Out of the big four professional services firms, KPMG’s Director of Sports Analytics Ryan McCumber started by telling Bullpen that they want to own the sports analytics space with their primary assets, game scheduling, real time in-stadium analytics and the Sports Analytics World Series a global conference series showcasing innovation, technology data and analytics which are driving sport.
By focusing in on KPMG’s game scheduling software McCumber explained how they are solving a complex problem that goes against decades of tradition. Instead of painstaking guessing, KPMG has developed a sophisticated algorithm that runs through trillions of possible solutions and schedules to land on the best one which addresses simultaneously competing league-wide objectives including rest, travel, stadium availability and broadcasting.
One of the aims is to bridge the gap between the wide variances in scheduling for teams as historically some teams have been given more favourable schedules throughout a league season.
“In the NBA’s case, we go through 200 million schedules a minute and we went through 35 trillion at the end of the day to land on the final one,” added McCumber.
With schedules developed for the NBA, the NBL in Australia, Major League Baseball, plus leagues across Europe and Asia to be on-boarded, the aim is to ensure that an optimised schedule that utilises an evidence-based approach will produce a fairer and better schedule for all stakeholders.
Bullpen: I’ll focus in on KPMG’s game scheduling software, can you describe what it is?
Ryan McCumber: “It’s a solution that uses the power of the cloud and a proprietary algorithm model which allows us to basically optimise a schedule; who plays where, when, what time and day of the week. It then factors in all kinds of constraints such as travel, arena capacity, arena availability, then we look at historical gate attendance, historical TV ratings. Then on top of that we factor in extra things like preferred games and certain blackout dates.”
BP: How many stakeholders actually have input into that, do you speak to governing bodies, leagues, broadcasters who in turn offer up what they want and then you input that into your algorithm?
RM: “Pretty much all of them and in a general statement as opposed to mentioning one league it starts with a few people within league office who own the schedule, they will work through collecting information from the various clubs, stadiums and venues for things such as blackout dates or preferred games.
“For one stakeholder blackout dates and when home games could be played were the primary inputs. Another stakeholder prioritised broadcast because sometimes there’s contractual obligations to broadcasters that you have to meet, and other times it’s input from broadcasters. The next input comes from the players association and that’s about player welfare.
“While in Europe there was actually input from entities that we never considered. There’s input from the police, the mayor of every city as well as transport authorities who also have to sign off on the schedule.”
BP: Are there stark differences when considering the Australian market as opposed to the US, European and UK markets when it comes to determining what stakeholders demand out of a schedule?
RM: “The US is leading the way on player welfare and really trying to increase that. In Europe it’s league by league for player welfare, while they all say they care about player welfare certain teams put their own league first as opposed to say continental commitments such as the Champions League or Europa League. They won’t factor those in because they feel like their league is bigger than the continental competitions as the team might have only two or three days rest, whereas other leagues in Europe will factor that in and will make sure that they have an advantage in those competitions so they can succeed and hopefully provide benefits to the domestic league.”
BP: Does the algorithm learn itself or do you have to keep putting in inputs?
RM: “It’s an art and science, identifying the constraints is the science and then the art comes in of what value you’re going to assign to that constraint and that value is usually in numerical form.
“In just about every league in the world you can never really get to the perfect schedule.”
BP: The NBA season began on October 17, having usually started later in the month. The reasoning given is that the league did want to reduce the amount of back to back games. Was there any consultation with KPMG or did KPMG have any involvement with the expansion of the schedule?
RM: “You’ll notice some significant improvements in this season’s schedule. We have eliminated four and five games stretches.”
BP: To extend the season by a week or two seems so obvious. Do you think leagues and governing bodies rigidly protect what has been done in the past?
RM: “I don’t want to speak for the NBA’s contract terms, but a lot of them were for broadcasters. However, the NBA is going to start to put more games head-to-head with American football. That’s an issue that they have been reluctant to go against before.”
BP: What Australian organisations use KPMG’s scheduling software solution?
RM: “The NBL was the first one in Australia. We do have numerous proofs of concepts underway as well as other projects which I can’t disclose yet.”
BP: Can you explain the difficulty of scheduling the NBL’s fixtures?
RM: “We originally thought it was going to be simple compared to the NBA, it’s eight teams and pretty basic, the road trips are simple and the time difference is similar to the US. It really didn’t seem that complex to us. However, the complexity is that availability dates for venues was extremely limited and it’s always ever-changing.
“In the NBL, there are events you know are going to be fixed, such as Melbourne’s Hisense Arena not being available in January. In other cities, there were ever-changing arena availabilities and then the league was managing two different broadcast deals which were going to prioritise games in different time slots.”
BP: How long did you work on the NBL partnership?
RM: “We engaged them 12 months ago, started the process of showing them what capabilities we had with the NBA. Heads down, the real duration is about six to eight weeks really full on and about three months total to generate that schedule.”
BP: As this is the first of hopefully many seasons, the feedback from the first season should be interesting.
RM: “Ideally every year the schedules should consistently get better as we keep working on the scheduling model.”
BP: At the moment, we hear a lot of anecdotal evidence and complaints about unfavourable scheduling. Could you briefly describe how metrics get scored?
RM: “One metric is the total score another metric is how many games where there were four-day breaks. We put up overall KPI’s comparing the teams and we can see that some teams had favourable schedules and some were significantly unfavourable.”
BP: Does it then infer that athletes actually play too many games?
RM: “NBA coaches believe so, Steve Kerr, Gregg Popovich is notorious for it, sitting his star players or his older players in back-to-back games.
“There is a gripe with players that don’t play in back-to-back games and the NBA has had to address it. Within the model we tried to compensate for that, teams that may sit down players for back-to-backs can be penalised.”
BP: Have you experimented with other sports?
RM: “Using a proof of concept we did for the Big Bash League and other cricket competitions such as the Indian Premier League, those leagues revolve around its stars and where they’re playing. There’s not too many games and the problem is schedules get made before some star players agree to join certain teams and that schedule will initially not be optimised for that star.”
BP: What makes KPMG an ideal operator for game scheduling?
RM: “There are things we are bringing to the table that I would almost call disrupting the game scheduling market, we start to apply things like commercial revenue, game attendance and factual numbers that factor into the schedule where previously the leagues would go by hand and say, “this looks like a good game and this looks like a good game,” and there wouldn’t be science behind it, it would be more anecdotal.
“We went and created a new platform, a digital platform, and one thing in the NBA’s case, we were the first ones to release an actual portal to collect information digitally between all the clubs. Every club logs in, it documents what they’re doing, and they’re submitting their content such as priority dates, availability dates and that content then gets approved by the league.
“If there are complaints with scheduling, the league can take back control. Lastly, sharing this kind of data and giving a fuller picture makes for greater transparency.”