In development

Strategy Creator

Our research lab. We look for trading strategies and put them through tests that are tougher than usual — before we risk a single dollar.

Stage: forward test, on virtual accountsReal money: not yetNumbers updated: —
In short

The research in numbers

—ideas pre-registered and sealed
—variants evaluated on unseen data
—failed the tests
—virtual accounts in the forward test

Only one idea in ten passes

Every variant was evaluated on data we did not use to build it.

  • a steady risk-adjusted gain on the unseen period
  • a statistically significant result, not luck
  • stays positive even at double trading costs
  • positive in most years, not just a few
  • beats a random control test
Why do so many fail? Read the explanation →

Research pace

Ideas sealed per day, before testing.

After the exploration phase the pace slows on purpose: next comes real-time verification.

Method

From idea to decision in five steps

Rules are fixed before the results. That way we can’t fool ourselves by tweaking a strategy until it looks good on the past.

The idea

We start from a real market mechanism, not from a chart that “looks good”.

Sealing

Rules, costs and criteria are written before any calculation and sealed with a digital fingerprint. They can no longer be changed.

The test on unseen data

The strategy is evaluated once on a hidden period, with real measured costs, not assumed ones.

The forward test

Whatever passes runs on virtual accounts, day by day, and must reproduce the simulation exactly.

The decision

Following a protocol fixed in advance: a strategy stops at the first bad sign; real money, a little, only at the end.

Where we are

The forward test, in real time

— of 126 trading days

A forward test can’t prove a strategy in a few months — it can only reject it. That’s why good results aren’t taken as proof, while bad ones stop the strategy right away. The asymmetry is intentional.

Intraday · virtual accounts

Results day by day

Every strategy that passed the tests runs on a virtual account, on real market data and with real costs, but without real money. We publish each account’s results day by day; the strategies’ names and rules stay hidden. Updates automatically every day.

See all accounts Profiles and the mix calculator
—active virtual accounts
—in the green / in the red
—median result
—trading days of testing
What we learned

Lessons from more than 200 tests

We publish them because they explain why most ideas fail — and why we treat any result that looks too good with suspicion.

The real cost decides everything

Many effects really exist, but they are smaller than the difference between the buying and the selling price. We measure costs on real fills.

Too good to be true

A spectacular result on a simple idea is almost always an error: information from the future leaking into the data. We look for that first.

Unseen data is used only once

We had excellent ideas on the build years that vanished completely on the unseen period. That’s exactly why we keep a period untouched.

A loss doesn’t become a gain when reversed

Costs are paid in both directions. A losing strategy doesn’t become profitable if you do the opposite.

Nothing is skipped silently

The trades “missing” from a test are usually exactly the bad ones. We count everything that was generated, not just what was evaluated.

The past through the eyes of the past

We don’t classify history by how things look today. Every simulated decision uses only what was known at that moment.

Full explanation · ~8 minutes Why 9 out of 10 trading strategies fail — and what to watch out for Luck mistaken for skill, tweaking until it looks good, costs and information from the future — plus the warning signs when someone shows you a “robot that surely works”. Read the explanation →
What we work on

The data behind the research

More than 12 GB of market history and public data, cleaned and checked before any test.

US market, minute by minuteOpening and closing auctionsPublic SEC filings Short sales (FINRA)Futures positioning (CFTC)ETF fund flows Options quotesCurrencies and interest ratesInternational indexes and commoditiesEconomic calendar
For investors

Why it matters to you

Rules before money

Every strategy has a protocol written in advance: when it continues, when it stops, when it moves to real money.

We show what fails, too

Nine ideas out of ten don’t pass. We’d rather reject them in research than test them with someone’s money.

Small steps

Real money only after six months of a successful forward test — and in small amounts at first.

Strategy Creator is an ongoing research project. The results described are simulations and tests on virtual accounts, without real money. The information on this page is neither an investment offer nor investment advice. Past results, simulated or real, do not guarantee future results.