Out-of-Sample Tests of Bullish Regime 2-day RSI Signals

A reader suggested: “It would be nice to see a study of a low 2-day Relative Strength Index (RSI) as a technical indicator as suggested by TradingMarkets.com. “We focus on exchange-traded funds (ETF) to perform two simple tests of 2-day RSI trading rules as described by TradingMarkets.com in “How to Trade ETFs: The 2-Period RSI and Entry Strategies for Traders” and “The Improved R2 Strategy: 84% Correct with Just 6 Rules”. For the test of the first set of rules, we arbitrarily select the Technology Select Sector SPDR (XLK). For the test of the second set of rules, we use SPDRs (SPY) as specified. Since the in-sample rule selection period described in the latter article extends through 2006, we conduct the tests with subsequent data to avoid data snooping bias. Using daily adjusted closing prices for XLK and SPY over the period 3/20/06 through 3/17/09 and the Stockcharts.com RSI template, we find that:

For the first test using XLK, we use the following rules:

1. XLK must be above its 200-day moving average on or after 1/3/07 (outside the apparent rule selection sample period).
2. When the 2-day RSI indicates oversold by dropping below 30, buy XLK at the close.
3. Since the first article does not specify an exit rule, we borrow the exit rule from the second article: when the 2-day RSI subsequently indicates overbought by rising above 75, sell XLK at the close.

These rules generate 24 round-trip trades of XLK during 1/3/07-3/17/09, with statistics as follows:

• 23 of the trades trigger in 2007, and only one triggers in 2008.
• Average holding period is about 1.5 weeks.
• 17 of the 24 trades (70%) are profitable, before trading frictions.
• The average return per trade is 0.25%, before trading frictions. In other words, profitability based on average return requires keeping the combined effects of round-trip transaction fees and bid-ask spreads below 0.25% per trade ($25 on a $10,000 trade).
• The standard deviation of trade returns is 3.4%, with a high of +2.9% and a low of -13.4%.

With reasonable trading frictions and trade sizes, these results probably indicate net unprofitability.

For the second test using SPY, we use the following rules:

1. SPY must be above its 200-day moving average on or after 1/3/07 (outside the apparent rule selection sample period).
2. The 2-day RSI on Day 1 must close below 65.
3. The 2-day RSI on Day 2 must close lower than Day 1.
4. When the 2-day RSI on Day 3 closes lower than Day 2, buy SPY at the close.
5. When the 2-day RSI subsequently rises above 75, sell SPY at the close.

These rules generate just 12 round-trip trades of SPY during 1/3/07-3/17/09, with statistics as follows:

• 12 of the trades trigger in 2007. None trigger in 2008.
• Average holding period is about 1.5 weeks.
• 10 of the 12 trades (83%) are profitable, before trading frictions.
• The average return per trade is 0.69%, before trading frictions. With reasonable trading frictions and trade sizes, these results probably indicate net profitability. However, there have been no trades signalled since 2007.
• The standard deviation of trade returns is 3.4%, with a high of +3.1% and a low of -7.7%.

In summary, limited evidence from out-of-sample tests of TradingMarkets.com 2-day RSI trading rules for ETFs are mixed regarding a belief that they are reliably profitable. The reported in-sample results may have substantial data snooping bias.