Methodology
An investing community for skeptics. This page explains exactly how the AI Investing Competition works, so you can verify every result yourself.
Season 1 Rules
Format
13-week season. Each week, five contestants submit two picks each: one stock from the S&P 500, one cryptocurrency.
Submission timing
Picks are submitted by Sunday 11:59 PM ET and locked in publicly Monday morning before market open. Late submissions count as a forfeit for that pick (zero points).
Scoring
Each pick is scored based on its percentage return for the week, calculated from Monday’s open to Friday’s close. The contestant with the highest combined return (stock return + crypto return, equally weighted) wins the week and earns one point. Tiebreakers: highest single-pick return for the week.
Season winner
The contestant with the most weekly wins after 13 weeks is the Season 1 winner.
Stock universe
Picks must come from the current S&P 500 index. ETFs, mutual funds, and stocks outside the S&P 500 are not eligible.
Crypto universe
Picks must be from the top 50 cryptocurrencies by market capitalization at the start of each week (per CoinMarketCap). Stablecoins are excluded.
Duplicate picks
If two or more contestants pick the same stock or the same crypto in the same week, the picks stand. We don’t force differentiation. (Reasoning: differentiation would advantage whoever submits first. Letting duplicates stand keeps the competition fair across contestants.)
The AI prompt
All four AIs (Claude, ChatGPT, Grok, Gemini) receive the identical prompt below, sent at the same time each week. No follow-up questions. No clarifications. Whatever the model responds with is what gets recorded.
The Crowd’s process
Guild of Gain members vote in a weekly poll. The stock and crypto with the most votes become The Crowd’s submission. Members can submit one vote per week.
The Season 1 AI Prompt
What we publish, what we don’t
We publish: every prompt sent, every AI response in full (not just the pick), every week’s results, the running scoreboard, the methodology behind every score.
We don’t publish: nothing. There’s no hidden layer. Everything that happens in the competition is on this site or on the public scoreboard.
If we ever change the rules mid-season (which we’ll avoid), we’ll publish the change with the date and the reason, and the change won’t apply retroactively.
Why we run it this way
There are a lot of ways to set up an AI investing competition. We picked this format because:
One prompt, no prompt engineering. Most AI comparisons cheat by tuning the prompt for each model. We use one prompt for all four. If a model needs to be coddled to perform well, that tells you something about the model.
Public picks, public results. A lot of AI hype is built on retrospective claims that can’t be verified. Our picks are timestamped, public, and unedited. You don’t have to trust us. You can check.
Including The Crowd. Retail investors are usually the unspoken control group in finance. Are the AIs actually better than a group of motivated humans? We’re going to find out.
No money on the line. Real-money trading introduces friction, slippage, and account complexity that distorts the test. Pure paper picks let us isolate the question: which contestant is best at reading the market this week?
The competition isn’t a stock-picking service. It’s a research project run in public. Don’t trade based on the picks. Watch the season, draw your own conclusions, and let the data inform how much weight you give to any AI’s market analysis.
See the contestants and standings: Head to the competition page to see who’s competing and how they’re doing.