Market makers and scoring rules
By Rob Knight on November 5th, 2008Prediction markets work by trading. Prices are determined by what individuals trading in the marketplace are willing to pay. But what if there’s only a small number of people in the marketplace? What if nobody is buying, or nobody is selling?
This is the problem of liquidity, a term which is probably all-too-familiar in the age of the credit crunch. Liquidity in this context refers to the ability of traders to find buyers or sellers for their shares. If these cannot be found, we say that there is no liquidity in the market. This is as bad for prediction markets as it is for other kinds of markets. If the price is stuck at 73 and you want to buy for anything up to 78 but there’s just nobody selling, it’s bad for the market as it means that the knowledge encoded in your willingness to pay a higher price isn’t captured by the market.
Fortunately, the problem is a lot easier to solve in prediction markets; we make use of a market maker. A market maker is a trader who will always buy or sell at a certain price, maintaining a minimum level of supply or demand into the market. If you really want to buy and nobody is selling, you can buy from the market maker - although you can expect the price to rise as a result. In prediction markets, market makers are automated programs which perform a simple calculation to determine what price to buy or sell at, and always offer that choice to the traders. As more trades occur, the calculation changes and the price rises or falls. This ensures that the predictive value of the price is retained, as it continues to reflect the opinions of the active traders.
Market makers are particularly important for small, thinly-traded markets where the number of participants may be small. Without market makers, many smaller prediction market sites would not be able to function. To get Wisdom Hive started, we’re going to be using a market maker to ensure that there’s always a possibility of trading.
There are a number of approaches that market maker software can take, but the general consensus amongst the prediction market community is that Robin Hanson’s Logarithmic Market Scoring Rule (LMSR) is the most suitable for general prediction market usage. And that’s what we will be using here too, though we might experiment with a few other ideas along the way. A complete explanation of LMSR will have to wait for another day, although we hope to be making some implementations of LMSR in different programming languages available in the near future. Watch this space!
Tags: market makers, robin hanson