Workflows

End-to-end guides for the most common paths through BitPredict.

These workflows show you how BitPredict's features connect across a complete journey — from finding a strategy to deploying it, or from an idea to a live custom strategy. Each step links to the relevant how-to for detailed instructions.

01Trading

Discovery → Backtest → Paper Trade → Go Live

The standard path for anyone who wants to take a pre-built BitPredict strategy from discovery to real capital deployment. Each stage validates the strategy further before increasing commitment.

1

Browse Strategies

Use leaderboards, AI search, and filters to surface candidates. Check Analytics Overview for market-wide context.

Reading a Strategy Detail Page
2

Analyse the Detail

Review all 6 tabs: equity curve, pipeline DAG, regime breakdown, trade ledger. Star favourites.

Reading a Strategy Detail Page
3

Run a Backtest

Validate on a 1–2 year date range with realistic risk parameters. Check trade count (aim 50+).

Run Your First Backtest
4

Optimise (optional)

Tune stop loss and take profit levels. Avoid overfitting — sort trials by Sharpe, not return %.

Run an Optimization Job
5

Paper Trade

Assign to a Virtual Account for 2–4 weeks. Monitor daily. Compare live performance to backtest.

Set Up a Virtual Account
6

Demo Trade

Move to Bybit Demo testnet for real exchange execution with zero capital risk. Run for 1–2 weeks.

Connect Demo Trading Account
7

Go Live

Request live trading access. Assign strategy to live account with conservative initial position sizing.

Key Outcomes

  • You've validated the strategy across 3 environments: backtest, simulation, and real exchange
  • You have a realistic expectation of signal frequency and execution quality
  • You understand the strategy's regime-conditional performance
  • You know how alerts work and have them set up before going live
2–4 weeks (mostly waiting for paper trade results)
02Strategy Builder

Idea → DAG → Optimise → Activate

How to go from a strategy idea to a live, signal-generating strategy using the visual DAG editor. This workflow covers the complete Strategy Builder lifecycle.

1

Browse Templates

Start from the template library. Filter by approach (trend/mean-reversion/momentum) and bar type. Open 2–3 templates to understand DAG structure before building your own.

2

Design the Pipeline

Sketch your signal logic: what data, which indicators, what conditions trigger entry, what rules exit. Write it in plain English before touching the canvas.

3

Wire the DAG

Drag nodes from the block library and connect them. Data Source → Indicators → Filters → Logic → Entry → Exit. Validate with the built-in validator.

Build a Strategy DAG Editor
4

Run Inline Backtest

Test your DAG without leaving the builder. Check trade count and basic metrics. Iterate on the DAG if results are poor.

5

Optimise Parameters

Submit an optimization job. Define parameter ranges for SL, TP, indicator periods. Sort trials by Sharpe. Import the robust winner.

Run an Optimization Job
6

Out-of-Sample Validation

Run one final backtest on a date range NOT used during optimisation. If performance holds up, you have a robust strategy.

7

Save as Draft

1 credit. Review one more time. Share the JSON with a colleague if you want a second opinion.

8

Activate

10 credits. Strategy goes live. Set up alerts immediately. Monitor for the first week closely.

Key Outcomes

  • A fully custom strategy running live in BitPredict
  • Confidence from multi-stage validation (backtest + optimisation + out-of-sample)
  • Alerts configured so you're notified on every signal
  • A reusable DAG template you can clone and modify for future strategies
1–3 days
03Machine Learning

Data → Train → Optimise → Integrate → Deploy

The full ML workflow: from selecting training data to deploying a custom model inside a live strategy pipeline. Requires intermediate knowledge of ML concepts.

1

Prepare Your Data

Go to Data → Data Quality. Find a bar type with ML Suitability score above 70. Check for adequate sample size (1,000+ bars).

Understanding Bar Types
2

Configure Training Job

Select symbol, bar type, date range, model architecture (neural network recommended), features, and forecast horizon (1–5 bars ahead).

3

Submit Training

Cost: 1 credit per 2 minutes of compute. A standard job takes 5–20 minutes. Monitor status in Training History (Pending → Running → Completed).

4

Evaluate Results

Review accuracy and loss metrics. A good classification model should exceed 55% directional accuracy on the test set (random baseline = 50%).

5

Run ML Optimisation

Submit an ML Optimisation job to systematically tune hyperparameters. Same credit model as regular optimisation. (optional)

6

Open Strategy Builder

Drag an ML Model node from the block library onto the canvas. In the inspector, select your trained model from the dropdown.

7

Wire the Model

Connect your Data Source to the ML Model node. Connect the model's output signal to your Entry/Exit logic. The model's prediction becomes a trading signal.

8

Backtest & Activate

Run a backtest on the ML-augmented strategy. Validate performance. Save as draft (1 credit) then activate (10 credits).

Key Outcomes

  • A trained ML model integrated into a live strategy pipeline
  • Understanding of which data features improve model accuracy
  • A systematic process for retraining and updating the model over time
  • An ML-augmented strategy generating live signals
2–5 days
04API & Development

API Key → Integrate → Build → Automate

How a developer gets from zero to a production integration with BitPredict's REST API — fetching data, subscribing to signals, triggering backtests, and building custom tooling.

1

Generate API Key

Settings → API Keys → Generate. Select scopes matching your use case. Store the key as an environment variable — never hardcode it.

Access the API
2

Read Auth Docs

Visit /api-docs/authentication. Understand how to pass the X-API-Key header. Test with a simple curl command to the /market-data endpoint.

3

Install the SDK

`pip install bitpredict-python` or `npm install @bitpredict/sdk`. The SDK handles authentication, rate limiting, and response parsing automatically.

4

Fetch Market Data

Pull OHLCV + regime data for your target symbol and bar type. Build familiarity with the response schema before touching write endpoints.

5

Subscribe to Signals

Use the Alerts API or poll the /strategies/{id}/signals endpoint. Design your application to handle duplicate signals gracefully (idempotency).

6

Trigger Backtests

POST to /backtests with your strategy config. Poll the status endpoint until completed. Retrieve results and store them in your own database.

7

Handle Rate Limits

Implement exponential backoff: on 429 response, wait `min(2^retry_count, 60)` seconds before retrying. Check X-RateLimit-Reset header.

8

Monitor & Maintain

Subscribe to the changelog at /api-docs/changelog. Test against the sandbox environment before deploying API upgrades to production.

Key Outcomes

  • A working API integration with authenticated requests
  • Signal delivery wired into your own application or notification system
  • Backtests triggerable programmatically from your tooling
  • Rate limit handling that won't break under load
1–2 days
Workflows · BitPredict