Strategy Creator

Build a strategy from 48 indicators with the visual builder, describe it in plain English, or write the code — preview it live on the chart, backtest it inline, save it to your library, and send it to The Forge.

The Strategy Creator (/strategy-creator) is where you make a new strategy. You can build one from indicators without writing code, describe one in plain English and let your LLM draft it, or write the Python yourself. Whichever way you start, you see its signals on a live chart as you edit, backtest it inline, save it to your personal library, and — when you are ready — send it to The Forge, where the pipeline takes over.

This page is for anyone designing strategies by hand. If you want the autonomous system to invent strategies for you, see crucibles and the research daemon.

Forven is a research tool. A strategy you build here has proven nothing until it survives the gauntlet. Backtest numbers are illustrative, not predictive, and nothing here is financial advice.

Three ways to build

The Creator has three modes. You can switch between them while you work.

Visual

A no-code builder over the indicator library:

  • A searchable indicator palette — 48 vectorized indicators in six families (below), each with its own parameters and defaults.
  • Conditions — compare indicators to each other or to values (crosses above, is below, and so on), and combine them in nested AND / OR groups.
  • Long and short sides — separate entry and exit rules for each direction.
FamilyIndicators
Moving averageSMA, EMA, WMA, DEMA, TEMA, Hull MA, VWMA
TrendMACD, PPO, ADX, Aroon, Vortex, TRIX, Supertrend, Parabolic SAR, Donchian, Ichimoku, Linear Regression
MomentumRSI, Stochastic, Stochastic RSI, KDJ, CCI, Williams %R, Rate of Change, Momentum, Chande Momentum, TSI, Ultimate Oscillator, Awesome Oscillator, Connors RSI
VolatilityATR, Normalized ATR, Bollinger Bands, Bollinger Width, Keltner Channel, Std Dev, Historical Volatility
VolumeVWAP, OBV, Accumulation/Distribution, Chaikin Money Flow, Money Flow Index, Volume SMA, Force Index, Ease of Movement
CryptoFunding Z-Score, Open Interest ROC

The crypto indicators read Forven's enrichment streams. The full catalog is also served by GET /api/indicators.

AI

Describe the strategy in plain English — for example "Buy when RSI drops below 30 and price is above the 200 EMA; sell when RSI goes above 60." — and your connected LLM turns it into a validated rule spec in the visual builder, where you can inspect and edit every condition. It uses your own model key (see models & providers); with no model connected, AI generation can't run, and the rest of the Creator works as normal.

Code

Write the strategy as Python against the BaseStrategy interface: a generate_signals(df) method, a TYPE_NAME, and a STRATEGY_CLASS. The editor starts from a template. Code strategies run through the same registration validator and sandbox as any other custom strategy — see writing custom strategies.

Check data before you generate

In AI mode, Check data runs a read-only preflight — no model call — against your description. It lists the markets, timeframes, and data inputs it detected, flags any that are missing locally, and links to the Data page to collect them. Generation repeats the same check on the server, so checking first is optional but saves a wasted model call.

A passing check means the local files exist. It does not certify that there is enough history, that enrichment values are non-null, or that the idea is any good — the backtest and the gauntlet still decide that. Named inputs Forven has no integration for (for example CDD, HODL waves, ETF flows, or VIX) block generation until you revise the idea; downloading an arbitrary file cannot satisfy them. The same preflight runs before an agent develops a candidate from a crucible, and a task blocked on missing data can be resumed once you have collected it.

Preview, then backtest

  • Live preview. As you edit, the chart shows the indicator overlays and where the strategy would enter and exit — no run needed. It is the fastest way to see that a rule fires where you think it does.
  • Market scope and execution settings. Pick the symbol and timeframe, the trade direction (long only, short only, or both), and the execution settings: stop-loss, take-profit, trailing stop, time stop, ATR stop multiplier, sizing mode (full equity, risk fraction, fixed notional, Kelly, or ATR risk), initial capital, leverage, fee, and slippage. These travel with the strategy.
  • Inline backtest. Run a backtest from the Creator and read its out-of-sample result right there, or open the full report. For a more deliberate run, Open Manual Backtest.

Your library

Name a strategy and Save it to your personal library, My Strategies. From the Open a strategy… menu you can load:

  • one of the built-in templates, or a blank canvas;
  • anything in your library;
  • any built-in strategy;
  • optionally, strategies the app itself generated.

When you open an existing strategy and change it, Forven asks whether to overwrite it or save a new one. Library entries can be duplicated and deleted.

Send to Forge

Send to Forge turns a saved library strategy into a pipeline strategy container at the quick_screen stage, keeping its market, timeframe, parameters, and execution settings. From there the normal pipeline applies: quick-screen guardrails, the gauntlet, and the gates. The library entry shows an in forge marker once it has been sent. Save your latest edits first — the Creator will not send unsaved changes.

You can also bring a strategy in from elsewhere: Import takes a strategy export (from another install, or from The Forge's export) and creates a new quick_screen container from it.

Steps: build and send your first strategy

  1. Open Strategy Creator and choose Visual (or start from a template).
  2. Add an entry condition — for example RSI crosses below 30 — and an exit condition.
  3. Set the symbol and timeframe, then watch the live preview: the markers should land where you expect.
  4. Set execution settings, then run the inline backtest and read the out-of-sample numbers.
  5. Name the strategy and Save it.
  6. Send to Forge. Open it in The Forge and let the gauntlet decide.

What you'll see: a chart that updates as you edit, an inline result card with out-of-sample metrics, your strategy in My Strategies, and — after sending — a new container in The Forge at quick_screen.

Caveats

  • A good preview is not a good strategy. Markers landing in the right place says nothing about edge. Read out-of-sample, then let the gauntlet judge.
  • AI drafts need review. The natural-language path produces a validated spec, not a validated idea. Read the generated conditions before you trust them.
  • Data first. A strategy that needs data you don't have will be blocked, not silently run on zeros. Use Check data and the Data page.