Published on Feb. 3, 2020, 4:58 p.m. - 3 Comments - Sports: Horse Racing
How often does the favourite win a horse race? And is it profitable to back the favourite in each race? With a simple backtest we investigate this simple approach on historical data.
When observing betting markets you might have the impression that favourites win in most cases. Your conclusion might be that it makes sense to back favourites. In this backtest I would like to analyse how profitable it is to bet on favourites in UK horse races.
This is a very simple, hypothetical strategy where the favourite in every UK horse race is backed at the Betfair Starting Price (BSP). The staking plan is a flat, simple 1 point risk per selection. Hence the maximum loss per selection is 1 unit.
First I would like to clarify what I mean with favourite: With favourite in a horse race I refer to the selection in a horse racing market with the lowest starting price odds. Lowest odds means highest probability of winning. The problem here is that you do not know who is the starting price favourite before the starting price is calculated. Imagine that there are two horses in a race with very similar odds just a moment before the off. You don't know which one will be the starting price favourite as the odds might still drift before the off.
In the backtest I can easily select the horse with the lowest starting price odds which will introduce a look-ahead bias. I will ignore this bias in the backtest since I am assuming that this bias is a random, non-systematic bias. If I was to deploy the strategy as a betting bot I would probably check the odds one minute before the off and then take the selection with lowest odds as favourite which might differ slightly from the actual starting price favourite.
The issue with this strategy is that prior to placing a BSP bet, we do not know the odds, hence we cannot be sure who the favorite is according to the odds. There might be races where two horses have very similar odds prior to the beginning of the race. However, for the backtest we will simply pick the selection with the lowest odds as our favorite.
Historical data with Betfair Starting Prices for UK horse races are available on promo.betfair.com in csv format. With a simple Python script it can be imported into a SQL database for further analysis.
Once the data was imported into a SQL database, a backtest can be carried out using the Betfair Starting Price. The favourite is selected in a subquery having the minimum starting price. The profit is starting price minus one for the stake in case the favourite wins and minus one in case the favourite loses. For the backtest the following SQL query is used:
SELECT *, CASE WHEN win_lose THEN (bsp - 1.0) ELSE -1 END as profit FROM data WHERE bsp = (SELECT min(bsp) FROM data as d WHERE d.event_id=data.event_id)
The cumulative profit and loss as well as the backtesting results are the following:
On a first glance the result is quite positive. However it is not the whole picture since we did not account for any commission.
In order to obtain a more realistic backtesting result, we need to take into account the betting exchange commission. For this scenario I will assume a commission level of 2% on profits. In the SQL query a factor of (1-0.02) is included to calculate the profit in case of a winner:
SELECT *, CASE WHEN win_lose THEN (bsp - 1.0)*(1-0.02) ELSE -1 END as profit FROM data WHERE bsp = (SELECT min(bsp) FROM data as d WHERE d.event_id=data.event_id)
This yields the following result:
Following this backtest it becomes clear that such simple approach is not profitable: The odds of the favourite in UK horse races are too low, there is no edge to be exploited. With a realistic commission of 2% on profits deploying this strategy as a betting bot would constantly generate losses. Also the opposite approach "Laying the Favourite" is not successful either as illustrated in a separate backtest.
So what can you take away from this backtest? Obviously that it is important to take into account commission and fees which can have a significant impact on betting strategies and profit and loss figures.
We would be curious to hear your feedback on this backtest. Do you use a similar approach, maybe using some filter? What are your thoughts on this? Please tell us in the comments!
Please log in to leave a comment.
If you would like to learn more about this strategy, please do not hesitate to contact us.Contact Us!