The right way to implement a multi-outcome prediction market: Linear programming
David Pennock February 19th, 2008
Over on Oddhead Blog, I give a lengthy pitch for linear programming as the “right” way to implement a multi-outcome prediction market.
I argue that the simplest and most common approach — to treat a multi-outcome market as a bunch of independent single-outcome markets — is wrong, even though it’s the approach taken by most prediction markets, bookmakers, and financial exchanges.
I also argue that
- IEM’s implementation is one of the worst
- Intrade’s is slightly better but not much
- Newsfutures’s and Chris Hibbert’s phantom bids approach is even better
- A host of people* hit on the best approach, many well before I did, advocating linear programming as a natural matching engine for multi-outcome markets
| *Including Baron, Bossaerts, Chen, Economides, Fine, Fortnow, Kilian, Lange, Ledyard, Nikolova, Pennock, Peters, So, Wellman, and Ye. |
- All Best Posts Ever , All Guest Authors's Posts , Exchanges & Markets , Explainers , Market Designs , Mechanism Designs
- Comments(4)
I hope to see more linear programming implementations. On a practical note, there are some reasons why a continuous auction is more fit than a call auction, and it might be hard to get a continuous auction going while constrained to not taking on risk. Maximizing liquidity and taking no risk are basically at odds. I am told that the economic derivatives were pulled because one of the banks ended-up taking (insufficiently profitable!) positions to tighten the market. Whatever the mechanism, there will likely be an instance of human price-setting by the market-maker/sponsor, and minimizing the damage, especially in the early stages, is important.
Great points. Agreed. Interesting about the econ derivatives closure.
It should be possible to run the linear program in continuous mode, identifying matches as soon as they arise. This is I believe a strict generalization of the typical continuous double auction, so should have a least as much liquidity as a CDA.
Don’t mind this comment. I’m testing the comment system…
[...] see Hibbert’s “Market Makers for Multi Outcome Markets” and Dave Pennock’s “Right Way to Implement a Multi-Outcome Market.”) If they’ve done something like this, it should make their markets work more efficiently. [...]