Seeding and Fairness

Is seeding fair? This would seem to be a fundamental question, but it’s one that doesn’t have a straightforward answer. Before further interrogating the data from the simulation of the Western and Southern, with its odd seeding style, for its effect on fairness, it may be well to revisit the question of just how fairness and seeding relate more generally.

In a nutshell, seeding usually enhances the meritocratic values embodied in fairness (C) at the expense of the egalitarian values embodied in fairness (B). But the decision whether to seed is not simply a way of expressing a preference for fairness (C) over fairness (B). The fairness effects of seeding may be incidental to some other goal.

In any professional sport, it’s likely that that other goal is spectacle. Tournaments are seeded to enhance the number or the quality of matches that generate revenue through the sale of tickets or broadcast rights. Tournament organizers like for a tournament to have a narrative arc to it. The final game, it’s hoped, will be the long-awaited showdown between the two favorites. So it’s important that they not meet each other too early – if the great showdown happens in the first round, everything else may seen anticlimactic.

But seeding can be used for other mercenary purposes as well. The National Football League condones a sort of reverse seeding, where the leading teams from last year’s play are more likely, rather than less likely, to meet each other before the playoffs. This sort of seeding reverses the usual relationship between seeding and fairness – fairness (B) increases at the expense of fairness (C).

For the usual fairness relationship to hold, the seeding has to happen in a way that gives an advantage to better teams, and there’s usually some element of this in the scheme. But seeding can also be used to reward entrants for some other behavior that’s deemed to be in the best interest of the organizers.

An interesting example of this used to occur in Major League Baseball in connection with the mid-season all star game. After a particularly dreary tied all star game in 2002, the league decided it needed to add interest to what is otherwise simply an exhibition game by making it count for something. The league winning the all star game would get the home-field advantage, essentially a favorable seeding, in the World Series later that season.

Starting this year, however, the practice was changed so that the home-field advantage in the world series goes to the team with the best regular-season record. It was apparently felt that it was inappropriate to seed the world series according to the results of an exhibition, and so a stronger fairness (C) enhancer was used.

The ATP points used to seed tennis tournaments like the Western and Southern are partly a measure of player skill, and so their effect on the tourneys is partly an enhancement of fairness (C). But the ATP points scheme also represents the association’s efforts to enhance revenue by increasing their inventory of salable tennis competition. ATP points don’t just reflect merit. They also reward players for playing a lot.

A similar use of points is found in backgammon. The American Backgammon Tour awards accumulating points each year, and while the players at the top of the ABT points table tend to be good players, no one would seriously contend that the ABT points table represents a real skill ranking. Something analogous often happens at the local club level, where players accumulate points for success in their weekly club meetings. Again, the players at the top of the club ranking table tend to be good players, but they’re also players who attend regularly.

There’s often an imperfect relationship between actual skill and the points schemes used to seed tournaments. If the connection is weak enough, we may find that the seeding applied impairs fairness (B) without doing much to enhance fairness (C). In the next post, then, I’ll use the Western and Southern simulation to consider whether this is happening in professional tennis.