Printer-friendly

Contact Author

April 28, 2003

Doctoring The Numbers

Hot Starts, Part II

by Rany Jazayerli

Welcome to Part 2 of our look at the importance of hot starts. If you haven't already, read Part 1 first. We'll wait for you to get back.

Last time, I looked at how teams fared at season's end after starting the season with a particular record, varying the data by looking at starts of varying lengths. While I pointed out general trends in the data (as well as the exceptions that proved the rule), I did not sum up the data concisely into a single, coherent formula to predict a team's final record. That's what today's article is about. In Part 3--yes, there will be a Part 3--I want to examine how the interaction between a team's record at the start of the season, and its record the previous season, affects its final winning percentage.

WARNING: This article contains lots of graphs, formulas, and other items found in an egghead's tool kit. If numbers aren't your thing, you're welcome to scroll down to the end and skip to the conclusions. For those of you who like to know what went into the batter (no, not that kind of batter), read on.

As I pointed out in Part 1, there are measurable differences in the final performances of teams based on their record after even just 10 games. The following graph compares a team's winning percentage after 10 games (meaning there are only 11 possible records, ranging from 0-10, 1-9, etc, all the way to 10-0) to its final winning percentage. The graph shows plots for all 1300-odd teams in the study:

The blue line represents a "best-fit" linear regression that comes closest to representing the fortunes of every team in a single formula. That formula, expressed numerically, is:

Y = .398 + .205 * X     r2 = 0.197