I build TradingView indicators and strategies, audit backtests in Python and connect signals to brokers on demo and paper accounts. Every tool states its claim and ships with a way to verify it: look-ahead, repaint, fill and cost checks are part of the work, not an extra.
Fixed scope or hourly, agreed in writing before work starts.
Indicators, strategies and alerts written from your rules, with realistic commission and slippage, a non-repainting build and a look-ahead check you can run yourself.
A review of an existing backtest or strategy: look-ahead and repaint, fill assumptions, costs, overfitting and sample size. You get a written report with what survives and what does not.
TradingView alerts or Python signals routed to a broker on a demo or paper account first, with idempotent orders, reconnection handling and a log you can audit.
Client code, strategies and data stay private. Nothing derived from a client's work is published or reused without written permission. The public scripts and repos on this page are my own work.
Each one measures something a chart can hide.
Measures the gap between Heikin Ashi and real prices, and what the same signal gains when a backtest fills at HA prices instead of the market.
Measures how old the value request.security returns is for up to three symbols, and counts the bars where it was carried forward from an older bar.
Checks one session string in a DST-aware zone and in a fixed offset, and counts the bars each one misses or adds when daylight saving time moves the clock.
Finds the bars where a stop and a target both fit in one range, and settles them with lower-timeframe data instead of an assumed path.
Counts intrabar repaints and ghost signals live, and A/B tests three request.security forms.
Re-prices the same trades at 0x, 1x and 2x costs, with a 95% CI and the break-even cost.
Tests the best month or weekday against a shuffled calendar, so the best of twelve is not mistaken for an edge.
Six classic patterns compared with an all-bars baseline, forward returns with a 95% CI.
An ICT entry model as a strategy, with a built-in look-ahead A/B so the non-repainting claim can be checked.
Gamma-exposure levels from a level string, with a freshness verdict on the open interest behind them.
Runnable repositories, no account needed.
A naive backtest finds an edge in data with none; the audit takes it apart artifact by artifact.
One strategy walked from a naive backtest to an offline order path, and how much of the claim survives each stage.
A 1-second backtest re-executed on real bid/ask and book data, removing one fill assumption at a time.
Idempotent order handling and state resync against the IBKR TWS API, verifiable without an account.