Rowleyfile Investigates: Seasonal Unpredictability In Racing Results

An old racing saying goes that “Autumn rides in on the tail of the last horse in the St Leger”. One of racing’s received wisdoms is also that results become much more unpredictable as the ground turns at this stage of the year.

With the world’s oldest classic due to take place at Doncaster on Saturday, it is a fair bet that many racing enthusiasts will start to turn their attention away from Flat turf racing before long.

Not so fast.

I decided to look at the level of predictability in Flat turf racing results in different months of the year, viewed in conjunction with the average state of the going and the variability in that going.

The message is that the middle of the season is, indeed, the most predictable time of year, and the time of year at which going varies the least. But the difference in this predictability is small, and, for what it is worth, September is an average month.

The effect may well be overstated, in other words: see what you think.

First, a few details of methodology.

The sample covers UK Flat racing from 2009 to 2013 inclusive.

I took Timeform’s description of the state of the going as my guide – this is superior to the official version for a number of reasons, dealt with here  – and I digitalised these descriptions to acknowledge that going categories are non-linear, in that, for instance, “soft” covers a wider range than “good to firm”.

This is how the going descriptions were recoded, in line with observation: Firm = 1.0; Good to Firm = 1.6; Good = 2.4; Good to Soft = 3.4; Soft = 4.8; and Heavy = 6.8.

I also considered handicaps only, measuring predictability as the degree to which the Betfair Starting Prices of winners were more or less expected than by chance given field size. A 5.0 BSP winner of a 10-runner handicap would be perceived to have a 20% chance of winning, when 10% would be random, and therefore gets a 2.0 “Predictability Index”. The higher the figure, the more predictable was the result.

Against each month is shown: the average going according to the above methodology (lower is firmer); the variance of going, as measured by Standard Deviation (lower indicates less variance); the predictability of results in handicaps, as explained above (higher is more predictable) and what that would mean in terms of average odds for a winner in a standardised 10-runner race.

There then follows the rank of these months for Going Variance (GV) and Predictability (Pred), with lower numbers indicating higher variance and greater unpredictability.

The bookend months of March and November have been left out, but are given for information. Sample sizes are small and potentially misleading, though, interestingly, they are the two most predictable months in terms of standardised BSP in handicaps.

October has the greatest variance in going (as well as the softest average going if you ignore November), and it is ranked second in terms of unpredictability. Similarly, April has the second-most variance in going and is the most unpredictable month.

At the other end of the scale, June is the second-least variable month in terms of going and is the most predictable month in terms of results (if you ignore March and November).

Outside that, however, months are much of a muchness.

September, the month in which this is written, and in which punters’ thoughts may start turning to the beauty of the winter game (as I like to refer to all-weather racing), is ranked third of seven in terms of variance and fourth of seven in terms of predictability of results.

As can be seen by comparison with the totals at the bottom of the table, September is a remarkably “ordinary” month, judged by these measures. And, I hope you would agree, the difference between the most predictable and least predictable months – equivalent to the difference between standardised winners averaging 6.3 BSP and 6.7 BSP – is in any case far from large.

There may be perfectly good reasons for punters to start drifting away from Flat turf racing at this time of the year, but implicit unpredictability should not really be one of them. Or not just yet, anyway.

New Timeform customers get £10 FREE downloads! Click HERE and simply enter code TFF10 when you register.