Be sure to read part one for background information.
The last post looked at recent performance using RSI2 signals to trade the SPY. What I want to do is to compare recent performance against past performance to see if RSI2 is losing its edge.
The charts below demonstrate how effective various RSI2 threshold have been at timing SPY over the time period from 1.1.2000 to 1.1.2010.
Using Low RSI2 Thresholds to Trade SPY
The above graph shows very positive average performance over the next 10 days. All thresholds finish in positive territory, and only one dips into the negative, but only for a day or so. The upward bias is obvious.
Also note how well RSI2 < 2 performed, gaining better than 1% on average after 5 days.
Now lets look at what happened if we bought SPY following a high RSI2 reading…
Using High RSI2 Thresholds to Trade SPY
The above graph shows a clear downward trend. Similar to the low RSI2 results, an extremely high, i.e., 98 or greater RSI2 reading generated a significant downward bias over the period tested.
What the above results reflect is the tendency of the market from 1.1.2000 to 1.1.2010 to revert to a short-term mean when RSI2 was either low or high. In fact, extremely low or high RSI2 readings resulted in stronger mean-reversion over the next 5 -8 days.
Now Let’s Re-visit Recent RSI2 Performance on SPY…
I have re-posted graphs from the part one showing recent (1.1.2010 to 11.11.2011) RSI2 performance so that we can compare them to the longer term graphs posted above.
Using Low RSI2 Thresholds to Trade SPY
In contrast to the longer-term tests, recent average performance has a tendency to degrade into negative returns. True, some of the thresholds have generated returns greater then 1% from 3 to 7 days after the signal, yet all but one finish negative after 10 days. I suspect the higher average returns are due to the high volatility of the last 2 years.
Note that the most extreme readings of < 2 performed the worst over the next 7 days. This is exactly opposite of the longer-term tests.
Using High RSI2 Thresholds to Trade SPY
In contrast to the longer-term tests, the last 2 years have found that high RSI2 reading have led to consolidation, small gains and small losses. We see that the most extreme readings, those over 98, have led to small gains after 10 days. Again, visually compare the two graphs and it is easy to see how high RSI2 readings have not led to pullbacks the same was as they did from the period of 2000 – 2010.
What Does It All Mean?
The RSI2 edge appears to have eroded or be eroding. However, this does not mean that an extreme RSI2 reading is no longer useful. To the contrary, an extreme reading may now be signaling a continuation of the short-term trend, rather than signaling a reversion to the mean. As I mentioned in the previous post, decades ago, RSI2 did not work well as a mean-reversion indicator.
The other point I wish to raise is that the market may be trending more in the short-term, before mean-reverting. Perhaps different RSI settings will catch this, for example, RSI3. Or, perhaps we need to be looking for short-term trend strength indicators. Having traded mean reversion setups for quite some time, I would love to hear about your favorite trend-strength indicator. I am not well-versed in that regard.
All thoughts and comments are appreciated.
For RSI(2) trading, I have found that Larry Connors’ “rules” and position sizing methods improve the edge a bit…especially his TPS set-up.
Disclaimer: I have not run any back testing on these rules, but I have traded them with some degree of success and failure.
The TPS rules are simple:
1.) Only go long if the ETF/Stock is trading above its 200-day SMA. Reverse for shorting.
2.) Wait for the RSI(2) to close below 25 for two days.
3.) Scale into the trade. Example: Say your total long position is 1000 shares. When the RSI(2) has closed below 25 for two days, by 100 shares or 10% at the close. If price closes below the previous entry the following day, buy 200 shares or 20% at the close. Add 30% and 40% if the ETF/Stock continues to close below the days previous entry.
This way, your largest position, 40% in the example, is at the most oversold spot.
4.) Sell when the RSI(2) closes above 70.
It’s counter intuitive. You’re adding to a losing position and traders are told from day one that you don’t do that. But here, I have found breaking that axiom helps small reversion systems.
Also, using the 200-day SMA as a filter helps you choose ETFs/Stocks that are trending in the general direction you wish to trade.
Since the RSI isn’t a trending indicator, taking overbought readings and trying to short against the prevailing bull trend is a low probability endeavor.
I’ll add this and wrap-up: Larry doesn’t believe in using stop losses so his back testing is without stops. I wouldn’t do that. I use a “fail safe” stop at the 200-day SMA in case the markets begin to waterfall as they did in August of this year.
Scaling into dip buying (aka doubling down) is a recipe for disaster.
-Source personal experience. Specifically from RSI systems.
Your point about some kind of MA or upward trending filter is spot on. And I think Wood and I have had this conversation several times. Dip buying is very profitable in bull markets, but we have not seen one in 10 years (ok maybe that is an exaggeration). Sorry for the long-winded reply. My point is that RSI systems work best with a good filter. I personally like double exponential moving averages better than SMA because they isolate trends better. The question then comes down to the period. Of course it would be interesting if someone else knew a better filter altogether. Also, I have just altogether discounted using RSI as an indicator for shorting, but I would be interested in hearing anyone’s experience.
Wood thinking about breaking out that chili recipe this Sunday for football. It was a huge hit last year!
Red, entered the chili into a cook off last weekend. I didn’t even place. There were 17 entries and the way they voted for 1st, 2nd, 3rd, was screwy. 2 of the 3 that placed were pretty bad, in my opinion.
Keith, my experience and testing has shown that it is best not to use stop loss with mean reversion set ups. Yes, it can be deadly not too. I think though that position-sizing is so important because it can allow you not to use stops and still stay in the game after some bad beats.
Thanks Wood. Great stuff.
I agree that scaling into a losing position is deadly if you’re a long term investor or trend trader, but I haven’t had many problems with short swings. It works very well actually. You need to be disciplined, for sure, and options or a fail safe stop loss will help.
A Moving Average filter or another indicator as a filter is key. Not taking overbought shorts when the your ETF/Stocks is trading above the 200-day MA (EMA or SMA) will help reduce a number of unprofitable trades.
Having a dynamic profit target is also helpful. Connors suggests using a close above the 5-day SMA or an RSI(2)close above 70 (longs) or below 30 (shorts).
Another way to improve an RSI(2) system is to enter the position a certain percentage down (longs) or up (shorts) from the previous close. This is another Connor idea.
Example: say your using the RSI(2) 90/10 system and you get a long signal on a stock that has a closing RSI(2) value of 7.00. It’s trending above its 200-day SMA as well. The next morning, you put in a limit buy order 3% below the previous close.
Buying intra-day weakness brings the ETF/Stock into an even more oversold condition potentially increasing the odds. You exit if it closes above the 5-day SMA (sell the following morning).
I do agree with Wood that the RSI(2) has lost some of its edge.
Sorry…the above comment was suppose to be in reply to Redshark. My goof.
Keith, interesting ideas about entry and exits for trades. I guess my overall question, and to the point that you in Wood are making that RSI has lost it’s edge, is has RSI really lost it’s edge or is it simply a victim of the Lost Decade of investing? My thought is that RSI2 only really preforms in strong bull markets.
Keith, I’m probably going to code up the connor’s idea and test it out. Thanks!
Red, RSI2 worked very well over the last decade. I do think it has lost some its edge recently. Volatility has been high, which is typically good for mean reversion. Even so, RSI2 is not doing very well.
Cool deal. It will be interesting to see if any of those ideas improve the results. Thanks for the great work as usual Wood.
I understand what you’re asking Redshark, but I don’t think that’s the case.
RSI(2) systems were heavily developed and traded through out much of the past decade with great success. Traders like Wood here, Michael Stokes over at the MarketSci blog and Larry Connors have a lot of back testing data that shows RSI(2) systems traded very well up to around 2008-ish. RSI(2) even held up well through the Dot Com bear market if you filter using a Moving Average.
What seemed to dull the RSI(2) edge was the 2007-09 crash. Things that changed:
? Open intervention by the Fed (QE1, QE2 and now “Operation Twist” or QE3).
? Historic volatility. The market is overshooting with more frequency. Just look what happened the past August and then again in October. That October bull move is nuts.
? The markets are driven by Macro headlines more now and technicals, while still functional, are being overshadowed.
? The proliferation of ETFs, including the 2x/3x leveraged ETFs, are driving the market into a singularity of sorts.
I thought the RSI(2) traded well for much of ’09 and into 2010, but as Wood points out, it hasn’t done well at all this year. Even filtering with a 200-MA didn’t help it much as you would have taken a sizable loss in October had you gotten short.
I do think the RSI(2) and frankly, any system designed around a single technical indicator produces better results in a bull market or an upward trending market. But that’s probably because bull markets tend to ignore news headlines whereas bear markets commonly knee-jerk to every news headline that hits the wire.
Bottom-line is that reversion to the mean by using the RSI(2) has dulled or changed recently so how do we improve it to adapt to the changing market conditions? Or will the market right itself in due time?
Regarding trend-strength indicators, IIRC you’ve got some sort of “proprietary” uptrend determinant in your market overview (SPX). If the calculation you are using spits out a number and “uptrend” is greater than that number, then determining the strength of that uptrend could be achieved by a PERCENTRANK calculation over the past x periods, no?
Kill, yes, the percentrank should work well for that. Some of David Varadi’s indicators do just that.
Call me stupid, but was is IIRC?
IF I Remember Correctly.
And thanks for these RSI studies.
Thanks for doing this study, I was backtesting my RSI2 long screen recently for various periods in ’10 and ’11, and got flat to poor performance except in very solid uptrends like the Oct.4 ’11 rally. Then I remembered Scotty’s “Didja try reversin’ the polarity, Jim?” and ran the same ‘long’ screen as a short strategy and got mostly pretty good results in markets including shallow ups, shallow or strong downs, and even some sideways chop. Didn’t work in strong uptrends of course. I guess the algos own RSI2-long now, but perhaps there’s still a play in there somewhere.
The only way money can be extracted out of markets is by taking speculation out of speculation(means trader has to be neutral in market direction).Good way to do it would be–Assume S&P 500 is at 1300 & daily trading range is 13 points.That is 6 1/2 points up/down over say 1000 runs.
Volatility expands transiently in every time span & that is when mean reversion is the best.6 1/2 into 3=19 1/2.My method would be sell short at 1320 take profit at 1311 & stop loss at 1329(as buy order next delivery month).Reversal order again is subject to same rules.Trading is always counter trend and continuous in both up/down markets.There is no input from the OBSERVER.Every thing is based on opening price every day & volatility has to expand at least 250%.Empirical evidence shows that this happens al least 5 times more often than SD numbers lead us to believe.That is why option writers go belly up lot sooner than they are led to believe.Trader has to make money from random motions of prices of the underlying asset.There is no other salvation.Even the creation of UNIVERSE is just combination of zeroes/ones all at random.ONLY after the creation universe operates subject to the PHYSICAL LAWS.ALL the miseries for trader come from guessing the next market move.I would be glad to communicate with any one to show that THIS works & also invite your input.email [email protected].I am a private trader & I sell nothing-just a trader day in & day out.The method can be applied to any asset in any MARKETS(stocks,bonds etf futures commodities currencies indices).ALL comments from serious practioners would be answered even through email