Friday, 5 July 2019

The AFL though the lens of Cherin-Gordon

To an Australian encountering US team sports, and soccer in Europe, it would no doubt feel surprising to discover how many games are played in Major League Baseball, the National Hockey League, and the National Basketball Association. Whereas no major sport in Australia plays more than 25 games per season, those three leagues play 82 games (NBA and NHL) or 162 games (MLB). European soccer leagues play between 30 and 46 games per season.

In his Untangling Skill and Luck from 2012, Michael Maubossin noted that a 50 percent ratio of skill to luck can be achieved by:
  • 12 NFL games
  • 69 MLB games
  • 15 NBA games
For ice hockey and soccer, estimates are extremely variable, ranging from 35 to 75 games for ice hockey and 25 to 50 for soccer.

Last year – though I discovered this only today – journalist Simon Cherin-Gordon offered a further insight into the difference in season length between gridiron (NFL) and the other major leagues in North America. In an article titled ‘Why Are We Playing 82 Games, Anyway’ – written to advocate a shorter NBA season, which Maubossin had shown to be more practicable than any other US or European team sport – Cherin-Gordon said:
“Much of this has to do with the inherent nature of each sport. A baseball star will only bat four or five times in an average game, or pitch one out of every five days. Runs are hard to come by, and lower scores lead to higher variance and make a short season untenable (the same can be said for the NHL, which has an 82-game season just like the NBA).” 
“[Gridiron] Football, meanwhile, is a sport where the best players are involved in somewhere between 60 and 70 snaps per game (which constitutes roughly half of the game’s plays). Its scoring is less random, and the best team usually wins. It is also as violent a sport as there is, thus frequent games and a long season are out of the question” 
“While the NBA season falls just in between that of the MLB and NFL in terms of length, it has far more inherent similarities to [gridiron] football than it does to baseball.” 
“If you think 60–70 plays per game is a lot for a
[gridiron] football player, try more than doubling that. Basketball players play both offense and defense, and NBA games generally feature around 200 possessions. The best players play between two-thirds and three-quarters of those possessions (32-36 minutes).” 
“That’s just one of many ways the NBA takes out variance. It is the only sport where every successful play is quantified on the scoreboard. There are no base hits or first downs. When you do score in other sports, it is a more random occurrence with more impactful ramifications. Crossing home plate or the goal line one time is worth exponentially more than putting the ball in the hoop. There are also less [fewer] players involved, and less [fewer] quirky advantages to playing at home. The best team in an NBA game wins more often than they do in any other sport.”
What strikes me about this list is that if we apply what Cherin-Gordon says to the AFL, we have the following comparison:
Sport “Possessions” per game % played by best players “Possessions” played by best players
≅240 (pitches)
≤⅙ (16.67 percent)
≤40 (pitches)
≅140 (downs)
≅½ (50 percent)
≅70 (downs)
≅¾ (75 percent)
(Australian rules) football
≅600-700 (total possessions)
≥⅞ (87.5 percent)
This table does not measure how many team possessions there are per football game. However, in basketball the rules make teams to alternate possession after every score, which gives weaker teams more of a chance than in football, where a team can extremely easily score and then regain possession without one opposition possession. (In one game in 1954, it is known Richmond kicked five goals without a Melbourne player touching the ball, and as the Demons actually won that game this is unlikely to be the most extreme possible case.) Football is also analogous to basketball in that every “successful” play is rewarded on the scoreboard – although if a team kicks out of bounds on the bounce or a pack forms that team can regain possession without an opponent necessarily touching the ball.

The table above suggests that, although the AFL season is much shorter than those of baseball, ice hockey or soccer, it is not proportionately so relative to the requisite length for a passable skill-to-luck ratio. In fact, if in Cherin-Gordon’s words,
“MLB needs 162 [games], the NHL needs 82 [games]... the NFL needs 16 [games]”
then the AFL would “need” no more than 10, or maximally 12, games per season.

However, although Australian football has exceedingly low internal variance via randomness in player performance, external variance from variation in weather and ground conditions was highly significant before the climatic “magic gate” of 1997/1998 and the closure of Waverley Park and the AFL’s old suburban grounds. Subsequently this external variance has been effectively eliminated by:
  1. rapid poleward expansion of the subtropical arid belt, totalling 800 kilometres since 1964
    • this has:
      1. reduced the incidence of wet weather and increased evaporative drying of grounds when it does rain
      2. also reduced the possibility of extremely windy conditions that favour shorter players
  2. artificial drying of grounds during rare cases of wet weather, and improved drainage
  3. the opening of closed-roof Docklands Stadium where weather conditions are consistently dry
This history of significant external variance undoubtedly explains why the AFL has a longer season than required for a passable skill-to-luck ratio. In addition, on softer surfaces football was less physically taxing than the present-day game, as seen by the fact that players sometimes played mid-week Foster’s Cup matches during the season. More that that:

  1. even a minor reduction in games would be extremely unpopular with football fans, as shown in the 20-game 1993 season
  2. a reduction to 10 or 12 games would require a radical restructuring of the AFL’s season
  3. many pairs of teams would – as in the NFL between 1978 and 2001 – go up to twenty seasons without opposing each other
  4. it is highly plausible that the game would become even more vigorous and physically demanding with only half as many games, so that the reduction in injuries would be much less than Cherin-Gordon supposes (this also would apply to the NBA)
Most likely, only major increases in severe – likely only in career-ending – injuries would cause the AFL to reduce the number of games on its schedule. This plausible in an even hotter and drier climate than observed today thanks to Australian and Gulf States greenhouse pollution, but even if judged inevitable it could create a vicious circle of harder, more anaerobic play and more injuries, because if AFL players knew they had only 10 or 12 home-and-away games to play, they would almost certainly play harder than they do now knowing they have to get through 22.