## Dome at Cold Revisited

There were 4 dome teams playing in wintery weather today, and all four lost. DET lost at PHI, MIN lost at BAL, IND lost at CIN, and ATL lost at GB.

A few years ago I looked at how dome teams tend to struggle in cold weather. I wanted to know if dome team underperformance in the cold was indeed true and if so how big was the effect. The answers were: Yes and huge. Last season I redid things using actual game temperatures and found just as big an effect.

With the weather as it is on the east coast today, I looked at this phenomenon once again. We have two more years worth of data thanks to the addition of 1999. Plus, I was able to reconstruct nearly another half season worth of data by replacing missing game temperatures from the gamebooks. I also broke out teams that played in retractable-roof stadiums by season.

Here are the results if we count retractable teams as dome teams. There's a case to be made that retractable home environments are closer to dome environments than open air stadiums. The chart below plots road team winning percentage according to game temperature.

Keep in mind sample sizes become small at temperature extremes. The key observation is not that dome teams lose in the cold, but that they lose more often in the cold than they do in moderate temperatures. You can see that visiting dome teams win nearly as often as their open-air counterparts do when the temperature is mild or warm.

Here are the results if we separate dome teams, retractable teams, and open air visitors. The sample sizes get even smaller at the extremes, particularly  for the dome/retractable categories. For example, there are only 4 cases of retractable-roof teams playing in the 19-28 degree temperature bin.

### 4 Responses to “Dome at Cold Revisited”

1. Anonymous says:

This would be another situation where a p-value or confidence interval would be helpful. With sample sizes as small as this, I look at the graph and have no idea whether I'm looking at a type I error.

2. Nathan Lazarus says:

I'm struck by the 0 victory percentage for dome visitors in 9-18 degree weather and the 10% in 19-28. What's the sample size on those?

3. Anonymous says:

I'm struck even more by the data point leftmost on the blue line. Clearly the result of small sample size, but seems to suggest that there is a temp below which two outdoor teams would rather the especially cold game be played in the *other* team's city. Wierd.

Also, tongue in cheek, this should provide an answer to those still claiming PManning isn't good in the cold weather: that used to be true when he was a Colt, but now that he's a Bronco it is clearly false. If it is really cold, however, he'd rather play either at home against a dome visitor or be on the road in another cold city.

But Peyton's answer yesterday was better than that.

4. basballguy says:

Are we able to see the actual data involved? Like say each team's record for each weather bracket?