TLDR
Markets are:
- Better than models for short horizon prediction, but
- Harder to wield, until now
Hayek meets ML, if you will.
1 Efficacy of existing markets If you are not remotely sympathetic to the first assertion, then space is too limited here but by all means join this discussion, or better, this one where I revealed my hijacking of the M6 Forecasting contest using a market. My contribution to the age old question of market efficacy (in the specific context of realtime operational intelligence) is also the subject of my book which you are welcome to listen to, at least in part.
2: Convenience and efficacy of a market-like thing you can create right now as easily as a model Could you create a small market and actually expect that market to be a powerful prediction device? Yes. I think it helps to just do it and see. You can publish live data such as your own prediction model errors. This will create a streamlined cousin of a market (read more here) where bots and their authors will try to find signal that you missed. There’s even a colab notebook that will help you compare the accuracy of the market you create to any model you believe might be a substitute. The only real question is whether you are as quick to grok this as others have been.
The reason: “trade friction” is virtually zero. Participation by humans and inaminate autonomous algorithms is seamless. If you are willing to cut and paste one line on bash command or open a colab notebook and run it, you’ll be on the way to being convinced of that fact, since this will initiate your participation in a stripped-down improvement of an options market (no email or ISDA required). And you might be interested to know that hundreds of algorithms from the Python ecosystem are rated offine and ready to enter the fray. See here how trivial it is to invite them to compete.
The bigger picture
There are other important ideas such as recursion, privacy, and the reformulation of control theory and reinforcement learning. To encourage you to dive into this world I’d ask you to look at what people are saying about the book and then wade deeper into this future vision of bespoke quantitative business optimization, powered by a million copies of what you see here, and improvements on the same.
-+-
Documentation map
Edit this page.
