Fortunately, there is an objective means to resolve this dispute: churn.
In sport, churn provides a straightforward measure of the uncertainty of outcome. Churn is simply the average difference between the relative rankings of the competitors at two different measurement points. One can measure the churn at an individual race by comparing finishing positions to grid positions; one can measure the churn from one race to another within a season by comparing the finishing positions in each race; and one can measure the inter-seasonal churn by comparing the championship positions from one year to another.
The latter measure provides an objective means of tracking the level of seasonal uncertainty in Formula 1, and F1 Data Junkie Tony Hirst has recently compiled precisely these statistics, for both the drivers' championship and the constructors' championship, (see figures below). In each case, Hirst compiled the churn and the 'adjusted churn'. The latter is the better measure because it normalises the statistics using the maximum possible value of the churn in each year. The maximum can change as the number of competitors changes.
The results for the drivers' championship indicates that churn peaked in 1980. Given that the interest of many, if not most spectators, is dominated by the outcome of the drivers championship, this suggests that Formula 1 peaked circa 1980.
The results for the manufacturers' championship are slightly different, suggesting that uncertainty peaked in the late 1960s, (although the best-fit line peaks in the middle 1970s).
One could, of course, make the alternative proposal that the churn within individual races is more important to spectators' interest, but at the very least we now have an objective statistical measure which provides good reason for believing that Formula 1 was better in the 1970s and early 1980s.
Hi Gordon
ReplyDeleteLooking at race churn is on my to do list, but I wanted to also think about it in the context of streakiness (eg in terms of streaks of laps where drivers are in the same position etc), which represents another of the chapters in the Wrangling F1 Data With R book ;-)
I've also been looking at graphical tools for trying to help pick out stories at the driver level with a race - one recent experiment revisits an idea I first started sketching a couple of years ago - battle maps - which are a bit like a cut down race history chart (eg http://blog.ouseful.info/2015/01/31/rediscovering-formula-one-race-battlemaps/ ) Howver, these are not statistical analysis/modeling tools, they're far more exploratory, to support people looking for narratives hidden in the race data.
Excellent, will have a look-see.
ReplyDeleteBy the way, which year had the highest churn in the manufacturer's championship? I can't quite see from the graph.