18,048 evaluations | 3 entries
Most of what a strategy does is refuse to trade.
Every mark above is one fifteen-minute evaluation of an ICT/SMC confluence model on BTC, from February to August 2026. Three of them reached a full entry. That is what a confluence filter is supposed to look like, and it is the first thing I can prove about anything I build for you.
I write Pine Script v6 indicators and strategies, then verify them against an independent implementation of the same rules — bar by bar, on a diff — until every disagreement has a named cause. You get the script and the evidence it behaves the way the description said.
What I do
Pine Script v6, written to spec
Indicators and strategies built from your description or your marked-up chart. Alert conditions that fire once, on closed bars, carrying the payload your automation actually expects.
Revisions and repair
Script that repaints, looks ahead, or double-fires. I find the cause and tell you what it was, instead of rewriting around it and hoping.
Ports, with parity evidence
Pine to JavaScript or Python, and back. I run both implementations over the same bars and hand you the diff — not an assurance that they match.
Signal to filled order
Exchange APIs, bracket orders, position sizing, liquidation-aware risk. The stage where a good strategy usually stops being a good strategy.
Work
Seven modules, a sweep-to-MSS-to-BOS chain, and a reference engine written independently so the two could be made to argue.
Trailing stops filling at prices the market never traded, and an objective function that selected for losing geometry. Both found by disbelieving the equity curve.
One distinction in how an exchange represents a stop loss, four bugs that looked completely unrelated until the model was written down.
Things you can look at without asking me
Running code beats a description of running code.