Sucking wind in La Paz – Sloan Sports Analytics Conference poster

Last year, I ran across an article analyzing whether high altitudes matter in South American international soccer. The authors run regressions with the away team’s altitude change as an explanatory variable and find that climbing but especially descending hurts the away team. Descending 2,000 meters seems to lower away team winning percentage by over 10 points. Pretty surprising finding.

However, these regressions only control for home team quality. It just so happens that Bolivia, Ecuador, and Colombia (i.e., the descending teams) are historically weak teams. What we have here is omitted variables bias — the naive analysis implies that descending harms performance, when in reality descending teams just aren’t very good.

On my quest for causation, I repeated the analysis and controlled for home and away team quality (you can get the full article from my website plus some press links). First off, controlling for away team quality (using team fixed effects) makes all the difference; the descent effect is wiped out.

What about climbing? Anecdotes abound, and FIFA tried a high-altitude ban for awhile, but it was met with ridicule. I’m not one to believe anecdotes and stories. With home team and away team fixed effects interacted, my effects compare games between the same two teams, but at different altitude.  For example, Bolivia plays near 4,000 meters in La Paz but also plays in Santa Cruz at 400 meters. My approach asks whether Brazil plays worse against Bolivia in La Paz than in Santa Cruz.

I find that climbing matters a lot. Averaging across all match-ups, Bolivia’s winning percentage increases by 45 points when teams travel up to La Paz, and Ecuador gets a 29 percentage point altitude boost in Quito. However, a puzzle remains: Bogota, Colombia is only 250 meters (800 feet) lower than Quito, Ecuador, yet Colombia’s winning percentage drops 20 points on average in Bogota.

I think I’m getting close to the causal effect here, and I got some great feedback on my poster at the Sloan Sports Analytics Conference, but the differences between Ecuador and Colombia encourage more work on this topic (I’m quite convinced by the strong Bolivia results). Omitted variables are always lurking; any ideas to explain these findings?

4 responses to “Sucking wind in La Paz – Sloan Sports Analytics Conference poster

  1. . A thought on the Bogota problem: What’s the difference between Columbia and the other “descenders”? Columbia is the only team where the majority of the team plays in Europe. It also follows from that (I’m too lazy to figure it out), that the vast majority of the team’s starters play in Europe. This means that they’re not getting the “boost” from training and living at altitude, they’re being shipped in to play internationals. Quick check? The team with the most players who play domestically, in this case Bolivia, should receive the greatest altitude boost, which they do.

    • Great idea – this is the best explanation I’ve seen. The easy test would be a difference-in-difference: take the difference in high altitude and low altitude winning pct when Colombia has starters who play in Europe, and compare that to the same difference when Colombia has fewer Euro guys on the roster. I bet there’s enough variation in that to do the test. I’ll try to get to this at some point.

  2. One other variable might be the fact that Colombia also plays national matches in Baranquilla, which is at sea level. Not certain how many matches are played at each venue (research is your strength, not mine) but If the Colombians are not doing all of their training/matches at altitude, even ‘domestic’ players may not get the benefit of altitude.

    • Thanks Bill – that’s a good point. It’s related to Adrian’s angle that Colombia may have more players on European clubs, meaning that they don’t train year round at altitude. The other high altitude countries also play at a low altitude site (I’m using these games as the “control” games), but maybe they don’t play there as often. Some things to check!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s