Is Bill James’ temperature a leading indicator?

Warning: nerd post. I’m actually playing with baseball stats instead of watching a game because my son is playing Call of Duty on the TV.

At Bill James Online (subscription site), Bill keeps track of each team’s “temperature.” The idea is to literally measure how “hot” or “cold” each team is. The system is calculated so that an average team has a room temperature of 72 degrees. The specific math is simple:

1. Each team’s temperature at the start of the season is 72 degrees
2. After each game, multiply the team’s pre-game temperature by .8958334
3. If they won the game, add 15 degrees

This system reminds me of Win Shares. I have no idea how he came up with the math (seven digits for the multiplier? Really?), but it seems to “work” in its own way. Essentially, the multiplier acts to “regress” the team’s record because 90% of a 120 degree temperature is a lot more than 90% of a 40 degree temperature.

Let’s look at the temperature in action. Here is a graph of the Braves’ record this season (the blue line showing games above/below .500) and their temperature (the red dotted line):

image

I purposely set the scale so that a .500 record is roughly on the same level as 72 degrees. On June 3, the Braves reached a higher temperature (122 degrees) than any other team this year.

The cool thing is that the Braves’ temperature has seemingly been a leading indicator of their record. When the red line has been above the blue, the blue line has risen. When it’s fallen below, their record has fallen.

Having said that, I don’t actually believe the Bill James Temp consistently acts as a leading indicator. But it does make a cool graph.


Dave Studeman was called a "national treasure" by Rob Neyer. Seriously. Follow his sporadic tweets @dastudes.
12 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Tree
13 years ago

It’s a little hard to tell, but it looks like you plotted the temperature after the game, when you should look at it before the game. (Temperature is peaking at the same time as their winning percentage; it should peak the day after, since it doesn’t come down until they actually lose.)

Dave Studeman
13 years ago

You’re right that the post-game temperature is posted on the same day as the game.  Doesn’t materially change the result.

RMR
13 years ago

I’m not sure your interpretation is accurate.  Notice how the peaks align?  If this was truly a leading indicator, the temperature peaks would occur before the win% peaks.  Rather, we’re just seeing that the temperature scale has greater variation. 

But since both are the result of the same good/bad performances on the field, it’s an issue of spurious correlation.  Changes in temperature are not occurring before the changes in win %, but rather simultaneously.

Tom M. Tango
13 years ago

I also don’t know that you’ve really matched the two scales.  If you change the range of the righthand scale by adding 20 or 30 extra degrees to both ends, thereby compressing that scale, you’ll probably be almost identical in the two lines.  Try it out.

Tom M. Tango
13 years ago

By the way, 1 – 15/72/2 =
0.8958333

James is using:
0.8958334

Had he used 14 and 70, the multiplier would have been 0.90.

obsessivegiantscompulsive
13 years ago

RMR, the point, I believe, is that when the temperature is higher than the blue line, the team is headed upward in terms of future performance.

It also appears to be acting like how stock market technical analysts like to use moving average lines to signal changes in momentum.  If the Braves record was a “stock”, “buying” when the temperature goes above the team record and selling when it goes below, the crossover would have “capture” more of the “stock” rise and fall.

Not good as a signal though, see how it went above and below a lot early on, no usable signal really.

I have to agree that it appears to be spurious correlation.

Tom M. Tango
13 years ago

Notice what happens for the most part: before each game, you take out 10% of the temperature, say going from 70 to 63 (i.e., minus 7), and if you win, you add 14 degrees (using my simpler scale), or plus 7 from the 70.

So, he’s adding +7 for a win, and -7 for a loss.

However, if it’s a team that already has alot of wins, say a 120 temp, they get minus 12 on a loss and +2 on a win.  If they are cold, say 20 degrees, they get minus 2 on a loss and plus 12 on a win.

Basically, a team that is really hot will basically max out to a certain temperature and stay there, and a cold team will bottom out.  And doing the opposite will quickly move them toward the center.

I really don’t see much there.

John K
13 years ago

I feel it’s important to point out that you made the increasingly common and annoying mistake of using the word “literally” prior to a figurative expression.

Dave Studeman
13 years ago

Tough crowd.  I’ll just point out my last sentence, the conclusion of the article: Having said that, I don’t actually believe the Bill James Temp consistently acts as a leading indicator. But it does make a cool graph.

Dave Studeman
13 years ago

I feel it’s important to point out that you made the increasingly common and annoying mistake of using the word “literally” prior to a figurative expression.

Yes, that was the idea.

studes
13 years ago

People are taking this more seriously than I intended. When I say that the team goes up when the red line is higher, I’m only referring to this graph.  As I said at the end, I don’t think the Bill James temperature is really a leading indicator.  It was just a cool graph.

I’m positive, by the way, that Bill wouldn’t advocate using the temperature this way.

Project_Zero
13 years ago

The temperature setup was well thought out. if the set “room temperature” was 72 degrees, the lower limit is 0 degrees, and the upper limit is 144 degrees under this system, so that 72 falls right in the middle. Not coincidentally, what is 144 x 0.8958334? 129. (Just about) Add 15 degrees to that, and you get the upper limit of 144. It’s quite simple.

There is no one formula to extrapolate the temperatures long term only given the win-loss numbers, since the temperatures are largely streak-dependent.

“when the temperature is higher than the blue line, the team is headed upward in terms of future performance.”

I’m not so sure, since the scale between games +/- .500 and “temperature” are bound to vary with the chance of extremes playing into this. Also, while the temperatures can vary from 0 to 144 degrees, the number of games +/- .500 theoretically could vary by +/- 81; the former is more likely to occur.