Until recently I rarely looked at the betting forecasts published alongside the racecards by the various racing services. Since most of my betting is system-driven, the betting forecast played no part in the horse selection process and hence was of little interest.
Even when I used conventional race analysis methods to find additional bets the betting forecast had no impact on the process. My adopted approach, one that I used for many years when I had usable Bookmakers’ accounts, was to analyse the form in depth in an attempt to isolate one or, at most, two horses which I thought would be in the frame at the finish.
During this process I would get a feel for the price I was prepared to accept, there was no complex mathematical formulae or computer-aided forecasting techniques, just an instinctive gut feeling possibly developed over the years from analysing hundreds of races and making many poor decisions along the way.
Consequently these prices may not have accurately reflected the runners’ chances of success, nor be reflected in the market, but they were the prices at which I would be prepared to strike a bet. Often they were much higher than the prices available, but from time to time I was able to get a satisfactory price, and then, regardless of the race outcome, win or lose, I would be content with the bet.
However I have perhaps been neglectful in this approach: the organisations that produce the betting forecasts expend a great deal of effort trying to get them as accurate as possible, and I know many bettors use them as part of their analysis process. So earlier this year I finally decided to stop ignoring this source of data and undertake some analysis.
The first issue was collecting the forecasts. There appears to be no online database of historical forecasts, at least I couldn’t find one, so I started collecting them manually intending to automate the process should I find anything helpful.
To simplify the collection I focussed on non-handicap races, flat and jumps, and just the first three in the market. I now have a small database of just over 1,500 prices with their associated race details. Admittedly this is a small sample but one that’s worth examining with a view to repeating the process at a later date when the sample has grown to a more significant size.
The betting forecasts collected for flat racing cover both turf and all weather, and overall the loss associated with these horses was -13p/£ at Bookmakers’ starting price which compared to a loss of -2p/£ at exchange prices before commission.
This heavy loss was split -16p/£ for juvenile races and -9p/£ for non-juvenile contests (-4p/£ and +2p/£ at exchange prices), but the losses were roughly equal when analysed by going (i.e. turf or all weather). Interestingly the loss increased across the market positions. So, for all flat races, the forecast favourite lost -6p/£, the second favourite lost -11p/£ and the third favourite a huge -23p/£. The equivalent exchange figures before commission were 0p/£, +3p/£, and -8p/£. Using the exchange place market would not have helped either since the losses were, before commission, -10p/£ for the forecast favourites, and -4p/£ for the others.
Of the 334 horses forecast to be favourite, 239 (71%) headed the market. Backing all of these to a level stake would have shown a welcome profit at Bookmaker’s starting price of +1p/£ which equated to an exchange profit of +8p/£.
The bulk of the profit came from the non-juvenile races, with two-year-old contests showing a loss, possibly due to the reduced amount of form available for these events. Backing all favourites that were forecast to occupy any of the first three positions in the market also made a profit of +1p/£ at Bookmakers’ starting price, and a profit of +7p/£ at exchange prices. Backing the favourite on the place market would not have produced such good figures, however if the favourite was forecast to be second or third listed in the market, the exchange place bets returned a profit of 11p/£ though with only 80 qualifiers the sample is very small making the data far from reliable, but worth noting.
Turning to jumps racing, we find that the overall loss for the first three in the forecast market was -5p/£ (+5p/£ before commission on the exchanges) which compares favourably to the flat racing figure of -13p/£. This lower loss was helped significantly by the non-handicap chases; these produced an overall profit of +11p/£, or +21p/£ on the exchanges. Even the place market was profitable for the chases at +15p/£ compared to -8p/£ for the hurdlers.
From the sample, only 19 chasers forecast as favourite headed the market, but they did make a good profit (+17p/£ at Bookmakers’ starting price and +23p/£ on the exchanges before commission). The 118 hurdle race favourites that were also forecast to be favourite, lost -2p/£ with the Bookmakers.
Taking the complete sample of 1,500 observations and analysing the chance of success as indicated by the forecast price against the actual success rate we find a pretty strong correlation suggesting that the betting forecast compilers are doing a decent job, despite the difficult challenge.
For those prices for which a reasonable number of cases are available the comparison between forecast and actual win rates can be very impressive. For example, a 17% forecast chance of winning produced an actual win rate 18%; the 20% price was reflected exactly in the win rate; and horses given a 29% chance of victory won at a rate of 28%.
Given this degree of correlation it is often thought viable to simply back those horses which beat the forecast price. The logic suggests that, given that these runners’ win rates are reflected by the actual win rate, simply backing those with a higher starting price than that forecast should make a profit.
Unfortunately, it doesn’t work like that. Backing all horses with a higher starting price than the forecast price lost -17p/£ with the Bookmakers, but those starting at a shorter price lost only -5p/£ which converted to a slight profit on the exchanges of +3p/£.
This investigation has certainly proved interesting and I will continue to update the betting forecast database over the coming months and periodically analyse it. Hopefully it will become a useful tool for future system development.