Defining Option Liquidity Filters to Avoid Backtest Overfitting

Summary

The document considers how to define a liquid strike when backtesting a tail hedge that pairs an index holding with monthly out-of-the-money put purchases. The proposed selection rule starts from a target strike near the underlying price, checks nearby strikes in a specified order, and skips the purchase if none pass an open-interest threshold. The author questions whether this is a sufficiently precise and defensible way to model liquidity, especially for an Indian options market whose trading activity increased over the sample period.

The central methodological concern is that loosely chosen filters or thresholds can be adjusted until a backtest looks favorable, creating overfitting. The discussion also notes that limited historical observations make it difficult to separate backtesting from forward testing. It poses the problem but offers no validated liquidity definition, alternative threshold, or performance results. A sound conclusion would therefore require evidence about historical tradability and careful treatment of how liquidity changed over time; the exchange itself does not supply that analysis.

Key ideas

• The proposed tail hedge buys monthly out-of-the-money puts alongside an index position.
• The strike selection rule checks a target strike and nearby alternatives using an open-interest threshold.
• Liquidity filters should be specified clearly to reduce discretionary backtest tuning.
• Changing market liquidity and a short history complicate backtesting and forward evaluation.
• The discussion raises these concerns but does not establish a validated filter or strategy result.

Tags

# Proper Definition of Backtesting Parameter


# Proper Definition of Backtesting Parameter












Currently I'm trying to test the efficacy of a tail-hedging strategy in which an investor goes long in an index and correspondingly buys 1-month OTM put options. For practical reasons, the options with the particular strike price that we buy each month must be liquid. My dilemma is in how exactly do I define the filters used to select the strike price.

Currently, there are 2 filters - open interest and the strike price itself:

1) Strike price - Suppose $K$ is 5% of the underlying price rounded to the nearest 100. Then the strike price is $K$ if the corresponding options are liquid, $K-100$ if $K$ isn't liquid, and $K+100$ if neither $K$ nor $K-100$ are liquid. No option purchased if none of them is liquid.

2) Open interest - A strike price with an open interest of at least 1000000 contracts is considered liquid.

Now the main problem is that the above definition of a liquid option is really vague. I read that it's a huge mistake to randomly vary or not properly define the parameters, and then to rely on the results of such a backtest. I'm testing this strategy for the Indian market, wherein option trading picked up from 2009 onwards (i.e., more liquid options were available) according to my project guide. This introduces another problem - I don't have enough data points to perform both backtesting and forward-testing.

So with all of the above background info in mind, how do I come up with a concrete definition of what a "liquid" strike price should be? Sure I can relax the definition and that way I'll have even more data points available from 2006 onward, but wouldn't that just be reckless overfitting?

Thanks in advance

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.