Decided to see if AI could build a better Value Portfolio. Here's the verdict so far

Claude became the de-facto way I did a lot of equity research, since it can do so many things, even with just public data from the web. I decided to run an experiment (still in progress) by giving it tons of curated data to see if it could build a value portfolio that could beat the market. I launched it at the end of April and here's the verdict so far, as of 07/17/2026:

|Performance YTD|Performance since Inception|S&P YTD|S&P since Inception|
|:-|:-|:-|:-|
|16.84%|12.88%|9.8%|5.2%|

And here are the holdings, in case anyone is interested:

|Symbol|YTD Performance|Weight|
|:-|:-|:-|
||
|NJR|\+28.9%|16.76%|
|TRV|18.5%|11.76%|
|GL|33.3%|9.14%|
|CASH|29.2%|8.04%|
|ADUS|6.8%|6.79%|
|FG|7.8%|6.56%|
|APAM|\-5.4%|5.63%|
|HCI|\-3%|4.58%|
|UFCS|40.5%|4.39%|
|KHC|7.5%|4.01%|
|CAG|\-16.4%|3.34%|
|SLDE|11%|3.12%|
|ADT|\-11.9%|2.54%|
|NRDS|\-26.7%|2.51%|
|ASO|\-6.8%|2.43%|
|SKWD|20.3%|2.23%|
|SBH|4.6%|2.22%|
|OSCR|92.8%|1.66%|
|UPBD|31.2%|1.15%|
|PRG|61.8%|1.12%|

And I'll share the methodology behind it as well:

Screening rules to define the universe:

* No cyclical industries: Oil &Gas, Airlines, Auto, Homebuilding, Semiconductors
* Key criteria for ratios:
* P/E: 1-35
* P/S: 0.1-10
* P/B: 0.3-10
* Price to Free Cash Flow: 1-35
* Operating Profit Margin: 0.05-1
* Gross Profit Margin: 0.20-1
* Current Ratio: 0.8-20.0
* Debt to Equity: 0.0-2.5
* Operating Cash Flow to Sales: 0.05-1.0

After defining the universe, the AI analyzes every financial filing and transcript call to firmly understand how these companies operate, their competitive advantages, and how they evolve over time, picking the top 20 candidates to hold.

Still very early, but I am encouraged by these results so far.