Trading strategies tested under identical conditions
Enter your rule, pick up to five markets and press Test. Three tests a day, no account, no sign-up, no e-mail. You get the same verdict as everyone else — on the same data, with the same fees, across the same fifteen time windows.
We also test on data that did not exist when you submitted. That data cannot be overfitted — not even by us.
1 — Indicators
2 — Rules
Allowed: > < >= <= and or not + − × ÷ numbers
and the names of your indicators. No code — nothing is executed.
3 — Markets
A rule that only works on one market has memorised that market. Several markets in a single run show the difference.
4 — Appear in the register as
A random handle is filled in by default. If you want to be named, overwrite it — otherwise no name appears. The handle stays the same so the attempt counter cannot be sidestepped by changing names.
This is what gets submitted
A readable file, not a program. Download it, pass it on, submit it again whenever you like.
Out-of-sample return per market
Forward test: still open. From submission onward your strategy is evaluated every Monday on everything that has arrived since — on prices nobody knew when it was submitted. It is the one number that cannot be massaged.
The result appears in the register further down once enough new prices have come in.
Same prices, same fees, same windows. The data revision and its checksum are printed in every report — anyone can recompute it.
Submit twenty variants and one will shine by chance. So every verdict states which attempt it was. Yours included.
From submission onward, only what happens next counts. Nobody can overfit that number — we cannot either.
3tests a day
pruefstand/p/a7f3c2 — bookmark it or pass it on, it stays valid20tests a day
9 €per month, incl. VAT
Three a day sounds like very little, and that is the point. Anyone working through twenty variants is no longer searching — they are waiting for one to shine by chance. That is exactly what this service warns about one heading further up. A limit that guards against it is not a gate at the door; it is part of the method.
Which is why no amount of money buys unlimited tests. The Workbench costs something because it computes more — every market instead of ten, a portfolio instead of a single rule, every week instead of once. It does not cost something so you may submit the same idea more often. The attempt counter keeps running on every tier and is never reset. It appears in your report and in the public register, visible to all — including for those who paid.