Extreme RSI(2) Readings Becoming Less Common
The indicator RSI(2) has received a lot of attention from traders and the blogosphere (we did our own take here and here) as an effective way to trade short-term mean-reversion in the markets.
For the uninitiated, RSI(2) oscillates between 0 and 100, with low readings indicating the market is oversold and high readings overbought. We track extreme RSI(2) as part of the free State of the Market report.
In this report, I’ll show that such extreme readings have become much less common over time, which has implications I think for anyone trading similar short-term MR strategies.
The graph above shows a rolling 5-year average of the percentage of trading days where the S&P 500 closed above 95 or below 5, from 1955.
Clearly since the early 1970’s, the number of days registering such extreme readings has been on the decline, from a high of 28% to about 11% today.
Looking at even more extreme readings paints an even starker picture. The graph below uses the same approach, but only includes days closing above 99 or below 1.
At less extreme levels, this affect begins to flatten out a bit. The next two graphs show readings beyond 10/90 and 20/80 respectively.
At least by this one measure, the market is making less and less extreme short-term highs and lows.
I think the “why” can be understood by understanding how the indicator is calculated. I won’t go into a uber-geeky discussion into calculating the RSI(2), but in a nutshell, the indicator is measuring both the magnitude of price changes (i.e. large changes in price tend to make the indicator move farther in a direction), as well as how many consecutive days the market moves in a single direction.
And as we’ve talked about on a number of occasions (ex. exploring multi-day runs, daily follow-through, and the moving average spectrum), the markets are becoming more contrarian in the short-term. That means the market tends not to move in a single direction for as long, which means that the market tends to register less extreme readings on short-term indicators such as this one.
I don’t think this says anything about the effectiveness of strategies based on indicators such as RSI(2), but it does say that they might trigger a bit less over time.
[Edit: click for a summary of posts related to RSI(2)]
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Evolving Markets, Trading Strategies | 18 Comments
Actually, it does say something about the effectiveness of short-term mean reversion strategies. The reason they work now, but wouldn’t have in the 70′s is that these extreme indicator values now indicate that the market is about to reverse, where as they used to not signal anything. In other words, the fact that they are reached less often means that they are actually tradable extremes. If they become more common, they will become meaningless.
One way to make an adaptive strategy is to look at indicator values relative to thier recent distribution. You can probably get good results that will be consistent over time if you sell when RSI is above 95% of the readings for the past 1000 days and buy when it is in the bottom 5%. The exact level that corresponds to a buy or sell will change over time, but the meaning is still the same.
Yet another thought provoking and useful post. Thank you.
Hi Michael,
Can you explain why did you consider 5 year ‘rolling’ average? What if, I take just year to year percentage change of extreme RSI(2) readings?
Or if, your concern was to have a large enough samples to consider the percentage to be statistically correct, then why did you take ‘rolling’ periods? [just thinking out loud]
Just some questions, perhaps it will be better if you answer them by mail.
Good Work
If you looked at shorter periods, you would come to the same conclusion, but the data would be noisier…I used the 5-year average to smooth out results and make the trend easier to see.
I always look at things “rolling” rather than divided into separate intervals because (IMHO) a calendar year is a completely artificial and arbitrary interval and could hide differences within the year. Rolling averages will capture those differences. It wouldn’t have made a difference in this particular post of course, but as a rule of thumb, I just go with rolling.
Good questions…michael
I see… quite similar effect to MA.
So your method I believe is to find the necessary points with a rolling window [acts as a filter, if I dare add]. In that case, will MA bring out the differences within the years? Or dampen it?
Exact same affect as an MA. Here are some previous posts that used a similar approach:
http://marketsci.wordpress.com/2008/07/17/evolving-markets-dynamic-systems-daily-follow-through/
http://marketsci.wordpress.com/2008/08/10/market-volatility-in-up-vs-down-trends/
http://marketsci.wordpress.com/2008/10/07/weekend-vs-weekday-follow-through-does-friday%e2%80%99s-sentiment-carry-over-to-monday/
Don’t overthink it too much…these aren’t trading strategies, these are attempts to capture the general trend of some characteristic of the market. Once we understand that characteristic, then we can take the next step of (maybe) turning the observation into something tradeable.
michael
hehe…
Sort of a habit to not take anything for granted. I understand this is an investigation of the markets and not directly a trading idea
Good stuff Soham. That’s how it should be. All the best, michael
Before the mid 80s, extreme RSI2 readings did not produce a reliable mean-reversion set-up. In fact it worked the opposite way where an extreme reading meant the trend was likely to continue.
So if one was trading an RSI2 strategy in the 1950s-70s, the buy signal would be what our short signal now is.
Bill Rempel used to have a couple good posts showing the data on this but since he took down his blog, I can’t link to them.
Agreed…covered that in detail in the latter half of this post:
http://marketsci.wordpress.com/2008/12/09/trading-with-rsi2/
michael
Hi,
I am very interested in your more quantitative, systematic approach to the market and wonder what level of schooling and in what field you recommend I study in order to build the skills necessary to trade the market as you do.
Thank you.
Sincerely,
Jake
RE to Jake: good question, but unfortunately, I don’t have a good answer. My formal education is in finance (undergrad and masters) but most of my non-MarketSci-related working career was in engineering (telecom). I think both of those were about as useful (and useless) at shaping my trading.
Finance was great for helping me understand the history and general shape of the markets, but useless for trading or strategy development. Engineering gave me the empirical approach to problem solving, but still useless for trading.
I would say to leverage whatever skills you have or can acquire, but understand that 90% of the important stuff comes from your own independent hard work and failures.
Good luck with the new blog!
michael
Thanks Michael. I greatly appreciate your response and your blog as well!
Peace,
Jake
Michael- thought for a future post…
This interesting research shows that (even though the effect is less pronounced these days) the market spends a disproportionate amount of time (as measured by RSI(2)) in the extremes.
The second chart shows that recently, the market spends 25% of its time with RSI(2) >= 90 or = 80 || 80 (without necessarily a break-through) and buying < 20.
You could do this either mechanically or do an analysis to build confidence levels- i.e. when it breaks 90, you are x% confident that it will keep going, 80 you are more confident, and so on. This would be really interesting if you could get it to work because it would be a way to use RSI as a momentum indicator, rather than a reverting indicator, which it is usually used for.
Thoughts?
RE to Justin: thanks for the feedback – love to hear reader pushing the collective discussion forward.
If I’m understanding you correctly, the problem I think with that approach is that the farther RSI(2) moves towards the extremes, the more likely it is to reverse. Did a post on that here:
http://marketsci.wordpress.com/2008/12/14/rsi-2-follow-ups/
Put another way, I couldn’t see the RSI(2) (at this moment in history) pointing to any kind of continuation when it gets even close to overbought/oversold levels (let’s say…and I’m just guessing here because I haven’t looked at thi question specifically, 70).
Thoughts?
michael
Exactly- that’s what I’d be looking for. It seems that it moves more rapidly through the middle. I agree that the higher it goes, the more likely it is to reverse- so the potential system would look for signs of that reversal, and ride it through the middle until it hits the other side of the RSI.
You could, as you said in your previous post:
“positions in or out as the RSI goes further overbought/oversold”
but I have a feeling this could be expensive, because it seems like stocks tend to follow their momentum for a while before reversing eventually (this is seen in Bollinger bands too- sometimes the stock will bounce off the band, but sometimes it will run right along the edge of it). A better (and more difficult) would be to look for a sign that the reversal is occuring- perhaps a breach of the 80 RSI(2) line to the downside after trading above 80.
RE to Justin: good comments – a couple of thoughts:
1. I think the idea of catching the RSI as it turns and riding it through the reversal makes sense. I’ve never done a post on it specifically on this blog, but in my own tests, it has merit.
2. You’re absolutely right RE: your comment about stocks…that’s why I don’t trade stocks (I only trade indices). Individual stocks are their own unique animal that tend to do all sort of unpredictable things. Indices smooth out some of that unpredictable things. Having said, I agree that buying purely based on a short-term OB/OS indicator can be painful (catching the falling knife and all).
In my own proprietary strategies (http://marketsci.wordpress.com/my-strategies/) I try to trade in all three timeframes – meaning I try to account for the intermediate and long-term states of the market. That’s a subject that I haven’t touched on too much on this blog because it’s starts to get a little too close to my own secret sauce. I leave it to readers’ imaginations to take that next step.
Again, good stuff.
michael