AI Model Stock Pulls a 1000% Gain in 1 day

General
1

I kept track of a two models on Wednesday Sept. 4. The random forest III had one of its top holding achieve 1000% return over the next 2 days. WHLR.

This was a 180 million dollar company by market cap this year, selling for 1.6 million dollar market cap the day the model bought it. All the valuation ratios were also 100x better than your average company out there. They had a class action lawsuit that resolved on the Thursday.

Wondering if I could replicate these “extreme” value stock picks in my own models. I tried adding an extreme factor weight that increases the weight of price to sales by 10x in my models if a company’s price to sales is less than .01. It did seem to add a couple percentage points to the return of my model, however not as drastic as you would think.

Do you think this model succeeding like this is random chance? Could you replicate it and how? Anything I try myself doesn’t seem to work reliably in all geographies (US, Can, Europe).

I feel like buying companies discounted 100x that have a 1/10 chance of surviving would be worthwhile since you’d be ahead as long as 1/10 survive - but doesn’t seem to work that well when I set it up that way.

The 5 year did show 55% annualized on the AI model so pretty reasonable.

I guess I should try copying the most important features in the model in the importance section eh?

2

Have you bought it?

3

This was a legal special sit, a pending class action lawsuit which got dismissed that's why the stock soared. I don't think any sort of factor strategy can be constructed solely on fundamental and price data which would capture this.

1 Like
4

Maybe a model for the nano stock universe somehow could. But I can't really distinguish between the theoretical 300% annualised returns that my model tells me are data illusions or real possibilities.