Survivorship Bias in a Tipster Sports Betting Market
Posted 21st February 2017
Just a short little article from some 'back of the envelope' thought experimenting.
I've run a little model to test the problem of disappearing tipsters at tipster supermarkets. The model went like this:
- Assume all tipsters bet once per day (odds 2.00).
- Assume on day 1 there is 1 tipster & every day a new tipster joins.
- Assume that every tipster will give up (and be removed from analysis view by the supermarket) at any time after he's had 100 tips when his yield drops below -5%.
- Assume all tipsters are tossing fair coins and have no edge at all.
What does the supermarket look like after 200 days?
From a Monte Carlo simulation (1,000 runs) I get the following figures:
- Average number of remaining tipsters = 56
- Average aggregate yield from remaining tipsters = 5.4%
Conclusion: Is it any wonder that tipster supermarkets give a false impression about forecasting ability, given their complete disregard for survivorship bias?
You can read more about Survivorship Bias in my Pinnacle Betting Resourcs Article. Plus lots more discussion about this and other biases that have an influence on betting in my latest book.