KURZHARR

Getting started

Build your first strategy, run it once, and find your way around the result it produces.

A strategy comes into existence by being run. There is no separate “save” step — you describe an idea, lock the bar it has to clear, and submit it. The first run creates the strategy; everything after that is a new trial under it.

Start from Strategies → New strategy. The builder is a short wizard, and it asks one thing at a time so you can see what is left.

Building the strategy

  1. Market. Pick the symbol, timeframe and date range to test against. This is set once here so nothing downstream has to guess it.
  2. Blocks. Choose the indicators and features you want available. This is the palette the next step builds from — the description step will only use what you select here.
  3. Describe. Write the strategy in plain language and the platform compiles it into a structured spec, or paste and edit the spec directly. Either way you review the interpreted spec before anything runs.
  4. Structure (optional). A closed check for indicators that are wired in but cannot actually affect any entry or exit — a common way to fool yourself into thinking a filter is doing work it is not.
  5. Your trades (optional). Check the spec against trades you actually took, so you can confirm it means what you think before you spend a trial on it.
  6. Commit. Name the strategy and lock its floors — the minimum trade count and significance it will be judged against. Floors are fixed before the run starts and cannot be changed afterwards.
Plain-language translation is a Pro-tier feature
On the Free tier you write the spec directly, or use the visual block builder (which produces the same spec with no language model involved). The interpreted-spec confirmation step is mandatory on every tier — the platform never runs a translation you have not seen.

Reading the result

Submitting queues a job that runs the full validation stack. When it finishes, the scorecard appears under the strategy. A strategy page has five tabs:

  • Result — the scorecard: the verdict, each check, the attribution, the deflated-Sharpe accounting for your account, and an equity replay that annualizes the return and draws a buy-and-hold line for the traded symbol next to it.
  • Chart — every trade the run took, on the price series, so you can see where the strategy actually fired.
  • Compare — trades from two trials side by side.
  • Forward — register the strategy to be tracked on data that did not exist when you built it.
  • Refine — change the strategy and run it again.

See the scorecard & verdicts for how to read each part, and the validation stack for what each check is testing.

Refining

On the Refine tab you change the strategy — in plain language, or by editing the spec — and then press Run as a new trial. Nothing runs until you do; changing the spec and running it are two separate acts, and only the second one costs anything.

Every run is a counted trial
The first run and the fiftieth refinement are recorded the same way: permanently, against your account. The deflated-Sharpe accounting raises the significance bar each time, because running many variants of one idea until something clears is exactly what that accounting exists to catch. Refining is deliberately not free.

Nothing you submit leaves the platform. Runs execute here, against data the platform already holds, and the scorecard comes back to you.