Turn an article into rules you can test
A computer can't “consider” a trade. Before you backtest anything, find out where the article stops and your own decisions begin.
The problem: articles are discretionary
Most trading ideas start as an article, a video or a chart someone posted. They're written for a human who will use judgment: “when MACD crosses and CVD is rising steeply, consider buying.” A backtest has no judgment. Every word like “steeply”, “consider” or “short-term resistance” needs a number or a definition, or it has to be dropped.
The danger is quiet. When you (or Claude) fill those gaps, the result looks like the article's strategy, but much of it is now yours. If you can't see which parts, you can't tell later whether the article's idea worked or your guesses did.
The tool: two labels on every line
The tickshift rules for the MACD + CVD strategy tag every line with where it came from. This is the actual document, shortened:
- From the articleMACD line and signal line, on either 5-minute or 30-minute candles.
- Our choiceLengths 12, 26, 9, on close. No signal for the first 35 MACD bars.
- From the articleDelta is buy minus sell volume, CVD is its running sum, and slope shows strength.
- Our choiceslope = sum(delta, 6 bars) / sum(notional, 6 bars), from −1 to 1.
- From the articleopen_long: flat, golden cross, and CVD steep rising.
- Mirror of the longopen_short: flat, dead cross, and CVD steep falling.
- Not testedNo stop-loss or take-profit, because the article gives none.
Every sentence in the article goes through three questions:
- What does the article actually say? Mark it From the article.
- What must I decide to make it run? Mark it Our choice.
- What did it leave out that I won't make up? List it under Not tested.
The labels matter because every “Our choice” is a knob. In Module 4 you'll see how to judge the best of 226 settings, and why the best of many tends to look better than it really is.
One more decision: when does a rule act?
Signals are decided at a bar's close and filled at the next bar's open.
You only know a 5-minute bar's close when it ends. Filling at that close, or at that bar's open, uses a price you couldn't have traded at when you made the decision. That's look-ahead, and it's the most common reason a backtest is too good to be true.
Practice: sort the sentences
1“Delta is buy volume minus sell volume, and CVD is its running sum.”
2“MACD uses lengths 12, 26 and 9, on the close.”
3“Open short on a dead cross while CVD is falling steeply.”
4“Close the long on a dead cross.”
0 of 4 answered · 0 correct
Check yourself
1A signal is decided at the close of the 10:25 five-minute bar. When is it filled?
2The article never gives a stop-loss. What do the rules do?
0 of 2 answered · 0 correct
Without scrolling up, write the three questions you ask of every sentence in a trading article.
Education only, not financial advice. The strategy shown is a case study of method, not a recommendation to trade.