Stugots, I have checked the accidents that were in Unitab's top 4 or equal top four rankings/selections.

There were 14 such accidents from an overall total of 31 accidents.

However, I don't know how many smiles there were from Unitab's top four rankings or their results (it would have taken ages to do so), so I can't say with any certainty if it is better to apply your suggestion as a filter.

But because almost 50% of the accidents were in the top four Unitab rankings I therefore have to assume it would have been better to omit them from the method. The resulting "smiles" strike rate may have been close to 99%, and so if I/anyone ever decides to apply some sort of stakings/recovery plan it might be less dangerous.

From today your suggestion is a filter.

If I get time I will look at betting Unitab's top four rankings.