Post New

P123 isn’t an inexpensive option for MI but their tool set is becoming more and more sophisticated.

You can now download historical data with normalized factor data and future target prices for your
personal machine Learning price prediction. I’m still in the experimental stages, experimenting with
Random Forest, Extra trees and SVM. Nothing has worked for me as well as I can do with traditional
screens in the SP500 arena with GTR1 or P123. ML seems to work better in mid or small caps using
2 or 3 month future returns to eliminate short term random results.
Currently one has to download data and use your own Machine Learning Software on your computer...not that easy!

But Marco of P123 posted a Preview of their in-development system where they give you
The tools to build your own ML system and historically test it (backtesting).

https://community.portfolio123.com/t/preview-scree...

RAM

RAM ..... I find this statement a bit strange

"I’m still in the experimental stages, experimenting with
Random Forest, Extra trees and SVM. Nothing has worked for me as well as I can do with traditional
screens in the SP500 arena with GTR1 or P123"

All traditional screening methods are nothing but intersection rulesets - RF ( Random Forest) is a ruleset ensemble. So its almost impossible for it not to reproduce any GTR1 based screener.

The KEY differences are that in ML based methods - there's inbuilt Train, Test, Validation samples to ensure your model is not OVERFITTING. While on Screens there's NONE. You are using the entire Lookback period to test the efficacy of your method.

In that case - I would PERSONALLY trust RF ( Its literally one of the most robust MLs out there , handling noise based data) more than your screen ie its pointing out a major Overfit risk.

Best

Yes, the P123 ML project is interesting.

Human stock screens are very different than AI generated screens. For example, Random Forest combines hundreds of decision trees, while a standard MI screen is one decision tree. To reduce overfitting, Random Forest uses simple decision trees, while humans use common sense. Standard AI methods do not have common sense or financial literacy, and so will produce different screens than humans. Modern AI methods might produce better stock screens compared to human expert rules based screens, or they might not.

The stock market has feedback with successful investors gaining more capital. AI methods will have to evolve or die, with a cycle:
First movers make a profit.
Others catch up and the trades become crowded and unprofitable.
New methods are developed and the cycle continues.

Successful AI methods will switch between investing strategies, something that has proven difficult for humans.

There are 2 limits to AI: data and processing. Both improve over time, and new algoriths are being developed, and so AI will overtake humans in investing at some point. But we're not there yet.

AI Powered Equity ETF (AIEQ) uses IBM Watson.
VanEck Social Sentiment ETF (BUZZ) uses sentiment from news and social media.
Qraft AI-Enhanced U.S. Large Cap Momentum ETF (AMOM) This AI picks large-cap momentum stocks.
WisdomTree International AI Enhanced Value Fund (AIVI) This AI model picks value-priced stocks.

                  Name                    Annualized Return  Annualized Standard Deviation
         SPDR S&P 500 ETF Trust                  10%                      18%
      Amplify AI Powered Equity ETF              -2%                      25%
      VanEck Social Sentiment ETF                -5%                      33%
QRAFT AI-Enhanced US Large Cap Mmntm ETF          5%                      22%
  WisdomTree Intl Al Enhanced Val ETF             3%                      17%
 Vanguard Total International Stock ETF           1%                      17%

Returns from Apr-21 to Feb-24 https://www.portfoliovisualizer.com/factor-analysi...

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.