Known mispricing opportunities only available for small traders

Question

Warren Buffett has famously said that he could generate 50% annual returns if he was working with small sums of money. (He cannot move the needle enough now with large amounts of capital). Perhaps two reasons could explain this, holding his skill level fixed, 1) there are just more smaller companies and therefore more opportunities available and 2) smaller companies are less closely covered and likely subject to mispricing. This all makes sense under a value investing assumption. Given his 50% claim, this sort of mispricing seems extreme.

Are there are any extreme inefficiencies (exploitable on the magnitude of 30%+ annual) using a quant trading viewpoint that are available to those with skill but small sums of money but closed to those with a lot of capital? What sort of examples are there? What is the evidence?

Answer 1

The claim that there are small opportunities that are overlooked by large institutions is increasingly untrue. Some large firms specialize specifically in aggregating a large number of low capacity strategies with low frequency signals. With good infrastructure, these firms can seek out a very large number of signals, each with very low incidence, much more quickly than a smaller outfit.

A high level workflow for this approach is to:

1. Have a good model construction and fitting pipeline.
2. Have data pipelines and normalized data representation that make it easy to construct design matrices for any arbitrary ticker in any arbitrary venue.
3. Have a common set of features that can be constructed over such normalized data, and expanded over a parameter space.
4. In an embarrassingly parallel manner, fit model(s) or signal(s) over all possible symbols on a first pass.
5. Where sensible, treat these model(s) or signal(s) as meta-features which are then fed into an ensemble model so that it can be aggregated into one strategy.
6. Traders hand-tune these model(s) or signal(s) either with the aid of simulation or post-trade log from live trading OR the model(s) and signal(s) are themselves fed into a monetization model that optimizes the execution trajectory.
7. Pass all orders to an internal matching engine (or "internalizer") that matches orders between different strategies before sending them out to the public gateway, and/or administrate these strategies separately (e.g. different siloed teams working for the same company), giving them separate accounts and/or session IDs and relying on venue-side self-match prevention to mitigate wash trading.

It's not unusual for such a firm to have thousands of "strategies" being monitored by just 1 trader. Individually, some of these strategies may have extremely low capacity (trigger rate in the single digit per day, and only requires <$100k of margin).

Answer 2

Small lot securitized product bonds or greatly factored down ones.