Becoming a successful bettor is all about finding expected value. Now, Pinnacle’s Betting Resources already have many articles about expected value, so let me just summarise briefly with an equation:
Expected value=(Bookmaker' s odds/True odds)-1
If the true odds are 2.00 and the bookmaker’s odds are 2.10, the expected value is 0.05, or 5%. If the true odds are 4.00 and the bookmaker’s odds are 3.50, the expected value is -0.125, or -12.5%. Serious bettors are only interested if the expected value is greater than 0%, or when the bookmaker offers odds that are longer than the true odds.
Of course, knowing what the true odds are is another matter altogether. Sports betting is not like dice or roulette. The markets are complex systems, for which there are no simple algorithms that will tell you what the true outcome probabilities are. We have to estimate them via data modelling, and perhaps a bit of intuition as well. The better your model, the closer your estimated probabilities will be to the true probabilities.
Pinnacle’s model, a combination of their own data analysis and the market information they acquire about their customers’ models, means it’s one of the best at estimating true probabilities. Of course, it won’t always be right, but on average it’s surprisingly good. For example, about 50% of their prices of 2.00 (after you’ve removed their margin) win; and 25% of their prices of 4.00 win. Individually, it’s impossible to know which odds were closer to the true odds and which were not so close, but on average, the errors are broadly cancelled out.
Because the last odds before a football match kicks off contain more information about the match than the odds when Pinnacle first published them, on average the last odds are closer to the true odds than the first ones. We might propose that the amount by which the odds move provides a measure of how much expected value there was available in the first odds. This is not to argue that the last odds are always correct, nor that the first odds are always incorrect. Rather, that the ratio of the two can be used as a proxy measure of expected value, given that we can’t know what the true odds are. Thus:
Expected value≈(First odds/Last odds)-1
Sometimes both the first odds and the last odds will be shorter than the true odds; sometimes they’ll both be longer; sometimes the first will be longer than the true odds and the last shorter; sometimes that will be the other way round. However, on average the ratio of the two should give you a good idea of how much expected value existed in the first odds. With this assumption in mind, how much expected value exists in the football match betting market? For a sample of 158,092 football matches and 474,276 home-draw-away betting odds, I divided Pinnacle’s first odds by their last odds, having removed their margin from the last odds, and subtracted one. How many odds had expected value greater than 0%? The figure was 29.7%. That’s actually quite a surprise. Nearly a third of Pinnacle’s first odds actually hold some expected value. We might expect there to be more occurrences of low expected value (for example 2%) than high expected value (for example 20%). But how much more? I ranked the size of the expected value for each bet in ascending order, then calculated a cumulative percentage. For example, 29.7% of odds held value greater than 0%. This has fallen to 21.7% for expected value greater than 2%. By the time we reach the threshold of 10%, only 6.6% of odds will hold such value, on average, and for 20%, it’s just 1.8%. I’ve drawn the trend in cumulative probability:
Evidently, there is a lot of expected value to be found in the football match betting market. However, don’t expect a lot of big gifts. The availability of increasing expected value falls exponentially; the greater the expected value, the far less of it you will find, assuming you even know where to look. And in this respect, of course, the much bigger problem for bettors is knowing the expected value is there in the first place. Using this proxy measure, you won’t know that it was there until Pinnacle have published their last odds and by then, the first odds have gone. Nevertheless, if you do have a forecast model that works, which can estimate the true odds well, you now know that when Pinnacle publish their first odds, there is, in theory, a lot of potential for making a substantial profit.