Skip to content
All library documents

Using the Gittins Index to Time Trading Strategy Changes

Article Quant Q&A · Author: LattePrincess
Read the original

Summary

The document considers whether the Gittins index, a dynamic allocation method for balancing exploration and exploitation in multi-armed bandits, can help decide when to switch among investment or trading strategies. It concludes that the index could support tactical allocation when strategies have sufficiently consistent risk and return profiles to inform forecasts.

The approach depends on forecasts of each strategy’s returns and a chosen measure of uncertainty, such as expected volatility or drawdown. The index does not remove the need to estimate these inputs, and its value depends on forecast quality. The discussion is conceptual: it supplies no empirical test or implementation details. It also assumes switching has no cost and does not explain how to incorporate transaction costs, changing risk, or estimation error.

Key ideas

  • The Gittins index can be considered for dynamically allocating among trading strategies.
  • Its usefulness depends on forecasts of strategy returns and uncertainty.
  • Uncertainty may be measured with volatility, drawdowns, or another defined risk measure.
  • The discussion assumes switching costs are absent and gives no empirical evidence.

Tags

StatisticsPortfolio constructionRisk management
Full text
# Is the Gittins index useful in determining when to change an investment/trading strategy?


# Is the Gittins index useful in determining when to change an investment/trading strategy?












I've been reading about multi-armed bandits and the explore/exploit trade-off that can be solved with dynamic allocation indices such as the Gittins Index Theorem. Could this be applied to when to change investment/trading strategies assuming no cost in switching such strategies? If the fund's risk aversion is an issue, would this be applicable to a risk-neutral actor (say a hedge fund)?

## Answer by amdopt (score 2, accepted)

https://quant.stackexchange.com/a/75647

Gittins is as useful as your ability to forecast returns and uncertainty.

Depending on what you use for 'uncertainty,' you may just be replicating processes that other financial metrics already accomplish.

That said, there is nothing wrong with using Gittins to attempt dynamic, tactical asset allocation. It's a matter of how well your forecasting works. You need to forecast returns, and whatever you use for uncertainty, such as expected volatility, expected drawdowns, some other custom function that quantifies the uncertainty input, etc.

If you have several trading strategies that produce consistent risks and have consistent returns that you are comfortable using to forecast, Gittins could be used to help with the timing of a switch between strategies.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.