Backtest: optimizer and backtester of trading strategies

In trading, it's easy to mistake a successful segment of history for a working strategy. To test the idea fairly, I created a backtester with an optimizer: it takes historical candles, counts signals, simulates trades, and searches for parameters that don't break with slight market changes. The output reveals the main point: whether the strategy truly has an edge or just looks good on one piece of data.

What's inside:
- 13 indicators for a single contract (Power Engine). Signals are consolidated into one format, with one exit mode - Zero Exit.
- Over 40 metrics: equity curve, MTM, MAE/MFE, PnL, drawdown, trades, commissions, swaps.
- Grid optimizer with constraints, checkpoints, and recovery. It looks not only at the peak but also at neighboring parameters to ensure settings don't break when the period changes.
- Results matrix: 13 indicators across 5 groups of timeframes, totaling 65 testing zones.
- 13 benchmark golden tests to ensure that changes in code do not silently break calculations.

Results are stored in PostgreSQL, with a REST API on top. AI accelerates development, but a human is responsible for the numerical results.

#Python #Backtesting #FastAPI #PostgreSQL #Trading #Analytics #pandas #DataAnalysis #SQLAlchemy #Alembic #numpy #scipy #pytest #Docker #TradingView #PineScript