May 2026
IN THIS ISSUE
Less data?
Bad data?
Zero data?
The Chatbot Alpha Trap
I learned a lot from day trading early in my investing career. I followed the economic calendar for each day’s data releases, with CNBC’s Squawk the Street on, sat at my Interactive Brokers trading screen, trigger finger on the mouse button, prepared for the split second that economic data was released. I instantly (tried to) gauge whether the data was good or bad, then either bought or sold e-mini S&P 500 futures contracts to profit from the rapid price action that followed.
The problem was that the futures prices reacted faster than CNBC and I ever could. Getting favorable entry points given the whipsawing volatility in prices was not feasible either. Sometimes I was right about the initial direction but then prices reversed after entry; other times I was wrong about the direction altogether (because sometimes bad news is good news).
Similarly, once high-speed algorithmic trading gained widespread awareness in 2009 (Stock Traders Find Speed Pays, in Milliseconds, New York Times), the Flash Crash in 2010, and Michael Lewis’s fantastic book Flash Boys: A Wall Street Revolt in 2014, it was obvious that neither I nor any human could beat quant algorithms at their own game. I had to either acquire the tech to play that game better than billion dollar quant funds could play (not a chance!), or I had to make better decisions over longer time horizons where profits depend more on decision quality and less on speed.
Perhaps the newest incarnation of this phenomenon is the AI chatbot. Just as I thought I could sit at my trading screen and aggregate CNBC’s economic data faster than other market participants, it is so natural to wonder whether AI can outperform other market participants by processing more information, more quickly, and more deeply, than humans.
Insofar as efficient markets quickly and accurately price all publicly known information about each stock, it follows that AI should be able to price this information faster and profit from market inefficiencies.
And yet, AI via large language models (LLMs) are trained on human-generated language. The SPIVA Scorecard shows that Large-Cap active fund managers systematically underperform passive indexing by increasing rates over time. As of December 2025:
66.8% of funds underperformed over 3 years
89.0% of funds underperformed over 5 years (Likely distorted from COVID in 2020-21)
85.6% of funds underperformed over 10 years
89.9% of funds underperformed over 15 years
If human active managers cannot reliably beat passive indexing, why would a chat bot trained on human language be any better at picking winners?
So of course there are folks putting AI models to the test (Bloomberg). A recent trading contest called Alpha Arena pitted eight major frontier AI systems, including Claude, Gemini, and ChatGPT, against each other for investing. Each was given $10,000 to trade over two weeks (yeah yeah, I know this is too short, shoulda been longer for stronger reliability). The total portfolio collectively lost about one-third of its capital in that short time, and individually across all 32 sets of results only 6 models finished with profits. Not surprisingly, the bots exhibited the same flaws as human day traders - they traded too much, mistimed their entries, and sized positions incorrectly.
At least for now, ChatGPT-based active management seems to accelerate human fallibility more than improving human decision-making.
ONE MORE THING…
Less data? Less data might help diminish short-term reporting incentives… but the better solution is behavioral awareness, not limiting data. Kalshi traders confident SEC will end mandatory quarterly earnings reports
Bad data? President Trump fired the head of the BLS for producing “rigged” data that made him look bad. This study found that having high quality, high integrity data generates economic benefits of $25 for each $1 spent on the agency’s budget. Fix the cause, not the messenger. The Value of Reliable Statistics | NBER
Zero data? Typically investors consider Price to Earnings ratios or similar valuation metrics to determine whether a stock’s earnings are strong enough to generate sufficient returns on their invested capital in a reasonable amount of time. Well, SpaceX’s upcoming IPO is asking folks to evaluate an investment with zero earnings. Buyers beware, you cannot divide by zero. 6 Charts on SpaceX’s Pre-IPO Financials | Morningstar
The information and opinions contained in this newsletter are for background and informational/educational purposes only. The information herein is not personalized investment advice nor an investment recommendation on the part of Likely Capital Management, LLC (“Likely Capital”). No portion of the commentary included herein is to be construed as an offer or a solicitation to effect any transaction in securities. No representation, warranty, or undertaking, express or implied, is given as to the accuracy or completeness of the information or opinions contained herein, and no liability is accepted as to the accuracy or completeness of any such information or opinions.
Past performance is not indicative of future performance. There can be no assurance that any investment described herein will replicate its past performance or achieve its current objectives.
Copyright in this newsletter is owned by Likely Capital unless otherwise indicated. The unauthorized use of any material herein may violate numerous statutes, regulations and laws, including, but not limited to, copyright or trademark laws.
Any third-party web sites (“Linked Sites”) or services linked to by this newsletter are not under our control, and therefore we take no responsibility for the Linked Site’s content. The inclusion of any Linked Site does not imply endorsement by Likely Capital of the Linked Site. Use of any such Linked Site is at the user’s own risk.