That fixed $2,500 boundary for a longshot bet is simply too arbitrary; it's like saying a car is expensive only if it costs over £25,000, ignoring that for a student, £5,000 is a huge sum, a 5x multiple of their annual savings.
A better approach ranks bets by their relative impact, perhaps using a multiple of the average transaction value in that specific market.
Otherwise, we risk missing about 40% of genuinely significant signals in smaller markets because their absolute value doesn't hit that minimum threshold.
For example, a £500 bet on a prediction market where the average stake is £50 is a 10x deviation and far more telling than a £2,500 bet in a market averaging £10,000, which is only a 0.25x deviation.