Rule-Based Trading: Testing RSI(2) and Trend Pullbacks Honestly
Two published rule sets—Connors’s RSI(2) pullback and Raschke’s ADX “Holy Grail”—show the difference between buying dips inside a trend and buying the resumption of a trend. Here are the rules as published and the checks that keep a backtest honest.
By PYE Markets • Published 2026-10-19 • 7 min read
Key points
- Indicators are easiest to evaluate when they are written as exact if-then rules that can be tested on historical data, rather than read by eye.
- Two common families of rules are trend-following/pullback rules (for example, Raschke’s ADX > 30 and 20-EMA pullback, or a 20-day breakout) and trend-filtered mean reversion (for example, Connors’s RSI(2) rule: price above its 200-day moving average, 2-period RSI below 10, exit on a close above the 5-day moving average).
- A backtest only means something if it includes realistic trading costs, avoids repeatedly tuning parameters to the same data, and holds up on data that was not used to build it.
- RSI(2) Rule (Connors): Above 200-Day MA & RSI(2) < 10
- Holy Grail (Raschke): ADX(14) > 30 & 20-EMA Pullback
- Backtest Check: Costs + Out-of-Sample Data
Technical indicators are calculations on price and volume. Used loosely (“RSI looks low”), they are hard to evaluate. Written as precise rules with a trend filter, they can be tested on historical data before any money is at risk—although a good backtest is evidence, not proof.
1. Connors’s RSI(2) Pullback (Mean Reversion)
In Short Term Trading Strategies That Work (2008), Laurence Connors and Cesar Alvarez describe buying short, sharp pullbacks only while the broader trend is up. The published rules are:
- Trend filter: the closing price is above its 200-day simple moving average.
- Trigger: the 2-period RSI closes below 10 (lower readings, such as below 5, are more selective).
- Exit: sell on the first close above the 5-day simple moving average. The published exit is this signal, not a fixed stop; any catastrophe stop a trader adds is their own risk decision, not part of Connors’s rules.
2. Raschke’s “Holy Grail” (Trend Pullback)
Instead of buying dips, a trend-pullback rule waits for a strong trend and buys its resumption. In Street Smarts (1995), the Holy Grail setup requires the 14-period ADX to be above 30, waits for price to pull back to the 20-period exponential moving average, and enters with a buy stop above the high of the pullback bar (the reverse for short trades).
3. Keeping a Backtest Honest
It is easy to adjust indicator settings until a backtest looks smooth, and then see it fail live. Ernest Chan’s Quantitative Trading (2008) discusses the main defences against this data-snooping bias: keep the number of adjustable parameters small, include realistic commissions and slippage, and test on out-of-sample data that played no part in designing the rules. Checking that the same parameters work on related markets (for example SPY, QQQ and ES) is a further practitioner check.
Two Published Rule Sets Compared
| Element | RSI(2) Pullback (Connors & Alvarez, 2008) | Holy Grail Pullback (Raschke & Connors, 1995) |
|---|---|---|
| Trend filter | Close above the 200-day simple moving average | 14-period ADX above 30 |
| Entry trigger | 2-period RSI closes below 10 | Buy stop above the high of the bar that pulls back to the 20-EMA |
| Exit | First close above the 5-day simple moving average | Trader-defined; often a retest of the prior swing high |
| Typical trade profile | Designed for frequent, short trades | Designed to catch a continuation of an existing trend |
Both rule sets are precise enough to test. What matters is testing them honestly—with costs, with few parameters, and on data that was not used to build them—before drawing conclusions.
Editorial Disclosure: Rules are as published in Laurence A. Connors & Cesar Alvarez, Short Term Trading Strategies That Work (2008), and Linda Bradford Raschke & Laurence Connors, Street Smarts (1995). Backtesting cautions draw on the data-snooping and transaction-cost discussion in Ernest P. Chan, Quantitative Trading (2008). General information, not personal trading advice.