Every screen on ChartMath comes out of one pipeline: propose, validate, ship, monitor. This page describes that pipeline, including the parts whose job is to say no.
We scan a curated universe of 500+ active US-listed equities, plus US futures and crypto, across 7 timeframes from 1-minute to monthly. Curated means liquid enough to trade and clean enough to backtest honestly.
A pattern that worked 6 times is an anecdote. We require a minimum trade count before a strategy result is shown at all, and we display the sample size next to the win rate so you can weight it yourself.
A rule tuned on the same data it is graded on will always look brilliant. That is why step 04 exists: every live screen keeps getting graded on market data that did not exist when it shipped.
Markets change character, and an all-history average can hide that. A setup that prints in a strong uptrend can bleed in a choppy or high-volatility tape. So we tag every historical trade with the conditions it fired in: trend direction and strength, the volatility environment, the broader risk backdrop. When a suggestion surfaces, its evidence is weighted toward how the setup behaved in conditions like today's, not just its lifetime average. What worked in the last bull run does not get a free pass.
Edges fade. Monitoring compares live behaviour to backtest expectations, and persistent degradation flags a screen for review. Review can end in removal; we would rather lose a screen than let it quietly lose credibility.
Backtests are history, not prophecy. Regimes change, fills are not frictionless, and a 70% win rate means losing 3 trades in 10 forever. ChartMath is educational research, not investment advice.
If you want the longer version of how we think, read Becoming Systematic, our three-essay series on sample-size thinking and what a systematic week looks like.