14 Comments
User's avatar
Andy Gardner's avatar

Hence, the popularity of Daily Double and Trifecta bets in horse racing. The challenge in other markets is finding the other side (and margining requirements) for an arbitrary combination bet. Hence the illiquidity and wide & opaque spreads in Exotic Options.

Expand full comment
Kevin's avatar

Why would the liquidity be high for combinations? Wouldn’t you need someone else to be taking the opposite bet, and thus liquidity by default would be low for obscure combinations?

Expand full comment
Robin Hanson's avatar

That's the magic of the combo market maker; you should be impressed.

Expand full comment
Kevin's avatar

I would be impressed if I understood how it worked. I'm imagining you have a zillion people betting for each of event A, not A, B, and not B. Now I show up and I want to bet on the combo market, say, "A and B" and "not A and not B". But I can't because there's no bets that would counteract mine. How can a combo market maker solve that problem?

Expand full comment
Robin Hanson's avatar

The market maker doesn't just match offers that people make, it has its own offers on everything.

Expand full comment
Kevin's avatar

How? What if I'm really accurate about these uncorrelated bets and I win them all. Where does the money come from? Like in this specific example of liquid markets on A and B, and I show up wanting to bet that A and B will resolve the same way.

Expand full comment
Arqiduka's avatar

In your example, am I moving the market for C by 101 dollars even though I only invested 99 dollars? I can see this being prone to severe manipulation.

Easier for the market to invest the funds and pay out to the winner.

Expand full comment
Eric Falkenstein's avatar

I don't follow. Does the long-shot bettor on B know that he is betting on B|~A? Does someone else provide collateral for B if the favorite bettor loses his $100 on A?

Expand full comment
Robin Hanson's avatar

All possible bets are supported by a combo market.

Expand full comment
Eric Falkenstein's avatar

Are those in your example betting explicitly on B and ~A in the second bet?

Expand full comment
Robin Hanson's avatar

I'm not assuming they do. All of the four combos of A and B would likely be bet on.

Expand full comment
Eric Falkenstein's avatar

Are you assuming longshot bettors consider betting on A and B almost perfect substitutes for "A and ~B" and "B and ~A", given their probabilities are adjusted downward only by 1%?

Expand full comment
Robin Hanson's avatar

no

Expand full comment
Eric Falkenstein's avatar

OK, moving on. But I don't understand

Expand full comment