A real-money test of an automated trading system, run in public. The system trades a single S&P 500 futures contract in a real brokerage account. Before the first trade we wrote down what would make us consider opening a fund with it — and what would make us walk away. Those rules are on this page, and they do not change during the test.
| Item | What it means |
|---|---|
| Portfolio strategy | A portfolio of seven independent signals that trade together as one. The signals look at different things — the shape of the day, overnight moves, recent highs and lows — and their instructions are combined into a single position. |
| Market | S&P 500 futures — the "micro" contract, one tenth the size of the standard contract. Each point of the index is worth $5 to us. |
| Size | One contract at a time, never more. The reference account described below is sized to cover one contract's full value, so the position is not geared against it. |
| Reference account | All percentages are measured against a fixed reference account of $38,500 — the full value of one contract at the start of the test (the index level then, times $5 per point), not a margin figure. It keeps the numbers comparable over time. |
| How orders are placed | Automatically by the software, which works out a limit price for every order with its own algorithms (anchored to the market's bid and ask). If a limit has not filled within about 20 seconds it is cancelled and the remainder is completed at market. There is no manual intervention. |
| Item | What it means |
|---|---|
| Test period | 2026-09-01 → 2027-09-01 — twelve months. |
| The first weeks | Every session from the first day is part of both the record and the checks. Nothing is set aside or treated as a practice run. |
| Minimum data | At least 230 trading sessions with confirmed results (a year has about 252). A short period full of luck is not enough to decide anything. |
| Limit | Value |
|---|---|
| Biggest acceptable fall | $3,850 — 10% of the reference account — measured from the highest point reached. If the account falls further than that, the test records a risk-limit breach and one of the checks fails. |
| Position size | One contract. Always. |
| Other markets | None. No options, no hedging, no side bets. |
| Manual interference | None, except to deal with a documented technical failure — and any such event is written up and published. |
We show the S&P 500 index over exactly the same days, so you can see what simply owning the market would have done. It is there for context: beating the index is not one of the four checks. The checks are about whether the system does what it was supposed to do, not about beating a benchmark.
These were written down before trading began. The numbers attached to them were fixed on the day this page was published (see section 12) and then frozen.
We replay the same rules over the same days and compare the two sets of daily results. The check is deliberately one-sided: live trading must not fall more than 10 percentage points a year below the replay. That floor is generous on purpose — it is about the size of the trading costs and the noise you cannot remove — and it is only a floor: if live trading ends up ahead of the replay, nothing about the launch case is worse, so there is no ceiling.
Every published trade must match the broker's records, and every gap must have a written explanation. At the end, nothing material may be left unexplained.
Two things are checked, and they are deliberately not symmetric.
The return is measured against the market, not in absolute terms. Our two years of historical testing happened to cover a rising market, so an absolute-return target would be unfair in a year when the market falls: the strategy could beat the index comfortably and still look like a failure. So the test asks how far ahead of the S&P 500 the account is running, and requires that gap to stay above the floor that the strategy's own history suggests. There is no ceiling — being further ahead of the index than history suggests is not a failure.
The worst fall is still an absolute limit. It must stay inside the range that history suggests, and inside the fixed risk limit in section 5. That part of the check is unchanged.
Trading costs must stay under $2.50 per round trip, the risk limit in section 5 must not be breached, and every operational problem must be resolved and disclosed.
All four checks must pass, and there must be enough data. Then — and only then — the fund may be considered. If any single check fails, no fund is launched under these rules. Passing is necessary but not sufficient: legal, operational and business readiness also have to be in place.
We do not publish the trading logic itself: which signals we use, how they are tuned, or the exact rules that trigger a trade. That is our work, and publishing it would be no different from handing out the recipe. What we publish is enough for anyone to check the claims on this site — how results are measured, what the targets were, and whether they were met.
Honesty about timing matters as much as the rules themselves:
The published results are the broker's numbers, not ours: they are rebuilt from trade confirmations, and the data files on this site carry a fingerprint of the previous version. An independent, dated registration of this commitment is being arranged with a public research registry; once it exists, its reference will be shown here and on the front page.
Practical details — how the numbers are collected, what "confirmed" means, and how the fingerprints work — are recorded on the Performance and Scorecard pages.
This site is a public research record. It is not an offer to sell, or a solicitation of an offer to buy, any investment. No money is accepted here, no account is opened here, and nothing on these pages is investment advice. Trading futures involves substantial risk of loss. The email sign-up in the banner collects addresses only for progress reports, with your explicit consent, and has nothing to do with any investment.