AI quant research · stocks, futures, crypto

Too many AI strategies. Which ones are real?

AI now churns out thousands of "winning" trading strategies. 100 Basis verifies them with precision: rules locked first, every attempt counted, real trading costs and tax included. So you know what's real before you put money in.

Get early access → See what we found

Rules locked before testingEvery attempt countedNo paid picks

Trade ahead of pension-fund rebalancing

Stock and bond futures · 2010–2024

✕ Not real

Paper's Sharpe

1.11

Our test

0.34

w/o 10 best days

−0.8%

Rules locked before test✓
Profitable after costs✓
Clears the bar after costs✕
Survives without its 10 best days✕

A recent research paper. Most of the gain came from a handful of crisis days.

95%

confidence required that a result is not luck

60M+

market data points in our research database

taken from brokers, funds or strategy sellers

100%

of attempts counted, none hidden

How it works

AI trading strategies that look too good to be true.

Try a hundred ideas and one will look brilliant by luck alone. We check whether a strategy holds up the way a careful scientist would, and show every number.

Rules locked first

Written and time-stamped before any result is seen. No tweaking after.

Every attempt counted

Ten tries and one winner isn't a winner. The luck check counts them all.

Sealed years

The most recent years stay locked until the end, then are opened once.

Real costs & tax

Then compared with simply holding a total-market index fund.

Findings

What the numbers say.

Popular ideas, tested with the same rules every time. More every week.

Natural gas "storage report day"

Didn't hold

−0.23%per report day, before costs

A published paper found gas futures earn most of their gains on weekly storage-report days. In 2010–2024 those days lost money.

Front-running commodity index rolls

Gone

1.0 bpa trade before costs, −3.4 bp after

Worth about 3.6% a year in 2000–2009 according to the original study. In energy futures since 2010, almost nothing is left.

CME bitcoin gap "always fills"

Myth

63%filled vs 73% by chance

After a weekend gap, price more often keeps going. Trading toward the "fill" lost about 1.5% a trade.

28 futures trading ideas

0 of 28

0passed after costs

Rebalancing flows, index rolls, Fed days, storage reports and more. None survived real costs and the luck check.

Early access

Know what's real before you put money in.

Join the waitlist for early access.

For individuals and firms

Two kinds of accounts.

Individuals join the waitlist. Funds, advisers, trading firms and fintech platforms work with us directly.

Individual

Waitlist open

For investors, traders and anyone using AI to build strategies. Launching soon.

Research newsletter: one tested strategy per issue
Honest crypto research
Check your own strategy, or your AI's
Every result published with its rules and dates

Join the waitlist

Enterprise

For funds, advisers, prop firms, data vendors and fintech platforms. In development: tell us what you need.

Independent verification of strategies and backtests
Research ledger: every attempt your team makes, counted and time-stamped
Custom research across stocks, futures and crypto
Results by API for your platform

Enterprise

Talk to us.

Tell us what you need verified or researched. We read every message.