Victor,
This is crazy impressive; even a 15% top bucket on EasyToTrade is Impressive! I’ve built multiple models and found many local minimum buckets when searching for an optimal solution. e.g., I followed the rabbit hole you created, starting with Price Volatility, Then Moving to Earnings Volatility, then Moving to Estimates Volatility, anchoring each as the primary driver against other sprinkled nodes. I’ve not even come close to your performance; attached are my feeble attempts that, while positive and valuable as a hedge against my main strategy, are still lacking; note one of the systems, I reversed the process and looked for the bottom bucket vs the top trying to take advantage of NA as Negative.
Also, I’ve noticed that even if I have a reasonable top bucket performance, I may have a lower performance when I apply that to the simulated strategy. Heck, applying a single factor CurQEPSStdDev while the RS is lower leads to outsized performance in the SS. (See Below)
Example Model built from your starting point with 4% across 25 Nodes.
This leads to this SS (new Rule Close(0, $VIX) < 20 Hat Tip Hemmerling + Ranking), which is quite good, but I’m hesitant to trust it!
Then I tried to build from Scratch without anchoring from your model and flipped the ranking to take advantage of NA as Negative. Where the anchor was on CurQEPSStdDev (40% weight)
This leads to this SS (Close(0, $VIX) < 20 + Ranking ), which is reasonable as a hedge, but again I’m hesitant to trust it!
But If I just anchor against CurQEPSStdDev (100%) I end up with the following Which is incredible for a single factor.
Finally I’ve had minimal success flipping a long model as shown below.
So no success in flipping and tailoring a strong long model to a short model. I don’t recommend it.
So all this long-winded analysis comes down to the question of “HOW in the world did you get to 15% top bucket performance?”
Tyler