Tuesday, October 30, 2012

Army's computer used to predict 1957 BAs

I'm going to introduce today's topic with a couple of paragraphs lifted from elsewhere on the internet. Each is slightly edited.

This description comes from the CED Magic site: 

RCA introduced the Bizmac computer in 1956. It was the company's first computer with commercial potential. Initially it was installed for military use, as in the above picture where Arthur Malcarney of RCA explains the control panel to Brig. General Nelson Lynde of the Army Ordnance Tank-Automotive Command in Detroit. The army purchased the $4 million machine containing 25,000 vacuum tubes for inventory management. 

This appears on ACM Digital Library site:

The RCA BIZMAC computer was developed as a major element of the RCA BIZMAC system which was primarily intended to handle cyclical accounting, such as inventory control, as employed at the Ordnance Tank and Automotive Command in Detroit, Mich. The design also permitted  the computer to be used in areas other than cyclical accounting; e.g., digital system simulation and statistical analysis.

The Bizmac was hot stuff in its day, as it $4 million price tag attests. As mentioned, the Army bought the first system to keep track of its two billion spare parts for tanks.

The system was written up in all the popular scientific magazines of the day, and probably was also thoroughly covered in scientific journals and periodicals for military wonks.

So what's the connection to a baseball blog? 

In an effort to showcase computer power in a way that the average Joe Taxpayer could understand, prior to the 1957 baseball season, somebody put Bizmac to work predicting major league batting averages for a dozen star players in the upcoming season.

Statistics from the previous five seasons were fed into Bizmac. After what was no doubt hours, if not days, of numbers crunching, the computer made it predictions. The numbers were reported in several periodicals, including The Sporting News' March 20 issue.

This chart lists the players' computed averages, their actual marks for 1957, and the variance.

Player                        Bizmac    Actual   Difference
Mickey Mantle          .342           .365        +23
Richie Ashburn         .328          .297        (31)
Ted Williams            .322          .388        +66
Harvey Kuenn           .319          .277        (42)
Minnie Minoso         .317           .310        (7) 
Carl Furillo                .314           .306       (8)
Ray Boone               .313           .273        (40)
Nellie Fox                 .309           .317        +8
Stan Musial              .305           .351       +46
Ted Kluszewski       .304           .274        (30)
Duke Snider            .302           .268        (34)
Yogi Berra               .297           .251        (46)

The average deviation between the computer's prediction and the actual 1957 stats was 10.12%. The composite average of Bizmac's predictions was .316. The actual 1957 marks came to an average of .305; Bizmac overestimated the players' performance by just under 3.5%. I'm horrible at math, so I have no idea how to analyze Bizmac's performance. 

Today, I'm sure that there's an app on the iPhone that could do as good a job--for whatever that's worth. 
No, that not the Bizmac computer on the desk in the foreground.
Bizmac is down the right-hand side, as far as the eye can see.

No comments:

Post a Comment

Your comments, criticism, additional information, questions, etc., are welcome . . . as long as they are germane to the original topic. All comments are moderated before they are allowed to appear and spam comments are deleted before they ever appear. No "Anonymous User" comments are allowed.