Last Updated on November 21, 2020 by Oddmund Groette
I’m testing a lot of mean reversion strategies to find the most suitable one. Here is another twist:
- IBS must be lower than 0.25.
- The close must be lower than the lowest the previous 5 days.
- If those two simple criteria are met, go long at the close.
- The exit is at the close 5 days later.
This is a very simple strategy. No fancy tools and hardly any calculations. This is the result from January 2005 until October 2012: 102 trades, 68 winners and average of 0.98% per fill. This is way above the average return for any 5 day period which is 0.11% in the same period. Win ratio is 66.6% and the average winner is bigger than the average loser (617 USD vs 498 using 5000 USD per trade).
Here is a table including the profit factor. I was recommended to include the profit factor by one of the readers (although in my opinion, one can see that this strategy has a high profit factor due to low drawdown(?)):
|Avg per trade||#trades||#wins||Profit factor||Annualized %|
As we can see its profitability climaxes at around 3-7 days. If exit at 5 days this strategy has 10.76% annualized return by just staying in the market for 510 days compared to 2826 days for buy and hold!
Here is the equity curve using exit after 5 days (due to a bug in my software the calculations is done using just 50% of the capital, thus total gain is twice as much as shown):
Here is another illustration showing buy and hold vs the strategy (from January 2005 until July 2012):
The blue line is the strategy and the pink is buy and hold. (This strategy also works very well without the requirement of having to end lower than 0.25 on today’s range. But it works better to include the requirement.)
What happens if we turn it upside down and go short? Using 0.75 as criteria on today’s range gives a lot more fills compared to long. I need to increase it to 0.9 to get a similar number of fills. Over the same period, this has yielded almost 10%. It’s a lot harder to find good short strategies than long strategies.