Robustness in the scientific world involves some very heavy high-level mathematics, well beyond the understanding of the ordinary man in the street. For the purposes of our comparative tests, however, we have kept the issue as simple as possible. After all, what it really boils down to is this: How would you decide which Selection System you should rely upon the most if the first had a success rate of 16 out of 20 bets, while the second had a success rate of 60 out of 80 bets?
The purpose of the Robustness Test is to see if you should trust the Selection System sufficiently to follow its advice completely, not just for the next bet but for the remainder of the season. We feel that if you can't rely upon a Selection System for the remainder of the season, you shouldn't consider trusting it for even one bet. So what we have done is to concoct the following simplified formula that can be used to test the robustness (reliability) of anyone's Selection System when the number of bets placed does not match your personal "comfort level":
We acknowledge that the idea of the "comfort level" is rather novel, and that it will therefore need further explanation.
Consider this: If you are betting on "singles" match results, where the odds are low because you are betting on the favourites to win, you would like to know that you stood a good chance of being correct with at least 70% of your selections. This is because the returns are generally so low that if you suffered any more losses, then your winnings on say 65% of the matches would not cover the losses on the other 35%.
The tables we present under the section entitled Robustness Inputs and Outputs (covering each of the different possible betting scenarios) show what we ourselves need to see as the answers to the Robustness Test before we would consider putting faith in our own Selection System.
If, for any given type of betting scenario, the Selection System you are using can't give you the minimum ratio we have quoted in the relevant table, then you won't make money on that type of bet.
The above Robustness Test Formula is meant simply as a tool to indicate the comparative reliabilities of two separate success rates where the number of bets placed was very different. So please don't get bent out of shape trying to analyse its content/formulation! In mathematical terms it is a crude mechanism, we agree, but for the desired purposes it is reasonably effective, in that it will very easily enable you to see when a claimed success rate is too weak for you to rely upon.
For example, applying the Robustness Test Formula to our two above examples (16 out of 20 (which seems to indicate an 80% success rate) versus 60 out of 80 (indicating a 75% success rate)), if the "comfort level" we want is 100, then the "robustness test" answers would be 57.69% and 67.92% respectively. This shows that the first reliability is therefore not as strong as it at first appeared (because it was tested against only 1/5 of the desired number of bets to be placed). On the other hand, the second success rate was tested against 4/5 of the "comfort level" number and therefore holds up well under test. If the "comfort level" were to be reduced to 90, then the answers would be adjusted to 57.95% and 70.96% respectively, so the 20 out of 80 successes is still seen to be a weak claim. That is not to say that it isn't a genuine claim, so if you decided that a "comfort level" of 20 bets placed is appropriate, then it would immediately get an 80% "robustness" rating! It is entirely up to you as to where you wish to pitch that "comfort level".