Halladay in heat

10:43 - Several posts ago, I mused aloud about Roy Halladay and his dislike of hot weather. I wondered to myself, and to you, about whether this had historically been a weakness of Halladay's.

Well, sound the horns, we have our first flash piece of sabermetric analysis here at the Star Baseball Blog. Not that I'm capable of doing anything like this. I'm still confused about the difference between Battlestar Galactica Seasons 2.0 and 2.5.

No, no, one of our thousands and thousands of loyal readers leapt to the challenge. Rob Pettapiece.

Rob first sent me a short note, claiming to have done the math. I think it's worth quoting him at length here.

"Using (Halladay's) starts from 2002 to 2007, there is no correlation between the game-time temperature and Halladay's Game Score (Bill James' metric to assess pitching performance - C.K.). Since hitters are usually more dangerous in the heat, this may mean that Halladay is actually 'better' when it's warm out."

Then Rob got shy about the numbers, and I smelled a rat. I demanded - DEMANDED - the numbers. He happily sent them along. I opened them up and suddenly the colour green looked like red to me. Here was a man, Rob Pettapiece, making FIGURES equal ANSWERS! I'm still a bit confused by the things you can do with stats, besides manipulate for your own nefarious ends if you are a mortgage broker. Rob wins, I lose. Halladay can do it all in any weather.

Tonight, assuming the Dome is once again closed, Halladay will have to struggle through in room temperature. Oh man, where's that April heat wave when we need one?

Here, by the way, are Rob's labours. And I will take my tongue out of my cheek long enough to thank him profusely for taking the time and effort. After all, somebody has to supply you people with real insights.

Here's Rob's graphic representation of 173 Halladay starts from April 4, 2002 to Sept. 26, 2007.

SOURCE: ROB PETTAPIECE

On the vertical axis, we have the game score, from a low of 6 on June 5, 2007 in 68 F temps to a high of 93 on May 29, 2005. During that best start, Halladay blanked the Minnesota Twins, allowing two hits, striking out 10 and walking no one. The game time temp was also 68 F, under a closed Dome.

And here are Rob's summary, showing negligble difference in a Halladay start even if he is sweating buckets up there:

Temperatures <50 50s 60s 70s 80s 90s Starts 2 11 49 78 24 9 Average GS 45.0 61.2 56.5 57.2 53.4 51.8 Overall GS Avg. 56.6 Correlation, temperature/GS -.0718

Awesome. I feel so Einstein-y. Or maybe Igor-y to Rob's Einstein. Either way, take that, puny subjective analysis!

Cathal Kelly

Comments

[Alex]

R-squared for the win!! (stats reference) Rice job Rob Pettapiece, you should get a job down in 'the truck'!

Now if you wanted a really painful job, try mapping out AJ Burnett's Game Scores against, oh I don't know, the faint glimmers of hope we occasionally have that he will pitch well.

[rob]

i can't even wrap my head around this blog post, let alone the link to all those numbers.

[Mike Piczak]

It was with interest that I read the blog posting on the correlation between game score and temperature. What has not been provided is two things:
1. whether the correlation stated (-.07) is statistically significant and
2. the linear relationship (or equation) between temperature and game score so that a manager could decide whether to put a particular pitcher into the game

A definition of game score would also be helpful to decide whether there is a theoretical relationship between the two hypothesized variables.

Mr. Pettapiece's supplying the original raw data would also be helpful.

Mike Piczak

[Mike Piczak]

I just ran the data taken from the table provided above and found the r squared (the proportion of the variation in the game score variable as explained by temperature) to be approximately 1.2% which means that about 98.8% of the pitcher's game score is unexplained by temperature. The correlation coefficient, r, would be the square root of 1.2% which would be about 10%. This suggests that there is no relationship between the two variables indicated by Mr. Pettapiece.

The line of best fit through his data is almost parallel to the x axis suggesting that there is a complete absence of any linear relationship between the x and y variables. In other words, for a given value of temperature, one could anticipate almost ANY value for y.

Mike Piczak