This entry was posted on Friday, September 28th, 2007 at 8:34 am and is filed under Education, chart patterns, market myths. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
I hear many trading “rules” that through my backtesting I have been able to debunk, but the one that I hear most frequently is that during market weakness a trader should focus on the stocks that have held up the best. In a market downswing if you are comparing 2 stocks the one that fell the least or has a higher RSI reading would be the better candidate right? We’ve all heard this but does this theory hold water?
I ran some very simple tests to show you what the data looks like. Here are the rules of this backtest:
Condition : McClellan Oscillator is below -100 (a pullback in the market)
I ran 4 separate backtests on the Nasdaq 100 stocks over the last 5 years. In order to trigger a trade the pullback condition must be met and the individual stocks can be in one of 4 classifications:
RSI under 25
RSI 25 -50
RSI 50-75
RSI 75-100
The 5 period RSI is used in each of these cases. I will now exit the trade after 5 days. Here is a graphic showing the results of that test:
Of principle concern in the test I want to see what the average profit and loss per trade in each of these RSI segments will be. Notice how as RSI increases the average profit declines. The most trades and the most common area for stocks would be in the 25 to 50 RSI reading after the market pullback. Also I should note that I allocated 1% of equity to each trade which allows us to hold every Nasdaq 100 stock. You can see this by the very low exposure readings but this allows me to ensure that I am seeing every trade.
Of particular interest is when you compare the 50-75 group to the under 25 group you will notice about the same number of trades occurred. Yet the average winner in the “weaker” group is 42% larger than the “stronger” group with readings of 0.97% and 0.68% respectively.
I have seen this demonstrated in not only the Naz 100 but also in any other grouping of stocks I have tested whether it is the SP500, Russell 1000, and any random grouping of symbols in not only bull markets but bear markets as well.
So do we have the Makings of a Market Myth? I think we do.
Have a Great Day!
Dave Johnson
September 28th, 2007 at 12:31 pm
You may be comparing apples and oranges. Stocks that hold up well during a market correction, in its strict definition, have good relative strength (usually compared with a benchmark index like SP500 or QQQQ), and this is not the same as its own RSI reading.
For the sake of argument, let suppose the market is in oversold/pullback mode as measured by the McClelland oscillator, but the major indices are still in uptrends and above the 200d MA.
Now, for example, during this period, a very weak stock (as compared with the SP500) that is in downtrend and stays well below the 200d MA may have a bounce that takes its own RSI reading back up to, say 60-75, while still under the 200d MA, and still weak relatively to the SP500. This means it starts to get overbought while in a downtrend, and maybe a good short candidate.
Now, a strong stock in uptrend mode (stronger than the SP500, i.e. its relative strength is higher than that of the SP500) may pull back during the market correction, and its own RSI reading may reach 20-30 level, which may signify it’s close to being oversold, and that it’s a good long candidate.
For example, the SP500 may pull back and have an RSI reading of, say 30, and the strong stock may also pull back and have the same RSI reading, but the strong stock’s relative strength may still be much higher than that of the SP500.
So a stock’s own RSI reading has nothing to do with its relative strength compared to a benchmark. The stocks you measured with RSI under 25-50 are not necessarily the weak stocks, and those with RSI over 75 may not be the strong stocks relatively to the SP500.
So, it’s natural that when an oversold market (still in uptrend) resumes its course, the strongest stocks that were also oversold (low RSI reading) during the market correction may bounce back more strongly, in general.
If you look at the most recent correction, stocks that held up well during this period bounced more strongly than the market. A rising tide generally lifts all boats, but the strong ones outperformed. They all have low RSI readings: AAPL, GOOG, RIMM, BIDU, AMZN, DE, CMI, FWLT, DRYS, EXM, GRMN, POT, OIH…, to name a few.
Note that all of these stocks have outperformed the SP500 for a long time, and during the correction, although almost all of them had RSI readings close to that of the SP500, they were still much stronger than the SP500 in view of the fact that while the SP500 did undercut its 200d MA a few times, these stocks were still way above their respective 200d MA during that time, and their gains up to that point were still way higher than that of the benchmark, i.e. their relative strength were much stronger than that of the SP500.
September 28th, 2007 at 12:43 pm
Ken excellent comments. I don’t try to interpret the results. In a condition where the McClellan is below minus 100 and comparing all the groups - the low RSI group outperformed 5 days out. Thats all that I am saying. Simple as that.
Our swing trading system plays on this edge. And you are right the Relative Strength (not RSI) could certainly be at a higher level. And I could bring the testing even futher by incorporating this. Maybe that can be our next test. If you get a chance send me the testing parameters and I’ll try to get them backtested.
September 28th, 2007 at 1:28 pm
Ken,
I think you’re falling into a common trap of trying to make test results fit into a theory you already subscribe to. While you explained a very reasonable and elegant sounding theory on why the results are what they are, the reality is that it does not matter why. The numbers are in black and white. Remember, when entering the world of system testing, leave any preconceived notions at the door. “Why” something does or does not work is irrelevant. What is important is being able to measure that something works and more importantly being able to measure when it stops working.
- John
September 28th, 2007 at 1:32 pm
I didn’t mean to say the results were incorrect — can’t disagree with the maths. In fact I’m quite happy with the results, because they’re consistent with how I trade as well. I’m not a system/mechanical trader, but it seems intuitive to me that when the market is oversold and starts to bounce, the stocks that are also oversold will benefit more than those that are already overbought (as measured by RSI reading) at that time. This may even be true for those stocks with poor relative strength compared to the benchmark.
I generally prefer stocks that had better relative strength than the SP500 before the correction (especially the leaders), and are oversold (say, RSI(14) between 30-40, or Slow Stochastics(5) less than 20) during the correction.
September 28th, 2007 at 2:08 pm
John, as mentioned in my reply to Dave, I agreed with the results, and didn’t try to interpret them. After a market corrects and starts to bounce, I never trade the stocks with high RSI readings, and always pick my candidates with low RSI readings, consistent with your test results. In addition to that, I just pick the ones that were stronger (higher relative strength than the SP500) before the correction.
Perhaps the confusion arises on my part as to what it really means by “stocks that held up best during the correction”. Dave considered those are stocks with higher RSI readings, and maybe that’s what most people measure as well. I personally consider them to be stocks with higher relative strength than the SP500 during the correction.
Say, the SP500 had 10% gain before the correction, and gave it all back during the slide, now it’s back to breakeven. And a stock with 100% gain that gave back 15% during the correction, although giving back more than the benchmark, still have much higher relative strength. So I prefer these stocks for trading after the correction, as long as they’re oversold by my measurements during the correction.
Apart from terminology, I think we all agree that stocks with LOW RSI readings result in better short term gain after a market correction, just as the results shown.
September 28th, 2007 at 3:29 pm
Ken,
I completely agree with your response and it sounds like you’re trading strategies are very similar to Dave and I’s. I think the myth Dave was dispelling was the tendency for the S&P to have a -4% week and people to point to the stocks that were +1% for the week as the proper ones to buy.
Do you trade an automated system with your strategy?
- John
September 28th, 2007 at 3:30 pm
Ken I think were speaking different timeframes. Were completely speaking a single wave. 5 day RSI and McClellan are a single thrust. A couple day event. So when I speak of “stocks that held up best during the correction†I mean this in with a very short term thrust. Hence the use of 5 day RSI. Pullback would probably be a better term - my apologies for any confusion.
Although when we test the logic you have discussed I think you will be very surprised. I hope you look out for the next post.
September 28th, 2007 at 5:24 pm
John, I don’t have a mechanical trading system, but has begun to develop an interest in system trading. I mostly trade using price/volume patterns and a few simple oscillators (RSI, Stochastics), moving averages and trendlines. I’m beginning to experiment with Wealth-Lab.
I think when it comes to relative strength, another thing that needs to be clarified is the relative gain/loss versus absolute gain/loss. For example, during the August correction, from the July’s intraday high to the August’s intraday low, the SP500 lost about 10%, while AAPL lost over 20% (a bit less if we only consider only the closes.) In any case, going by this measurement, it seemed AAPL was much weaker than the SP500, i.e, didn’t hold up quite well during the correction.
However, ignoring the fact that AAPL has outperformed the SP500 for several years, just from the beginning of this year to before the correction, AAPL had a gain of ~72%, while the SP500 had ~9%. At the bottom of the slide, AAPL was still ahead 38% for the year, while the SP500 had a minor loss.
In other words, AAPL gave back ~47% of its total percentage gain since the beginning of the year, while the SP500 gave back 100% of its respective gain, and then some. So even though the absolute % loss of AAPL (from the July peak) during the correction was almost double that of the SP500, for my purpose, I still consider AAPL relative strength was much higher than that of the SP500. And when both got oversold (by RSI/Stochastics measures) during the August slide, I preferred to trade AAPL for the bounce back. AAPL has now exceeded its July high (in fact, it’s at all time high), while the SP500 is still below its July high.
Of course, this is just a specific example, but lots of other stocks with much higher relative strength than the SP500 before the correction has continued to outperform it after the correction. That’s why I prefer those candidates.
I wouldn’t touch AAPL or any other stock if it had had high RSI/Stoch readings before the bounce started (even if they had superior relative strength.) Although some overbought stocks may continue to remain overbought for a long time, the odds seem to favor strong stocks that are oversold in an oversold market that is starting to bounce.
Dave, yes, we’re probably speaking different timeframes. And I probably won’t be too surprised if your backtesting would contradict my “logic” — it’s nothing more than some experience in reading chart patterns. In that case, I’d prefer to believe the statistics of your test results when it comes to short term trading.
I’m more of a position trader, holding for weeks if not months. I also have long term investment in a number of core holdings, such as AAPL, GOOG, RIMM, MA, CME, ISRG, GRMN, POT. Sometimes I trade around these core holdings, but I think I’d be terrible at very short-term/day trading. I tend to hang on to a bull market’s leaders for as long as I can, and get out of losing trade quickly. And I think part of the reason my long term investments are still doing well is because we’re probably still in a bull market until proven otherwise.