A New Ballpark Doesn't Fix Everything On Its Own: Insights from a Machine Learning Model

When (and if) the A's build a new ballpark, there is no guarantee that the new ballpark will generate dramatically higher attendance and more revenue. Winning is key, and an awesome ballpark is key. An awesome ballpark with a terrible team may end up being a net negative. Being downtown helps a bit but it's hard to say how much.

I used a machine learning model called tree boosting to predict MLB attendance from 1995-2018. I was able to use the model to simulate a few scenarios. I've posted detailed results on my blog (Part 1, Part 1b) but thought it would be worth cross-posting here.

Before delving into results, here's how I thought about a new ballpark affects attendance:

• One component is purely a "new ballpark effect" - if the A's had a new ballpark, but literally everything else was the same, how much does the new ballpark increase attendance? When I say everything else the same, I'm including how fans perceive and feel about the team.
• The second component is "fan engagement." A new ballpark should boost fan engagement for at least one year. Fans will feel more excited about the team and are more likely to go to games. Fan engagement is basically elasticity of demand—how many fans are willing to go to see a third-place team play on a cold night? Or you can think of it as a measure of "fan stoke."

First, I looked at how other new ballparks affected fan engagement. To do that, I pretended that the team stayed in their old ballpark. Additionally, I pretended that "fan engagement" stayed at a baseline, which I set as the last year in their old ballpark. For example, for the Giants, I used 1999 as the baseline year. Then I predicted attendance for 2000 to 2018, using all of the game and weather data in the model.

Here are some of the bad ones. Again, remember, we're pretending that the Mariners stayed in the Kingdome and their fans stayed "as stoked" as they were in 1999.

Next unsurprisingly, is the Marlins.

Okay, and now a few of the better ones. I would consider the Padres to be middle-of-the-road. This is maybe a reasonable expectation for how a new A's ballpark would affect attendance.

PNC Park is a really interesting case. It's commonly regarded as a top of the line stadium. But it didn't improve fan engagement much at all over Three Rivers Stadium. I think this is because they were very bad only one year after opening PNC.

We'll wrap up with the two best ones: the Phillies and the Giants. I don't know much about Citizens Bank Park but I haven't heard much about it being a stand out park. So that would suggest their great attendance numbers are more about a winning team, and a pretty passionate fanbase in Philly.

And then there are the Giants. The Giants completely transformed their franchise with AT&T Park. We all know that and have lived it. I was surprised at how much dramatic this is—the Giants did something no one else has done. There are are few more teams in my blog post.

New Ballpark Effect - A's Stadium Simulation

Now let's look at some predictions for a new A's stadium. This is just looking at the "pure ballpark effect." Again, this does not include a boost to fan engagement, which can be huge as we have just seen. So these are all very conservative estimates. I think I would be willing to bet on these as a lower end.

The way I made these predictions is a little weird. I pretended that the new stadium was constructed in either 2008 or 2011. It will be obvious which year because there's a big bump for first year attendance. I also plugged in a few different stadiums, but still used Oakland weather and population.

If an A's stadium was built in 2008, there would definitely be a first year bump. But the attendance would have likely declined pretty quickly in subsequent years since the A's were bad in this time period. We could see a PNC Park scenario. Fans would have be engaged in other ways, which is one reason why a "downtown" or super accessible stadium helps.

If the A's built a new stadium in 2012, then we would likely have seen an extended run of pretty good attendance. Remember these are conservative estimates and the model has never seen a year where the A's had an average attendance over 30,000 (since I used data from 1995 to 2018).

Conclusions

Well I hope this was interesting and you learned something! I think I did—I thought that parks like PNC and Safeco would definitively have increased attendance for the Pirates.

Here are my takeaways:

1. A new ballpark is no guarantee to increase attendance past the first year.

2. Winning in the new ballpark is extremely important to create fan engagement.

3. The stadium itself is also important, but maybe less important than winning (see: PNC Park). The long-term play is all about fan engagement - how to get fans to show up when the team isn't in first place, on a cold weekday night. Part of that may be convenience and the stadium experience.

If you want some more details (or want to know more about the methods), check out my blog post (Part 1, Part 1b)!