Winning Systems: Flat Racing Drifters

    In his latest update, Dr Peter May looks at horses in Flat races that drift in the market and whether they offer a path to profit for statistically-minded punters.

Back in 1985 I remember waiting in a Bookmakers’ shop in Oxford for the first show for a race featuring Nicky Henderson’s top class chaser, Classified. 

Those of you over a certain age will remember this Classified, as opposed to the more recent one, and his battles with the likes of Badsworth Boy, The Tsarevich and Beau Ranger, plus his excellent third to West Tip in the Grand National.

When the price came through and 11/2 was offered I quickly placed my bet. A few minutes later Classified was trading at 13/2. The punter standing next to me knew about my bet and asked whether I was going to back Classified again. I was unsure why he had asked this but his explanation made perfect sense: If you thought the horse was a good bet at 11/2 then it’s an even better bet at 13/2. It’s difficult to fault this logic, but does the data back up his theory?

 In order to test the theory that drifters are good bets it is first necessary to define precisely what is meant by a “drifter”.  For the purposes of this article I have defined a drifter to be a horse with an exchange price at the official race time which is longer than the price it traded at mid-morning after any adjustments for subsequently withdrawn runners were made.

Naturally there are many other definitions which could be applied, for instance the time frame can be reduced to the last 10 minutes of trading before the off when there is the greatest amount of betting action, or the price threshold could be increased so a runner only qualifies if the price exceeds the early comparative price by 10% or 20%, for example.

An examination of 100,000 race performances on the flat shows that, under this simple definition of a drifter, 60% of runners lengthened in price, and 40% either remained stable or shortened. The larger number of drifters is partly due to the fact that the morning over-rounds tend to be higher than those at off time and prices for the very long shots tend to be more cramped in the early stages of market development.

The handicap/non-handicap split is informative: In non-handicaps the proportion of horses which drift in price increases to 63%, which could be due to the fact that many of the runners in these races are less exposed in terms of form and the layers are not prepared to take too many risks when the market is being established.

Some punters believe that a drift in price is a good indicator that the horse is not going to win, so how do the win rates for these two groups of horses compare? From Table 1 it can be seen that there is a significant difference between the two sets, in fact this difference is much greater than I had expected.

Table 1: Analysis Of Win Rates

Race Classification

Drifted

Others

All

Turf Flat Handicaps

7%

13%

10%

AW  Flat Handicaps

8%

15%

10%

Turf Flat Non Handicaps

9%

16%

12%

AW Flat Non Handicaps

9%

20%

13%

For non-handicap races on turf, horses which have not drifted in price win at a rate that is 2.6 times higher than the drifters. As remarkable as these statistics may seem to be, they will not make us rich, for that we need to investigate the profit and loss data.

Table 2 shows the average profit for horses backed at the exchange price available at the official race time for the four main race categories. For handicap races the data follows a similar pattern to the win rate figures with drifters performing poorly compared to the remainder. However the two return figures are equal for all weather non-handicaps at a very high -11% and for turf non-handicaps the drifters preformed slight better.

Table 2: Analysis Of Profit/£

Race Classification

Drifted

Others

All

Turf Flat Handicaps

-0.05

0.01

-0.02

AW  Flat Handicaps

-0.07

-0.03

-0.05

Turf Flat Non Handicaps

-0.02

-0.09

-0.02

AW Flat Non Handicaps

-0.11

-0.11

-0.11

The fact that non-handicaps seem to buck the trend is probably due to the depth of form available for the runners which, in general, is not as comprehensive as it is for handicap races and more guesswork is used to set the early prices. It is important to remember that a lengthening price does not always mean the horse is going to be a non-trier, or that connections have decided that today is not going to be his day, it is far more likely that the horse was simply underpriced and the market is correcting for this.

The most interesting race type within the non-handicap category is selling races. From the 730 performances in this lowly race classification, 472 were associated with drifters and these lost 22p/£ on average; the rest made a profit of 9p/£ at a win rate of 24% over three times that achieved by the drifters which is more in line with expectations drawn from the handicap data.

Examining a few key factors we find that horses making a quick return to the track (within five days of the last run) return a 2p/£ loss in all turf flat handicaps covered by this sample, but those which contracted in price boast a profit of 11p/£ which is entirely attributable to all age contests; races restricted to just two-year-olds or just three-year-olds show a loss which again could be due to reduced levels of form. In fact for these two age-restricted race classifications the drifters made decent profits though the sample size was small with less than 200 observations. For all weather handicaps the pattern is pretty much repeated with quick returners that have shortened in the market making 22p/£ profit (669 observations and 31p/£ for all age contests) and the drifters just breaking even.

We already know that horses which shortened in turf flat handicaps returned a 1p/£ profit for the sample analysed (Table 1), but it is more interesting to see how this 1% profit was distributed across the last run finishing positions. Runners which were following on from a victory lost 1p/£ (2238 runs) which is the same as for the drifters, but horses which filled the runner-up spot on their latest outing made 6p/£ (2142 runs) with drifters losing 12p/£.

To limit the number of finishing positions to a manageable number I use “10” for all runners which failed to make the first nine home (therefore this group also includes those which failed to complete) the majority of which are shown as “0” in recent form lines. The sample included 2698 of these which also contracted in price, and on average they made a profit of 9p/£. However the majority of the profit was achieved by horses that were priced at 40.0 or higher in the morning so getting a bet on these would not be easy.

For turf non-handicaps the best horses to follow were those which ran second or third last time out and shortened (13p/£ from 1823 runs).

The best way to profit from this information is to identify which horses are going to shorten and then back them early. For instance, in turf flat handicaps horses which shorten make 1p/£ at off time, but backing these early would have made 27p/£. Naturally it is not possible to precisely determine early in the day all runners will contract in price, but it is something that race modellers attempt to do by using an odds line.

Once established for a race, the odds line can be compared to the prices available early in the day and a threshold set which is designed to capture as many horses that shorten as possible. The threshold can take the form of a ratio of the two prices, for instance: the price divided by the odds line price must exceed 1.2. I tested a method using this approach over the summer of 2014 focussing on turf flat handicaps. From 7 July to the last day of the turf season I generated qualifiers each day at 10-30am and backed them on an exchange. 

In total there were 481 bets, slightly lower than I expected and indicating that I had possibly set the threshold too high, however they did make a reasonable profit pretty much in line with expectations. The drawback is that stakes are limited at that time of the day, and for punters who prefer to bet in thousands it would be necessary to apply the technique much closer to the off time. However this method will now be added to my betting systems for 2015.

From the data presented here it can be seen that drifters do win but at a much lower rate than other runners and following them in flat handicaps cannot be recommended, in this instance the market does tend to get it right.

  • Race Type
  • Location
  • Date Range
More options
  • Race Class
  • Distance
  • Official Going
  • Race Types
×
×