Self-Adjusting RSI
Original article by David Sepiashvili
AIQ Code by Richard Denning
The AIQ code for the two versions of David Sepiashvili’s Self-Adjusting RSI is given below. To test the effectiveness of the self-adjusting overbought and oversold bands, I created a classic RSI entry rule as follows:
Enter a long position if:
1) Yesterday the 14 day RSI is below the oversold benchmark and
2) Today the 14 day RSI is above the oversold benchmark
Exit all long positions after 5 trading days.
Enter a short position if:
1) Yesterday the 14 day RSI is above the overbought benchmark and
2) Today the 14 day RSI is below the overbought benchmark
Exit all short positions after 5 trading days.
The long entries were tested during a bullish to sideways period of 10/15/2002 to 12/06/2005 using the NASDAQ 100 stock list. The key statistics from this test are summarized in Table 1:
Table 1:
| Strategy-Description | RSI Classic | RSI Self-Adj Ver 1 | RSI Self-Adj Ver 2 |
| Entry Rule Tested | RSI LE | RSI LE1 | RSI LE2 |
| Strategy-Long or Short | Long | Long | Long |
| Market Condition | Bullish | Bullish | Bullish |
| Start date: | 10/15/02 | 10/15/02 | 10/15/02 |
| End date: | 12/06/05 | 12/06/05 | 12/06/05 |
| Number of trades: | 546 | 1,913 | 1,330 |
| Winners | 301 | 1,048 | 713 |
| Losers | 241 | 856 | 610 |
| Neutral | 4 | 9 | 7 |
| Average periods per trade: | 7.18 | 7.23 | 7.29 |
| Average Drawdown: | (2.63) | (2.40) | (3.83) |
| Average Profit/Loss: | 0.42 | 0.81 | 0.55 |
| Winning Probability | 55.83 | 54.78 | 53.61 |
| Losing Probability | 44.14 | 44.75 | 45.86 |
| Average Annual ROI: | 21.28 | 41.12 | 27.47 |
| Annual NDX (Buy & Hold): | 25.10 | 25.10 | 25.10 |
| Reward/Risk Ratio: | 1.24 | 1.49 | 1.23 |
| Using List | NASDAQ 100 | NASDAQ 100 | NASDAQ 100 |
| Entry Rule | LE | LE1 | LE2 |
| Exit: Hold for (periods) | 5 | 5 | 5 |
Both Version 1 and Version 2 of the self-adjusting RSI resulted in better returns than the classic approach using the fixed benchmarks of 30 for oversold.
For the short test, I used the bearish period from 9/01/2000 to 10/15/2002 using the NASDAQ 100 stock list. The key statistics from this test are summarized in Table 2:
Table 2:
| Strategy-Description | RSI Classic | RSI Self-Adj Ver 1 | RSI Self-Adj Ver 2 |
| Entry Rule Tested | RSI SE | RSI SE1 | RSI SE2 |
| Strategy-Long or Short | Short | Short | Short |
| Market Condition | Bearish | Bearish | Bearish |
| Start date: | 09/01/00 | 09/01/00 | 09/01/00 |
| End date: | 10/15/02 | 10/15/02 | 10/15/02 |
| Number of trades: | 370 | 1,512 | 1,330 |
| Winners | 205 | 822 | 713 |
| Losers | 164 | 686 | 610 |
| Neutral | 1 | 4 | 7 |
| Average periods per trade: | 7.34 | 7.28 | 7.29 |
| Average Drawdown: | (3.64) | (3.88) | (3.83) |
| Average Profit/Loss: | 0.73 | 0.66 | 0.55 |
| Winning Probability | 55.41 | 54.37 | 53.61 |
| Losing Probability | 44.32 | 45.37 | 45.86 |
| Average Annual ROI: | 36.35 | 33.17 | 27.47 |
| Annual NDX (Buy & Hold) | (36.18) | (36.18) | (36.18) |
| Reward/Risk Ratio: | 1.35 | 1.28 | 1.23 |
| Using List | NASDAQ 100 | NASDAQ 100 | NASDAQ 100 |
| Entry Rule | SE | SE1 | SE2 |
| Exit: Hold for (periods) | 5 | 5 | 5 |
Neither Version 1 nor Version 2 of the self-adjusting RSI was able to out-perform the classic RSI short entry rule. It appears that the constants K and C need to be increased for the short entry rule. Further tests to optimize the K and C parameters were not performed.
EDS Code for Self-Adjusting RSI:
RSI Self-Adjusting.EDS
|
|
|
|
|
|