Accepted
Putting cash in US Treasury bills
Cash that waits earns the current rate, with no added risk.
The method
One idea at a time. A judge every time.
What sets MFT Bot apart is not a secret formula. It is the way each rule earned its place, and the way most of them lost theirs.
48
ideas examined
10
accepted
31
rejected
7
pending or under observation
The research log as of September 24, 2026
- 1
We write the idea down before testing it
The idea enters the research log with its success criteria, in numbers, before any result is seen. No moving the target after the shot.
- 2
We judge it on data it has never seen
Tuned on one period, judged on another. An idea that only works where it was tuned is rejected. That is the most common case.
- 3
We require several proofs
Earning more is not enough. It must also not fall lower, and it must succeed in every sub-period.
- 4
We count the attempts
The more ideas you test, the more likely one of them succeeds by chance. The number of attempts is declared and the bar is raised accordingly.
- 5
Artificial intelligence audits, it does not decide
An AI reviews the trades and suggests ideas. It never changes a rule and never places an order.
A few verdicts
Accepted
Lower fees for the same leverage
The Growth Engine gets its leverage from a ×3 fund rather than a ×2 fund: same exposure, lower fees, +0.4 points per year in the 2010-2026 test.
Rejected
Grids and martingales
They win every day until the day they wipe out the account. Banned in the code.
Rejected
Using an AI to read the news and filter purchases
Tested on 328 trades: no measurable improvement.
Rejected
Expanding to 132 markets
More positions, but they all fall together on crash days. Worse than with 36.
Rejected
Making the Guardian's positions bigger
More risk, no more gain outside the tuning period.
The mirror account
Every order from the bot is replayed at a second broker. We measure the gap between what the bot thinks it gets and what it really gets. In September 2026: 13-second delay, price as expected.
What we do not hide
- The 18-year and 33-year performance figures are simulations. No one has collected these gains.
- Every worst drawdown is one from the past. A future decline can be deeper.
- The Guardian is simulated with stocks chosen in 2026, which makes its past look better.
- Our clients can consult the full research log, failures included.