What Are The Odds Of Winning The World Cup?

What Are The Odds Of Winning The World Cup?



Just an hour from kick-off against Argentina, the market’s probability of Australia winning the final of the World Cup was fifty percent. Never has my country’s football prowess garnered so much respect!

(To be clear I refer to the conditional probability if they make the final, and this is inferred from the Betfair prices for making the final, and for winning, which trade separately).

Of course, this is absurd.

At the risk of offending my compatriots, there is no plausible model of football reality where, on average, Australia will start the final as the favorite or equal favourite.

Indeed the probability that Australia starts as the favorite in the final is surely less than 5%, even taking into account ability updating (which also applies to their future opponent), the gigantic size of our central defender, or what have you.

Australia has less than a one percent chance of facing Poland, for example, the only lower-regarded team.

This is an example of why design and intent can matter in small markets, depending on what probabilities you intend to infer. In this case we use two independent central limit order books and things seem to go awry. We don’t get the number we need. Even though it should be “spanned” by these markets.

I wish to shrink the benefits of markets down to a nano-scale, where they can infiltrate “AI” in fast-moving operational domains. Lastly, there is no single best way and there are challenges, clearly.

World Cup Economic Impact

But mechanisms generalizing markets can become inspired by statistics. Reinforcement learning. And elsewhere, they don’t need to become limited to CLOBs or other accidents or history.

There is a fascinating relative greenfield existing between market design, scoring rules, market making, and machine learning.

Written by Peter Cotton

Buy Peter’s book:

Microprediction: Building an Open AI Network