TheStockExperiment
ResultsPortfoliosRoundsHow it worksBlog
Next rebalance Dec 30
TheStockExperiment

An open research project. For informational purposes only. Not investment advice. Past performance does not guarantee future results. Full disclaimer

ExperimentResultsPortfoliosRounds
LearnHow it worksFAQDisclaimer
MoreBlogAboutContact
ResultsPortfoliosRoundsLearn

A live experiment · since March 2, 2026

Can machine learning pick stocks that beat the S&P 500?

Three model-built portfolios, each started with $1M. Every ~2 months the models rank all ~500 S&P stocks and buy the top 15. Every pick list is published and timestamped.

See the current picksHow it works
Next picks Dec 30
Current picks →

Growth of $1M since launch

Total return of each portfolio against the S&P 500. Hover the chart for exact values.

Meet the three portfolios

Same rules for all: 15 stocks, equal weight, rebalanced every 42 trading days.

* Ensemble #2 launched Apr 30; its first round is carried by its predecessor, ML Model #1.

Retired: ML Model #2, Mar 2 – Oct 29. Its rounds stay on the record below. Why it was retired

Round by round

Wins and losses both stay on the record.

How it works, in three steps

  1. 01

    Score every stock

    Each model rates all ~500 S&P 500 companies using prices, fundamentals and analyst data.

  2. 02

    Buy the top 15

    Equal weight, one company per slot. Every pick list is published and timestamped, so nobody can edit it later.

  3. 03

    Hold, then repeat

    Hold for 42 trading days (about 2 months), score again and swap. Every round is compared with the S&P 500.

Read the full method →