7 Comments
User's avatar
James clark's avatar

Have you explored aggregating these sentiments over longer periods ( weekly/ monthly? )

Expand full comment
Quantitativo's avatar

Yes! There are 4 additional explorations I will write about in follow up pieces:

1. Weekly/monthly rebalancing

2. Using more features (price action, supply/demand imbalance, etc)

3. Using non-linear models

4. Incorporating short inventory/availability data

Expand full comment
James clark's avatar

Looking forward to it!

Expand full comment
Matt L's avatar

one of your best articles yet!

Did you by any chance run the same experiment for going Long only? There can be - as you mention - issues with short selling both operationally and emotionally - therefore wondering what the result would be if you only went long the top quartile in your model?

Expand full comment
Quantitativo's avatar

Thanks! Yes, a long-only model also works, but it is not as impressive. In my view, it is not worth reporting... the key benefit of this model, imho, is market neutrality, which is only achieved with longs and shorts.

There's one improvement related to your idea, though: implement a simple market regime prediction model and tilt to long-bias if in a bull regime, short-bias if in a bear regime, or neutral if in a sideways market

Expand full comment
michael's avatar

Very cool article. Average pnl seems kind of low though. Did you factor in slippage?

Expand full comment
Quantitativo's avatar

Thanks! Yes, I’ve considered trading costs. But remember: this is a market-neutral strategy with a near-zero correlation to the overall market. That means its returns are largely uncorrelated with market returns, so when combined with a long-only market exposure (e.g., S&P 500), the portfolio's overall performance can benefit from diversification. The returns from this strategy could be additive on a risk-adjusted basis, potentially improving the Sharpe ratio of the total portfolio.

More important than the absolute returns is the fact that the return stream is uncorrelated to the market.

Cheers!

Expand full comment