuff into the assisted modeling tool and looked at the reporting outputs to understand which are the most important features, it actually turned out that the most obvious, the simplest ones, were the most effective.
MADDIE: 22:16	I think that explanation is super helpful. Context is always really important for understanding. And yeah, I loved seeing the assisted modeling incorporated into your webinar because I think that was really awesome just to see how intuitive and easy to use it was. And I'm curious. With you guys working with Alteryx users every day and working in Alteryx yourselves, what has been the reaction that you've seen from users when they see it? Are they ready to jump in, even if they're not a data scientist, as you guys were saying that you're not data scientists either, and you guys were able to use it super easily as well? So yeah. Curious what you've seen out there every day.
WILL: 22:55	Yeah. So the best part for me of seeing people use it is the excitement. And I think a lot of that maybe comes back to the whole aspect of the data science problem or advanced analytics you need to be a data scientist. And when presenting assisted modeling, it has often been the case that you can see something switch in a customer's mind from, not necessarily fear, but almost a questioning, "Can I do this?" to, "I really want to get involved. I want to start using it, and I want to start kind of building my own models." And I think the nature of it-- and as Luke said, I don't have a data science background. So even for myself, picking it up, and I've used kind of standard predictive tools, is that you get the explanations along the way. And I think this is key for those people is that-- a lot of the people I work with that they're interested to learn. They don't just want to kind of build something and leave it aside. So having those elements that they can learn, they can create a model, but also present it back and come away thinking, "Not only have I done that, but I've bettered myself. I can now present this back. I can speak more fluently," and almost that move, and we often talk about it, to the citizen data scientist. But it often starts that they'll come to us with one idea of what they might want to predict. You can show them the assisted modeling, and you speak to them next week, and they've built something, and they've had about 10 new ideas. Because they've sort of changed their mindset from, "Will I be able to do this? What do I need to learn?" to actually just kind of freeing them up to go, "Here's all the things I'd like to do. I now just need to get kind of my data ready, and I can start building and learning." [laughter] And that's the best part. Because we said it on the webinar as well is, "You'll never get a model right the first time, and you're not going to get 100% accuracy." So if you've got the mindset to kind of test and iterate, that's what we're trying to get them to do, and help them on them on their way.
LUKE: 24:56	Those are one of the funniest challenges to overcome when you're working with a new user is when they get super excited and suddenly have eight problems to solve, [laughter] and you're like, "Hold on. Focus here. [laughter] Let's just make sure you've got one solution," [laughter] rather than--
MADDIE: 25:16	Yeah, casting a wide net. [laughter]
LUKE: 25:18	--eight half-baked ideas. [laughter]
MADDIE: 25:21	Yeah. And that leads into another question that I wanted to ask. I'd love to hear more about how you collaborated with each other and tweaked the model and maybe collaborated with members of the community or people who joined the webinar. Because I saw that there were a few polls in the webinar, and you're kind of asking people, "What features should we include?" and, "How should we tweak this?" I would love to just hear any tips that you have for folks to collaborate in order to make their models stronger. [laughter]
LUKE: 25:50	Collaborations start as we both go into silence. [laughter]
MADDIE: 25:55	Good [time, right?] guys.
LUKE: 25:56	I am a big-- I am a big advocate. Yeah. So I mean, there are a load of different elements to that. One of the key things that we wanted to do was make it an open forum for the audience to speak back, which is why we-- I mean, to tell the truth, we put a couple of polls in, but technically, that was a bit more complicated than we should have gone down. [laughter] That was a learning point because the Q&A box was just flying through with messages. And obviously, as we're reading them, we can answer out questions and things. But the polls we put together made it a bit more complicated than necessary. We were just trying to stretch the webinar platform as much as we could. But the Q&A and having that reference of-- I think there were 250 people on the webinar. And people were really actually responding back to what we were saying and having both positive and some more constructive feedback. [laughter] That's not to mention some of Will's friends that came along and did everything that they possibly could to distract us. That element was really good. And like I said, from my perspective, because I'd put most of the work into building out the workflow itself in the predictive model. Will did other stuff, I'm told. [laughter] And then being able to discuss with him ideas and then take ideas from the community and the audience and then impart those into the model as well made it a lot more of a fun journey for us.
WILL: 27:26	I think the benefit for us in this example is that we're friends, and we both like football. And so it sort of feels a bit more natural in sort of coming together, but. And I think, not just sort of on the model, but if you take the collaboration as a whole of trying to create the webinars is like working together to kind of get the framework, and ultimately, the message that we wanted to deliver to the people attending, but then making sure we had time to kind of think individually to bring different ideas to the table. As Luke said, I did try and do something as a part of this collaboration. [laughter] So hopefully, I did something there, but. But yeah. What we would often do is to take time apart to have that individual thinking, that individual time to create your own ideas once you'd got a framework of what you want to get to. And then those sessions, as Luke mentioned, where you're together is bringing it all together, understanding which bits we think are going to be most impactful, which bits we take from what Luke had produced, which bits we'd take or not take from what I had produced to come together. And hopefully, that collaboration just showed off nicely in the webinars itself.
MADDIE: 28:34	Yeah. And I'm curious what you guys learned from the process, overall. And I think one of the things that stood out to me was inviting friends to the webinar to distract you. I think that that's always a good thing to do. But other than that, what did you guys learn from the process?
WILL: 28:52	I'd say it's the converse. Never mention it to your friends. [laughter] I thought it'd be a rare opportunity; they might be interested in what I do when it's football related. But I think they all got a bit excited. [laughter] So they won't be invited to any future ones. But generally, in terms of it, and like Luke mentioned at the start, first, I've always been interested in sports and interested in data, but I've never really gone kind of behind the scenes and taken the time to understand more. And ultimately, we wanted to present something back that people could learn from. And so I wanted to make sure that we were knowledgeable, and we were skilled in all of this. So it's actually been a great opportunity to expand our learnings. And I think, in having that and sort of, I guess, putting you outside your comfort zone of, "You need to go and produce something," and set a challenge, it then kind of inspires you/forces you to go and do that reading. And I personally found it incredibly interesting. It meant Luke and I, in our discussions, actually sounded a little bit more intelligent, or I'd like to think anyway, [laughter] maybe with the conversations behind the scenes. But also, there's a part that it's great to be able to talk to so many people in that industry and just to get different people's perspectives. As we've already mentioned, their approach is similar, but the data that they're working with and how they operate is really interesting. So it's just been a great experience, I'd say, of learning more and meeting new people and understanding the ways that they work.
LUKE: 30:24	I was reading the other day about [Forster?] Rugby Club in England, one of the premiership rugby clubs. They spoke to analysts and people from all areas of world sport when they were looking at bringing analytics more into what they were doing. And they brought in people from the NFL, people from the Team GB, cycling, things like that to try and understand different learnings. There are so many interesting books around sports or data. And it's not just money ball. There's a load of different bits of literature that you can read to learn from.
MADDIE: 31:03	Yeah. In one of your blog posts that you wrote, Luke, I know that you listed some resources for some other learnings. And I think it'd be fun if you guys wanted to share more of your favorite books or movies or podcasts or anything that's related to this, and we can link them in the show notes for people. I think they would like that.
WILL: 31:25	I think, actually, the one that got us sort of started on it, and this was a few months back - and Luke, did you recommend it? - but the Football Hackers book is, sort of, I would suggest, a really good gateway sports book in the-- again, because that does reference a lot of material, and internally, sort of in our team at Alteryx, quite a few of us read it, and that's really what--
LUKE: 31:48	That's a part of our book club, wasn't it?
WILL: 31:50	Yeah, a part of the book club, always learning. But I would say that that was a great starting point, and certainly one that I found incredibly useful. And I think when it's topics like xG, that was kind of the introductory explanations that I was getting and the first time I sort of saw behind the scenes of it.
MADDIE: 32:10	Yeah. I like the back stories of some of these models and methodologies that people are using. It's really interesting. Cool.
LUKE: 32:18	Yeah. I mean, and there's some really interesting characters, especially in football that got involved and did all this stuff. And the storyline is actually really interesting because it's written by a German guy who got involved, and he suddenly got into a really high position in one of the German football teams because of his knowledge of data. And you then start realizing that all of the famous names of managers that everyone just knows now, their history was not necessarily playing football, and they had that experience, but they're starting to be educated in the world of data and understand how to leverage that to better the team. And then there's loads of stories about how data is used to identify new players and work out who gets promoted into the team from the use team or the B team or whoever it is. And so yeah, there's some really interesting stories. I very much recommend that book. That's Football Hackers: The Science and Art of a Data Revolution.
MADDIE: 33:27	Yeah. Thank you. We'll be sure to link to that. So what can we expect next from you guys?
WILL: 33:34	Good question. [laughter] I mean, for us, and a lot of the reason for the kind of sport and analytics in football is a great start to it is that, ultimately, we've kind of created a framework now to help other people and probably continue, ourselves, to apply this to other sports and other competitions. There's pretty much a-- there's major international tournaments every year across different sports, across domestic leagues. There's stuff happening all the time. So I think, personally, from us, we'll probably get more involved into some of those. It's just a good opportunity to hopefully kind of apply more across different sports areas. And hopefully, what we can do-- and we've got Alteryx as a really good platform to bring it to life for a lot of people. I mean, ultimately, what we wanted from the webinar series was to get people more passionate. For those that either are interested in data and not in sport, they might go one way. For those that are really interested in sport but have never been interested in the data that, hopefully, they might see that transition, get more invested. So I mean, ultimately, that's mine. If I can continue working and watching sport, that's the dream. So if anybody from Reading Football Club is listening, [laughter] I'm all ears for any job opportunities there [laughter] as well.
MADDIE: 34:53	Nice plug. Okay. [laughter] Awesome.
LUKE: 34:57	Yeah. I mean, as Will said, there's plenty of projects and things to be going on with. If anyone's got any ideas around how we can sort of merge the world of data analytics that we work in day-to-day, whether it's in finance or marketing or whatever it might be, and passions that you have outside, not just sporting, but where we can get ahold of data and mess around with it, I think that that'd be a really good thing to do. [music]
MADDIE: 35:27	Well, thank you both so much for joining me. This is so fun.
LUKE: 35:32	It's great to be invited. Yeah. [laughter]
WILL: 35:32	Thank you for having us.
MADDIE: 35:37	Thanks for listening. To explore assisted modeling with Alteryx and to hear more from Will and Luke, including their 2020 Euros webinar series, check out our show notes at community.alteryx.com/podcast. Catch you next time. [music] Luke are you having a pint now?
LUKE: 36:02	This is not a pint. This is not a quiet pint. It's just a can.
MADDIE: 36:06	Okay. Just, do you call it a can?
WILL: 36:08	For anybody listening, it's Monday, 9 AM in London at the moment. [laughter]
LUKE: 36:12	That is not true. I just realized--
MADDIE: 36:14	And you're just cracking a beer.
LUKE: 36:14	--it's getting to the end of the day. [laughter] I've noticed that we're getting towards the end of recording. I had been looking forward to this Thursday afternoon beer, recording this podcast, and realized that I hadn't gone and got one. So yeah. That's my lunch.
MADDIE: 36:27	Well, to be fair, yeah, it was my idea initially when we had prep calls, I was like, "We should drink beers during this call." And I personally dropped the ball. So I didn't expect you guys to do it because it is literally 9 AM in Denver for me. So I was going to make a mimosa, and then I decided I didn't want to be hiccupping throughout this whole recording, [laughter] so. But I'm proud of you for having a beer like we originally planned.
LUKE: 36:54	One of the revelations of the last few months was getting a wireless headset because it's opened up the entire house to me [laughter] whilst I'm working. So yeah. Just able to go to the fridge, get myself a nice little IPA, crack that open.
MADDIE: 37:10	That's great. I'm glad that we could hear it, too, on the audio. I appreciate that. [laughter]

MADDIE: 00:00 Welcome to Alter Everything, a podcast about data science and analytics culture. I'm Maddie Johannsen, and today, we're going to [music] talk about sports. But as a quick disclaimer, I'm not the most avid sports fan. I mean, I know quite a bit. I usually know the rules of most sports, can recognize and tell you why certain plays are a big deal, but I don't always know who the best team is, who the top players are, or any of the drama or politics that goes on behind the scenes. But I do love the magic of sports. To cheer on a team, see an uplifting story, or watch as the underdog prevails, there's a certain something to it that I think we all tend to gravitate towards. But what about analytics in sports? From actual sports analysts to dedicated fans, data-minded folks everywhere are using analytics to unpack performance of their favorite teams and to predict outcomes. So in this episode, I sat down with two of my coworkers from the UK, Will Davis and Luke Minors, to talk about the 2020 Euros, the football tournament that took place in the summer of 2021. Luke and Will are big football fans, enough so that they put together a two-part webinar series about the Euros where they built a predictive model attempting to figure out who would win the tournament. We'll unpack what went into the predictive model, how they collaborated and tweaked the model for optimized performance, and we'll get into a discussion about how sports data and analytics can really add to the magic of the sport. Let's get started. [music] WILL: 01:29 Thank for having us, Maddie. Hi, everyone. I'm Will Davis, and I've been pretty upset for the last four days. I thought I'd start this off on a nice, cheery topic as we decided well in advance to film this or record this four days after England losing in the final. [laughter] But onto, I guess, happier grounds. So I work as a sales engineer for Alteryx. Essentially, my role is to help people help businesses address their challenges and problems. In my day-to-day, though, I love all sports, anything that I can get involved with. This does mainly involve football as I'm sure we're going to come on to talk a lot more about. But it's been great for me over the last couple of weeks of having the opportunity to combine my two passions, which are sport and football alongside data and using Alteryx to bring these together and to do some predictions with my friend, Luke. LUKE: 02:24 Yeah. Well, I guess [laughter] that leaves me to introduce myself. I would actually, yeah, tell you my job title, but I'm actually not quite sure what it is at the moment. [laughter] I also work for Alteryx. [laughter] But my job title has changed three or four times in the last six months. I think it's currently solutions consultant. But essentially, I help customers, both internally and externally, solve their problems with Alteryx, or more strategically, around conversations about digital transformation and things like that. Will and I have been working on this analytics and sport sort of series that we've been doing webinars and things about, and it's been really good. And with that, my love for sport comes from slightly different angle. I played rugby since the age of five. Obviously, that's not the only thing I know about. I'm very passionate about most sports as Will said, but combining those interests and that sort of background with what I learn in my day-to-day, speaking to customers and things, has been really, really interesting. MADDIE: 03:27 Yeah. And during the webinar, Will, did you say that you played cricket or that was kind of your favorite sport to play? WILL: 03:35 Yeah. So growing up, it was all cricket. So I'm from a tiny, little village in England where there's about 200 people live, and then cricket is pretty much the only thing that anybody does. So [laughter] it was less of a decision to go into it, but more a rite of passage. So I'm sure, for the American listeners, it's effectively baseball but takes a bit longer. [laughter] MADDIE: 03:56 Don't some of the games go for days? WILL: 03:59 Five days, yeah. Five days of pure entertainment, [laughter] I could add. MADDIE: 04:04 It sounds fun. Well then, actually, yeah, I was going to say, we don't really have rugby here, but I think that might just be my perception. Because I went for-- I live next to a lake, and I went for a run around the lake the other day. And I saw people playing rugby, and I was like, "Oh, that's interesting. I never really see this." But maybe it's like a-- maybe it's a thing here, and I just never noticed, so. LUKE: 04:23 I think it is. I think the rugby-playing population of the US is growing significantly or has done over the last couple of years. MADDIE: 04:30 Good to know. LUKE: 04:31 It's an interesting one because there's always a comparison of rugby and American football. It's purely probably on the shape of the ball and not very much else [laughter] because the games, they're similar. I think that maybe the destruction of the human body element of it as well comes in. MADDIE: 04:50 Yeah. It seems really brutal from what I've watched. But no, that's cool. Yeah. And I also remember last year, or I guess, two years ago when England was in the rugby finals of some tournament, I think, a big deal. WILL: 05:03 Yes, mentioning more finals [that?] England [crosstalk]. [laughter] LUKE: 05:04 The World Cup, yes. [laughter] MADDIE: 05:06 The World Cup, yeah, some big tournament. [laughter] LUKE: 05:10 Are we 10 minutes in, and [laughter] we've already, yeah, both our recent losses in finals of major tournaments. I'm so sick. MADDIE: 05:18 Well, I remember I went to-- I was in England at the time. I was in London for Inspire. And I went to a pub. I woke up early and went to a pub because the game was at 8 AM or something. And that was such a fun experience, too. Everybody was going crazy, but. LUKE: 05:32 Well, yes. That's the other classy British thing of going to the pub at 8 AM, or people were actually queuing at 6 AM to get into a pub for 8 AM and everyone was-- MADDIE: 05:40 Yeah. I felt really classy. [laughter] LUKE: 05:43 --yeah. [laughter] Yeah. It led to a pretty interesting day afterwards, especially because everyone's mood was absolutely through the floor. [laughter] MADDIE: 05:51 Yeah. Well, this is a good segue because I'd like to focus our conversation talking about the use of analytics in sports as you guys mentioned. And you recently completed a two-part webinar series talking about this and used the 2020 Euros as your examples. And to prepare for the webinar series, you guys did a lot of research. And I believe you spoke with sports analysts, correct? WILL: 06:15 Yep. MADDIE: 06:16 And can you share what you learned about their use of analytics and how that compares to how Alteryx users are using their data? WILL: 06:24 Yeah. It was really interesting to sort of see into the minds of everything that's happening behind the scenes. I think for people like Luke and I who are really interested in sport, and your main focus is just sort of on game day of watching what happens on the field. You sort of don't take into account anything that's happening behind the scenes. So to get an insight into the work they're doing, the data they have access to, and who they speak to and who they interact with was really interesting. I think one of the key things we sort of found and discovered is that, actually, the approach that they all take is incredibly similar to all of the customers that we speak to, day-to-day. Fundamentally, these people are all analysts. They've all got a challenge or problem that's presented to them whether it be by a manager who you see on the TV or just somebody kind of in your office, they're presented with it. They've got to go and crunch the data and provide some insight back in an actionable manner. And this ranged across kind of pretty much all the topics that-- ultimately, the way that they say they address these problems and approach them was incredibly similar. LUKE: 07:36 And actually, the thing that was interesting was-- our main aim for these webinars in the series was to take the passion that people have for sport and the interest and the knowledge that they own and try and leverage that in a really tangible way for them to learn something that's useful for their job. And one of the quotes I really liked from a guy we spoke to was-- and this is completely unprompted, but it fell perfectly in line with the angle that we were taking was, he said that all business intelligence problems are the same. And your three elements are, you need to understand the data you're working with, you need to know what question to ask to try and solve that problem, and then you need to be able to communicate the result or your findings to the stakeholders or whoever it is that you're working for, which basically rang true, and then we took that and leveraged that through our-- that's the key theme for looking at sports analytics and comparing that then to the business and how that affects everyone. MADDIE: 08:39 Yeah. That reminds me of the weekly challenges in community. And I think you shouted this out on the webinar, too. But just to kind of emphasize it again, there's just so many different topics. But regardless of the theme or the topic of the challenge, the practice can be very much the same. So yeah. Let's talk about how you can use Alteryx to make predictions in sports. And one of the things that you started out with in the webinar was an xG model. Tell me more about this. LUKE: 09:11 Yeah. So if you haven't heard of it-- had you heard of xG before, Maddie? MADDIE: 09:17 I hadn't. No. Yeah. I'd love to hear more. Because I got some from the webinar, but I'm sure people out there don't know what that is either. LUKE: 09:25 Yeah. So xG is a funny one because it is a term that appeared in the media around football probably a couple years ago now but was never really explained to the general public. And it's taken a while for people to kind of understand it. But it became a really good talking point for Will and I on our webinar. xG essentially stands for expected goals. And it's a metric, which essentially determines the likelihood that a particular player would score at a particular point in a football match based on whatever metrics you might want to include in there. So if you think about, on a really basic level, the most simple version of xG, is where they are on the pitch. So how far away from the goal are they when they kick the ball, and what angle are they at? So how much of the goal can they actually see? And that then gives a percentage chance likelihood of that goal, of the ball going in the net on that shot. And then as a model gets more and more complicated for xG - and this is the same with pretty much any model - you can add loads more factors in. So are they using their best foot? So are they using their left foot or their right foot, and which one are they better at shooting with? Are there defenders in the way that they have to avoid kicking the ball at? Are there-- I can't even think of another metric at this point. [laughter] But yeah. You can add all these different things in. And this was like a starting point for us when we tried to talk about doing predictive models because it's easily translatable into a business sense. Because if you were looking at a customer, and they-- if you were doing some analysis around your customers and looking at whether they might buy a product from you or whether they might sign up to a particular promotion or something, there are metrics around that that would persuade them to do these things. And this could be the demographic information or whether they have historical purchases of similar products or things like that that you can then include to help you determine what might happen in the future. WILL: 11:39 Yeah. I think it's a really good example as well, as Luke described, that when it was first presented, it was sort of-- people just talked expected goals, but they didn't give that explanation and that background. And I think that's why people didn't really take it on initially. There was a lot of kind of your traditional football heads who would go, "Well, that wasn't the actual amount of goals there were. So it doesn't matter." And I think the more it's been explained, the more people are kind of getting on board with it and understanding more about what's happening kind of behind the scenes and getting involved in the analytics. And I think that's what's important kind of business context as well is, you can't just present a number and expect people to go, "Oh, yeah. Well, let's use that. We can now do it." You do need to have that interpretability. You need to be able to explain it to somebody that might not be interested in football or the challenge or the dates that you're using so that they can go, "All right. Not only can I use this, but I can understand how you've got to that number." LUKE: 12:41 I think there was more reference to this. And I think it adds to the drama around football because you can start comparing it to what the expected goals were and how kind of interesting it is, the disparity between the actual result and the expected goals metric that was calculated. And the application of that expected goal has actually not just changed the way that the media portrays or talks about the sport or the layman in the pub that watches the game, how they think about the match, but some of the big stories about how data has impacted the sport itself around leveraging the data to look at these sorts of things because there's some really famous cases, Brentford being one of them. They're a sort of West London football club who've recently achieved promotion to the Premier League, which is the top tier of football in England. One of their key things was they stopped caring about the actual result of games - it wasn't like, "Oh, we won that game, lost that game - that they started using different KPIs, so the expected goals metric. Because if someone kicks a ball slightly to the left or the right, and it goes in the net and doesn't-- at the end of the day, you can't actually judge the whole team's performance on that one shot, whether it went in, and they won, or it didn't, and they drew the match. It's more about if they made enough chances to have a good xG, then that was actually a good performance. And if they continue doing that, then they're going to be a better team and improve results in that sense. WILL: 14:17 I should point out, before they became famous for that, I think their most notable factor was that they had a pub on every corner of their stadium. [laughter] So I think that the view of Brentford has changed over the last 10 or so years. [laughter] MADDIE: 14:32 I mean, that's a really good point because I always feel so bad for the goalie if they miss in any sport really, so hockey, soccer, football, whatever. If they miss one goal, and then they lose the game, and then I feel like there's just so much weight on the goalie or the keeper, whoever. But I feel like it shouldn't all come down to that one tiny thing. So yeah. Keep bringing in all those other factors. It definitely isn't-- LUKE: 15:00 I've never thought about that, actually. MADDIE: 15:02 Really? LUKE: 15:02 Well, as in, I imagine, now, that a goalkeeper would look at the expected goals metric of that particular shot and be like, "Oh, well, data says that there was only a 1% chance I could ever stop that from going in." And then, therefore, it's not his fault. [crosstalk]-- MADDIE: 15:19 Yeah. Maybe then to a layman sports viewer, maybe then that's just a perception that I have of like, "Oh, I hope they don't feel bad." So [laughter] maybe it's good that they're more data-minded in that way, and it takes some of the pressure off for them. LUKE: 15:35 Yeah. So data will be making their days better, improving morale of the goalkeeper. [laughter] MADDIE: 15:41 Well, and this is kind of a quick tangent. This whole conversation reminds me of an interview that we did with Stan Van Gundy a couple years ago now at that point. But Stan Van Gundy was a NBA coach. And one of the questions that we asked him in interview was if analytics and data was taking away from the style of the game. Was it making it rigid, or was it taking away from, I guess - what am I trying to say? - the fluid, fun, charismatic nature of the game? You know what I mean? But I mean, to your point, Luke, I think you were saying that it really enhances how people are watching football, and it's adding to the magic, I guess, in a way. LUKE: 16:28 Potentially, I think it's very divisive, isn't it? WILL: 16:32 Yeah. LUKE: 16:32 Because you've got us three, and I imagine the majority of the listenership and the Alteryx customers who are relatively data-savvy find this sort of data interesting and can understand how it might affect certain things, the same with technology. However, there are people who have a completely different background and opinion around technology in sport and how data is affecting different things. Would you agree to that, Will? WILL: 17:04 Yeah. It's always really interesting. And I think you pick a lot up when you're-- A, Twitter is a great place for it, and also, B, with just when you're in the pub watching and listening to people of those who are interested in the data aspect and those that just kind of want back to the traditional ways. And I think it's relevant to all things and all change. You'll always get people that are resistant to it and like things to be the same, what they're used to, and they don't like all these new numbers people. And you often hear it with some of the pundits as well of, they want people that are in football - they've been there; they've experienced it; they know it - as opposed to kind of these new data people. And I think that was an interesting part, again, of some of the conversations we had of how the analysts interact with those people that potentially don't have that, almost, future mindset, that are kind of stuck in their ways of how they had to present back an idea to them that was data-driven as opposed to, "I've played football for 10 years professionally. I know better than somebody who's looked at numbers for 10 years and has never kicked a ball." MADDIE: 18:12 Yeah. Two sides to that argument for sure. It is interesting just to kind of look at the data in that way and to watch sports with that in mind. And kind of going back to the features that you guys were talking about with the model that you demonstrated in the webinar, and this may be a basic question, but how do you know what features to add and when it's maybe more of an asset to keep it simple and only have a couple of features, so only including historical data and data about the opposing team versus expected weather outcomes or player health or anything like that? WILL: 18:52 I guess the simple answer is you can never be 100% sure without trying. And I think that's the approach we sort of took. So a lot of the work-- I'm going to give Luke his kudos here as the main feature man. I would say I was the sounding board for ideas and [laughter] providing that along. But for a lot of it, and our approach really was to start off to keep it simple with some of the features we fed in. So that was historical results going into that. And was it the world rankings that we brought in as well, Luke? LUKE: 19:23 Yep, the feature rankings. WILL: 19:25 Yeah. So in our first sort of iteration of keep it simple with the idea as well of introducing people to the concept and not wanting to overcomplicate it at that stage with lots of features, but as it progressed kind of through the tournament and as part of the webinar, we did want to add more features in. Now, as you said, when you're starting out with your data, some people might come from that position of, "We've only got a couple of features that could possibly go in." Some will have hundreds. I think that's where we were able to use Alteryx to help us out rather than kind of almost guessing in those first instances of what should go in. You can actually use the assisted modeling to do a lot of that for you. So that's going to pull out your feature importance as you feed all of these in. And then you can combine that with, I'm going to say our footballing expertise or where we'd like to consider ourselves as experts of almost, "These are the factors we know will have an impact," and then statistically, "Here's what assisted modeling is presenting back as those other ones that we should be including." LUKE: 20:28 I think this is where the use of Alteryx in whatever capacity you can allows you to learn these things as you go through the process. Neither Will nor I are data scientists by any stretch of the imagination, but learning through doing and some of the community, we've picked enough to be able to start building these things out and then teaching them to customers as well. One of the elements that we were taking with how we present it in the webinar was around-- the features sound really complex because we're talking about data points in an example, but if you were to have a conversation about the upcoming sports match that you were going to watch, you sat down in the pub with a beer in your hand, you would speculate over what the result might be based on these features. So, oh, we think it's going to be where? And therefore, this team is going to play better. Or that this quarterback is injured, and therefore, they're less likely to have as many passing yards or whatever it might be. Those elements are essentially features, and when you boil them down and put them into a predictive model, you can then understand how they have more of an effect on what you're trying to predict. And it was really interesting because we used the Q&A and the chat function in the webinar to start asking people what they thought might be effective as predicters or things that might determine whether someone would win a football match or not. And when we eventually put all that stuff into the assisted modeling tool and looked at the reporting outputs to understand which are the most important features, it actually turned out that the most obvious, the simplest ones, were the most effective. MADDIE: 22:16 I think that explanation is super helpful. Context is always really important for understanding. And yeah, I loved seeing the assisted modeling incorporated into your webinar because I think that was really awesome just to see how intuitive and easy to use it was. And I'm curious. With you guys working with Alteryx users every day and working in Alteryx yourselves, what has been the reaction that you've seen from users when they see it? Are they ready to jump in, even if they're not a data scientist, as you guys were saying that you're not data scientists either, and you guys were able to use it super easily as well? So yeah. Curious what you've seen out there every day. WILL: 22:55 Yeah. So the best part for me of seeing people use it is the excitement. And I think a lot of that maybe comes back to the whole aspect of the data science problem or advanced analytics you need to be a data scientist. And when presenting assisted modeling, it has often been the case that you can see something switch in a customer's mind from, not necessarily fear, but almost a questioning, "Can I do this?" to, "I really want to get involved. I want to start using it, and I want to start kind of building my own models." And I think the nature of it-- and as Luke said, I don't have a data science background. So even for myself, picking it up, and I've used kind of standard predictive tools, is that you get the explanations along the way. And I think this is key for those people is that-- a lot of the people I work with that they're interested to learn. They don't just want to kind of build something and leave it aside. So having those elements that they can learn, they can create a model, but also present it back and come away thinking, "Not only have I done that, but I've bettered myself. I can now present this back. I can speak more fluently," and almost that move, and we often talk about it, to the citizen data scientist. But it often starts that they'll come to us with one idea of what they might want to predict. You can show them the assisted modeling, and you speak to them next week, and they've built something, and they've had about 10 new ideas. Because they've sort of changed their mindset from, "Will I be able to do this? What do I need to learn?" to actually just kind of freeing them up to go, "Here's all the things I'd like to do. I now just need to get kind of my data ready, and I can start building and learning." [laughter] And that's the best part. Because we said it on the webinar as well is, "You'll never get a model right the first time, and you're not going to get 100% accuracy." So if you've got the mindset to kind of test and iterate, that's what we're trying to get them to do, and help them on them on their way. LUKE: 24:56 Those are one of the funniest challenges to overcome when you're working with a new user is when they get super excited and suddenly have eight problems to solve, [laughter] and you're like, "Hold on. Focus here. [laughter] Let's just make sure you've got one solution," [laughter] rather than-- MADDIE: 25:16 Yeah, casting a wide net. [laughter] LUKE: 25:18 --eight half-baked ideas. [laughter] MADDIE: 25:21 Yeah. And that leads into another question that I wanted to ask. I'd love to hear more about how you collaborated with each other and tweaked the model and maybe collaborated with members of the community or people who joined the webinar. Because I saw that there were a few polls in the webinar, and you're kind of asking people, "What features should we include?" and, "How should we tweak this?" I would love to just hear any tips that you have for folks to collaborate in order to make their models stronger. [laughter] LUKE: 25:50 Collaborations start as we both go into silence. [laughter] MADDIE: 25:55 Good [time, right?] guys. LUKE: 25:56 I am a big-- I am a big advocate. Yeah. So I mean, there are a load of different elements to that. One of the key things that we wanted to do was make it an open forum for the audience to speak back, which is why we-- I mean, to tell the truth, we put a couple of polls in, but technically, that was a bit more complicated than we should have gone down. [laughter] That was a learning point because the Q&A box was just flying through with messages. And obviously, as we're reading them, we can answer out questions and things. But the polls we put together made it a bit more complicated than necessary. We were just trying to stretch the webinar platform as much as we could. But the Q&A and having that reference of-- I think there were 250 people on the webinar. And people were really actually responding back to what we were saying and having both positive and some more constructive feedback. [laughter] That's not to mention some of Will's friends that came along and did everything that they possibly could to distract us. That element was really good. And like I said, from my perspective, because I'd put most of the work into building out the workflow itself in the predictive model. Will did other stuff, I'm told. [laughter] And then being able to discuss with him ideas and then take ideas from the community and the audience and then impart those into the model as well made it a lot more of a fun journey for us. WILL: 27:26 I think the benefit for us in this example is that we're friends, and we both like football. And so it sort of feels a bit more natural in sort of coming together, but. And I think, not just sort of on the model, but if you take the collaboration as a whole of trying to create the webinars is like working together to kind of get the framework, and ultimately, the message that we wanted to deliver to the people attending, but then making sure we had time to kind of think individually to bring different ideas to the table. As Luke said, I did try and do something as a part of this collaboration. [laughter] So hopefully, I did something there, but. But yeah. What we would often do is to take time apart to have that individual thinking, that individual time to create your own ideas once you'd got a framework of what you want to get to. And then those sessions, as Luke mentioned, where you're together is bringing it all together, understanding which bits we think are going to be most impactful, which bits we take from what Luke had produced, which bits we'd take or not take from what I had produced to come together. And hopefully, that collaboration just showed off nicely in the webinars itself. MADDIE: 28:34 Yeah. And I'm curious what you guys learned from the process, overall. And I think one of the things that stood out to me was inviting friends to the webinar to distract you. I think that that's always a good thing to do. But other than that, what did you guys learn from the process? WILL: 28:52 I'd say it's the converse. Never mention it to your friends. [laughter] I thought it'd be a rare opportunity; they might be interested in what I do when it's football related. But I think they all got a bit excited. [laughter] So they won't be invited to any future ones. But generally, in terms of it, and like Luke mentioned at the start, first, I've always been interested in sports and interested in data, but I've never really gone kind of behind the scenes and taken the time to understand more. And ultimately, we wanted to present something back that people could learn from. And so I wanted to make sure that we were knowledgeable, and we were skilled in all of this. So it's actually been a great opportunity to expand our learnings. And I think, in having that and sort of, I guess, putting you outside your comfort zone of, "You need to go and produce something," and set a challenge, it then kind of inspires you/forces you to go and do that reading. And I personally found it incredibly interesting. It meant Luke and I, in our discussions, actually sounded a little bit more intelligent, or I'd like to think anyway, [laughter] maybe with the conversations behind the scenes. But also, there's a part that it's great to be able to talk to so many people in that industry and just to get different people's perspectives. As we've already mentioned, their approach is similar, but the data that they're working with and how they operate is really interesting. So it's just been a great experience, I'd say, of learning more and meeting new people and understanding the ways that they work. LUKE: 30:24 I was reading the other day about [Forster?] Rugby Club in England, one of the premiership rugby clubs. They spoke to analysts and people from all areas of world sport when they were looking at bringing analytics more into what they were doing. And they brought in people from the NFL, people from the Team GB, cycling, things like that to try and understand different learnings. There are so many interesting books around sports or data. And it's not just money ball. There's a load of different bits of literature that you can read to learn from. MADDIE: 31:03 Yeah. In one of your blog posts that you wrote, Luke, I know that you listed some resources for some other learnings. And I think it'd be fun if you guys wanted to share more of your favorite books or movies or podcasts or anything that's related to this, and we can link them in the show notes for people. I think they would like that. WILL: 31:25 I think, actually, the one that got us sort of started on it, and this was a few months back - and Luke, did you recommend it? - but the Football Hackers book is, sort of, I would suggest, a really good gateway sports book in the-- again, because that does reference a lot of material, and internally, sort of in our team at Alteryx, quite a few of us read it, and that's really what-- LUKE: 31:48 That's a part of our book club, wasn't it? WILL: 31:50 Yeah, a part of the book club, always learning. But I would say that that was a great starting point, and certainly one that I found incredibly useful. And I think when it's topics like xG, that was kind of the introductory explanations that I was getting and the first time I sort of saw behind the scenes of it. MADDIE: 32:10 Yeah. I like the back stories of some of these models and methodologies that people are using. It's really interesting. Cool. LUKE: 32:18 Yeah. I mean, and there's some really interesting characters, especially in football that got involved and did all this stuff. And the storyline is actually really interesting because it's written by a German guy who got involved, and he suddenly got into a really high position in one of the German football teams because of his knowledge of data. And you then start realizing that all of the famous names of managers that everyone just knows now, their history was not necessarily playing football, and they had that experience, but they're starting to be educated in the world of data and understand how to leverage that to better the team. And then there's loads of stories about how data is used to identify new players and work out who gets promoted into the team from the use team or the B team or whoever it is. And so yeah, there's some really interesting stories. I very much recommend that book. That's Football Hackers: The Science and Art of a Data Revolution. MADDIE: 33:27 Yeah. Thank you. We'll be sure to link to that. So what can we expect next from you guys? WILL: 33:34 Good question. [laughter] I mean, for us, and a lot of the reason for the kind of sport and analytics in football is a great start to it is that, ultimately, we've kind of created a framework now to help other people and probably continue, ourselves, to apply this to other sports and other competitions. There's pretty much a-- there's major international tournaments every year across different sports, across domestic leagues. There's stuff happening all the time. So I think, personally, from us, we'll probably get more involved into some of those. It's just a good opportunity to hopefully kind of apply more across different sports areas. And hopefully, what we can do-- and we've got Alteryx as a really good platform to bring it to life for a lot of people. I mean, ultimately, what we wanted from the webinar series was to get people more passionate. For those that either are interested in data and not in sport, they might go one way. For those that are really interested in sport but have never been interested in the data that, hopefully, they might see that transition, get more invested. So I mean, ultimately, that's mine. If I can continue working and watching sport, that's the dream. So if anybody from Reading Football Club is listening, [laughter] I'm all ears for any job opportunities there [laughter] as well. MADDIE: 34:53 Nice plug. Okay. [laughter] Awesome. LUKE: 34:57 Yeah. I mean, as Will said, there's plenty of projects and things to be going on with. If anyone's got any ideas around how we can sort of merge the world of data analytics that we work in day-to-day, whether it's in finance or marketing or whatever it might be, and passions that you have outside, not just sporting, but where we can get ahold of data and mess around with it, I think that that'd be a really good thing to do. [music] MADDIE: 35:27 Well, thank you both so much for joining me. This is so fun. LUKE: 35:32 It's great to be invited. Yeah. [laughter] WILL: 35:32 Thank you for having us. MADDIE: 35:37 Thanks for listening. To explore assisted modeling with Alteryx and to hear more from Will and Luke, including their 2020 Euros webinar series, check out our show notes at community.alteryx.com/podcast. Catch you next time. [music] Luke are you having a pint now? LUKE: 36:02 This is not a pint. This is not a quiet pint. It's just a can. MADDIE: 36:06 Okay. Just, do you call it a can? WILL: 36:08 For anybody listening, it's Monday, 9 AM in London at the moment. [laughter] LUKE: 36:12 That is not true. I just realized-- MADDIE: 36:14 And you're just cracking a beer. LUKE: 36:14 --it's getting to the end of the day. [laughter] I've noticed that we're getting towards the end of recording. I had been looking forward to this Thursday afternoon beer, recording this podcast, and realized that I hadn't gone and got one. So yeah. That's my lunch. MADDIE: 36:27 Well, to be fair, yeah, it was my idea initially when we had prep calls, I was like, "We should drink beers during this call." And I personally dropped the ball. So I didn't expect you guys to do it because it is literally 9 AM in Denver for me. So I was going to make a mimosa, and then I decided I didn't want to be hiccupping throughout this whole recording, [laughter] so. But I'm proud of you for having a beer like we originally planned. LUKE: 36:54 One of the revelations of the last few months was getting a wireless headset because it's opened up the entire house to me [laughter] whilst I'm working. So yeah. Just able to go to the fridge, get myself a nice little IPA, crack that open. MADDIE: 37:10 That's great. I'm glad that we could hear it, too, on the audio. I appreciate that. [laughter]

This episode of Alter Everything was produced by Maddie Johannsen (@MaddieJ).
Special thanks to @andyuttley for the theme music track, and @TaraM for our album artwork.

Alter Everything Podcast

Podcast Guide

91: Analytics in sports

Panelists

Topics

Transcript

This episode of Alter Everything was produced by Maddie Johannsen (@MaddieJ).
Special thanks to @andyuttley for the theme music track, and @TaraM for our album artwork.

Alter Everything Podcast

Podcast Guide

91: Analytics in sports

Panelists

Topics

Transcript

This episode of Alter Everything was produced by Maddie Johannsen (@MaddieJ).Special thanks to @andyuttley for the theme music track, and @TaraM for our album artwork.

This episode of Alter Everything was produced by Maddie Johannsen (@MaddieJ).
Special thanks to @andyuttley for the theme music track, and @TaraM for our album artwork.