With all due respect to your methods though WB. the real probabilities are probably less than the betting odds let out. If for gambling poo sizel A (how many people bet on Driver X), and betting odd B for any particular one,
If A = 1/B, then the expected return would be 0 for the booki.
A has to be greater than 1/B and that would skew things slightly than an outright linear comparison. However if BetFair really does base its odds on betting money then that would probably decrease the skew to as little as possible.