The sports' Beane counters

Billy Beane was a pioneer. The manager of the Oakland Athletics baseball team, now immortalised by Brad Pitt's portrayal in Moneyball, was among the first people in sport to apply rigorous statistical analysis - called sabermetrics - to improve the way players were evaluated.

Beane most valued one statistic for hitters - on-base percentage, a measure of how often a batter reaches base for any reason other than a fielding team error.

This sabermetric concept, initially applied to baseball in "search for objective knowledge", is by no means unique to that sport. Now that football and rugby seasons have begun, the same search is conducted in those sports, largely out of sight of most observers. Our perception of game analysis is somewhat "dumbed down" by what we see in TV match previews or wrap-ups. Here, in-studio experts show slow-motion replays with animated arrows to demonstrate how goals and tries were scored, errors were exploited, and how key moments occurred. This is certainly part of game analysis but it is merely the tip of the iceberg.

That iceberg is created by enormous quantities of data, gathered and then analysed by statisticians and coaches to help plan team tactics for upcoming matches and to evaluate individual performances.

The volume of data is staggering. Companies exist whose sole purpose is to capture every detail of a match. These companies employ hundreds of game analysts who "code" every professional rugby and football match. Coding means watching a match and labelling every single event into a statistical file that is then used to evaluate performance. Quite literally, a match is watched by a person whose job it is to click mouse-buttons that label every single event. Every tackle, pass, lineout, scrum, ball-carry and kick is coded, tallied and filed away for a discerning eye, like Billy Beane's, to analyse and search for hidden value. Consider that it takes between 25 and 40 hours to analyse an 80-minute match, and you have some idea of the detail that exists behind the scenes.

For example, an analyst might code lineouts for the following outcomes: Who is the catcher, who lifts, what movement occurs before the lift and throw, what happens after the throw, what happens off the lineout (is the ball passed or retained in a maul), and if passed, where is the next phase likely to be set up or will the scrumhalf or fly-half kick for territory? This is then categorised for location on the field. Armed with this information, a coach can begin to anticipate what the opposition is likely to do. The same kind of analysis for tackles and rucks might reveal how many metres are gained per ruck by each player in every tackle, how many players the opposition are likely to commit in attack and defence, how long the ball is retained in each ruck, and where the next ruck is likely to form.

Key to all this, however, is to simplify the data minefield and avoid being paralysed by the sheer volume of statistics. That applies to coaches and players, who can hardly be expected to take the field with hundreds of data-derived scenarios in their heads. When game analysis software was first introduced, New Zealand rugby employed a dozen analysts who worked literally the entire weekend to provide coaches with about 50 important variables to explain results. They quickly realised the futility, and soon managed to simplify to four variables that took one person a day to provide. That's the real secret to finding value from data - simplicity.