Premium Market Analysis

Trading Strategies

PSI5 Mean-Reversion Strategy Update – December 8, 2020

The PSI5 mean-reversion strategy in SPY ETF is up 17.3% year-to-date. A performance comparison with two other popular mean-reversion strategies is included.

The PSI5 mean-reversion strategy is not a data-mined but based on a formula from a text in probability theory that models price action.

PSI5 strategy in SPY ETF backtest details

Time-frame: Daily (adjusted data)
Strategy type: Long-only
Market: SPY ETF
Backtest period: 01/04/1993 – 12/08/2020
Commission per share: $0.01
Position size: Fully invested
Position entry and exit: Open of next bar
Current position: Flat since December 2, 2020

Equity curve with monthly returns

The strategy fully recovered from March drawdown and it is up 17.3% year-to-date. The win rate in backtest period is 66.8% with a sample of 1052 trades. Profit factor is 1.60. Average holding period is 3.9 days. Exposure is 43.6% (percentage of time in market.)

Note that the equity drawdown during March did not exceed the historical maximum drawdown of about 23%.

Below we compare the performance of the PSI 5 strategy in SPY to that of two other popular strategies: RSI2 and WR2. Backtest period is from 01/29/1993 to 12/08/2020. Commission is $0.01/share. Equity is fully invested.

RSI2: Buy if RSI(2) < 10 – Sell if RSI(2) > 70
WR2: Buy if two consecutive negative returns – Sell if two consecutive positive returns.

CAGR 9.8% 6.4% 7.5%
Maximum DD -22.6% -30.7% -30.1%
Sharpe 0.68 0.54 0.50
Win rate 66.8% 66.8% 67.7%
Trades (long-only) 1052 306 579
2008 return 54.3% 3.7% -9.3%
2020 YTD return 17.3% -2.4% 5.7%

Note that the default settings for the RSI2 strategy may not be optimum. However, walk-forward analysis showed that optimum selection of settings resulted in severe trade-offs. Specifically, with anchored walk-forward testing based on annual periods, 2020 YTD performance increased significantly at the expense of a large drawdown during 2008. Without anchoring but based on five year rolling periods, 2008 performance was highly negative but YTD performance was positive. WR2 has no parameters to vary. For PSI5, the parameter of interest is fixed.

We disclosed the WR2 strategy a few years ago in this blog. This is a strategy discovered via data mining and that is possibly its biggest risk as opposed to PSI5 that is based on a sound theory of price action dynamics.

The RSI2 strategy was very popular in the past but in the last few years it has lost its robustness possibly due to widespread use.

A few comments about mean-reversion strategies

Mean-reversion strategies are risky because signals go against the short-term trend. Stop-loss application usually destroys profitability and cannot be used effectively.

Mean-reversion strategies are suitable only for professional traders who know how to manage risk at the strategy and trade level.

Charting and backtesting program: Amibroker

Data provider: Norgate Data

More information about PSI strategy: PSI mean reversion strategy information.

If you found this article interesting, you may follow this blog via RSS or Email, or in Twitter