Over the course of the next few weeks or so I will be charting the historical performance of buying SPY and other ETFs based on various levels of RSI2 readings.
This first post takes a look at SPY, using a RSI2 level of less than 10 to initiate buys at the close. Each trade is closed at the close 5 days later.
Click the chart to enlarge…
I have created a rolling 10, 20, and 50 trade average and plotted them according to the dates each trade was closed.
On a first glance, it appears that the 10 trade average may be able to be used to time RSI2 trades so that the indicator is used when there is a greater likelihood of it being effective.
While the 10 trade average doesn’t suggest a significant breakdown of RSI2<10 effectiveness on SPY, it does show that performance has increased in volatility. It also shows that there are many times when taking just a few RSI2 trades would result in compounding losses. Again, it may be better to wait for the 10 trade average to fall beneath 0% in order to take higher probability trades. And, as with many mechanical setups, the 10 trade average shows the important of the law of averages. Quitting after a string of losers likely means quitting right before a string of winners.
The 50 trade average does show that the effectiveness has waned over the past 15 years or so.
I will be happy to hear any of your thoughts or questions in the comments section.
I think this is a good place to jump off. There are many variations on this theme. I intend to examine less extreme RSI2 levels and different exit strategies. As I noted in the introduction, I think it is important to run the same tests using QQQ and possibly IWM or other country-specific ETFs.
One final note: MarketSci used to run a fantastic check-up on the effectiveness of mean-reversion. I don’t think he has posted about it this year, but it is worth pulling up the previous posts as his analysis was very interesting.
Thanks Wood, some good ideas there. Here’s a few questions:
Where does RSI2 < 10 stand in percentile or standard deviation terms relative to something like a rolling (1 or 3 months?) average? I've seen a rule of thumb that you want to be at 10% or less for these types of systems (ie RSI2 gets under 10 only 10% of the time).
What do you do when you get multiple signals less than 5 days apart – extend the trade?
Looks like we'd be doing a lot of trdes recently.
Good questions Bozo.
1. I’ll have to check. Never looked into that.
2. Multiple signals – the way I structured this was that only the 1st signal was traded. All other signals were ignored as the system was assumed to be in the trade. I could look at on the time when it was beneath 10 the previous day as well, or even for two previous days. Might be helpful. Probably, for that type of strategy, you’d want to scale in, but that is more difficult to represent in this type of study.
Great study as always. I really enjoy your work. Couple of thoughts.
1. I think you would find it interesting to run this through a trend filter (only take it if above the 200 day or golden cross)
2. The cousin of this is RSI > 90 when the trend filter is short. A little more erratic, but same idea.
3. Your findings make a lot of sense. I run a variation of this where I take RSI(2) 70 (or some other up “pulse”) as opposed to a time exit. Makes it easier if I want to switch from single to multiple entries as last poster suggested – I can exit everything at the same time. What I find is that when there is a drawdown after a closed trade, the probability usually increases for subsequent trades (and that effect is exacerbated if multiple trades show drawdown).
Cheers,
Tom C
Playing the devil’s advocate here..each day I get more and more focused on understanding Alpha Strategies X Timing Strategies…putting some ingredients to work, it is not that difficult to find strategies that beat Buy&Hold, and it is about what I use to name Timing ones. Unfortunately, I can’t say the same regarding Alpha ones…90 (sell) and 10 (buy) RSI strategies don’t beat the B&H including filters or not…What I would say here is how trick it can seems to have so simple trading rules, like the RSI, working for so long time, they all are based the B&H market stylized fact!
*based on