The Algo Catalog methodology
How we build and ship strategies.
Every strategy in our storefront flows through the same pipeline. No selective backtests, no cherry-picked snapshots. This page documents the stages, gates, and evidence behind each product page.
The pipeline
Concepts move through four process stages, then route to one of four terminal stages. Folders physically move between stages as a concept progresses or is killed. The state of the pipeline at any point is auditable on disk.
Gates and floors
Three threshold checks decide where each concept ends up: an auto-cull against minimum Sharpe and MAR, a Hall-of-Fame eligibility gate, and a lifetime-ROR routing rule.
Data and modeling resolution
All backtests use daily-resolution OHLC data from Norgate Premium (point-in-time index membership, survivorship-bias-free, dividend-adjusted). Norgate does not provide intraday (minute or tick) history, and we do not model intraday price paths. What this means in practice:
- Stops and limits are checked against the day's High and Low. If the range touches the price, the backtest assumes a fill at that price. Live fills may differ depending on intraday liquidity.
- Same-day stop-and-target collisions are resolved with pessimistic assumptions because daily data cannot reveal the intraday sequence.
- Open and close exits are modeled at the session's open or close print. Live MOO/MOC fills are typically close but not guaranteed identical.
- Gaps through stops fill at the open, not at the stop level. Stop-based risk control is approximate, not absolute.
Daily-resolution modeling is appropriate for the holding periods we sell (one day to roughly two weeks). The full disclosure lives in the Risk Disclosure document, § 11.
Robustness artifacts
- Parameter disclosure - total vs optimized parameter count, trades per parameter, statistical adequacy.
- In-sample / Out-of-sample overlay - cutoff date plus Sharpe retention across the boundary.
- 2-D parameter heatmap - production cell highlighted against a sensitivity grid.
- Monte Carlo path band - 1000+ trade-sequence simulations vs percentile bands.
- Trade truncation - how much P&L survives clipping the top 1% and 5% of trades.
The public/private boundary
Public: equity curve, monthly returns, comparative stats, robustness artifacts, cost of ownership, setup honesty, FAQ. Private (post-purchase only): entry conditions, indicator stacks, parameter values, exit rules, position-sizing math. Those live in the .rts file the buyer downloads.
House rule: we never sell a strategy we personally trade live. Premier production stays in-house; the storefront is the validated second-tier work judged honest enough to ship without information-leakage risk.