Thursday, March 13, 2008

Is metallic hydrogen a black swan?

I have set up a prediction market for superconductor temperature here. One problem with this bet is black swan events

When betting on the price of gold or the score in a football match you know it is very unlikely to move outside certain ranges. Scientific knowledge does occasionally make huge advances though.

To take one example say someone creates metallic hydrogen at reasonable temperatures. This substance is estimated to have a superconducting temperature at near room temperature. Well 230 85 K but the upper end of this is at the level to completely change the world (with the huge assumption that the pressure metallic hydrogen existed at is practical).

In this case the prediction for the superconductor temperature is in two forms
1. Something similar to the current rate of change happens. A few degrees a year.
2. Something completely new comes along with really different properties (figure 2 here).

Metallic Hydrogen would not even be that novel. It has been modeled theoretically and we know people are trying to make it. Usually in prediction markets you have scoring rules that give out money/points based on how close to the answer you were. In these cases where huge changes have a reasonable chance of occurring these sorts of scoring rules might be unfair.

Say there is a 90% chance that research will continue along its standard path to 188K. Say there is a 5% chance that nothing will change. And say there is a 5% chance the maximum temperature will suddenly jump up to 270K. Does this mean that the market will sit at 193K? Will people in the lab who create a room temperature superconductor go out and bet heavily on their side moving the market (and making them money)? Will people ignore the black swan of room temperature superconductor and move the market to a “sensible” level? Humans do have a bias to ignore unlikely highly risky possibilities.

Could this bias make scientific prediction markets more profitable for those that create huge innovations?

No comments: