Following is a video by Dave Evans using the rapid development MS Excel tools that are part of the DV_Indis Research Service to develop a Silver (SLV) Exchange Traded Fund system:
The acronym “LTR” simply stands for the Livermore Trend Ranking. The concept was inspired by literature on Jesse Livermore as well as legendary speculators like Nicholas Darvas. Both traders tended to favor stocks that were predictable and trended smoothly. Livermore was indifferent to direction- he was equally likely to short in bear markets, while Darvas preferred only those stocks that moved up. Both speculators wanted to be able to pyramid on their positions. Pyramiding means that they wanted a stock to “prove itself” by going higher/lower following purchase. At this point they would add to their positions in the direction of the trend. Having tried this strategy a long time ago, I realized that many stocks would simply pull back by the time I added my biggest position—leaving me at an immediate loss. I was beginning to experience the transition from trending markets to mean-reverting markets.
I realized that in order to find stocks that would be suitable for trend-based strategies or pyramiding, I would have to consider the “noise” in their daily behavior. I noticed by looking at price charts that stocks with smoothly trending price closes (and high/lows) seemed to have a higher tendency to follow through—that is an up day would tend to be followed by an up day, and breakouts had higher odds of success. This did not mean that a “trending” stock must go up—often this is a big point of confusion with high LTR stocks. Most people associate trending with going up. In fact a high LTR stock only means that the stock has less noise, it has nothing to do with relative strength or the probability that a stock will go up.
The calculation of LTR is highly proprietary, but is the best single measurement of a stock’s noise factor that I have tested. Intuitively it looks at the close to close volatility with respect to momentum. Academics (and option traders!) have long confused volatility as a pure proxy for risk. There is a big difference between “good” and “bad” volatility. If you look at high momentum or high relative strength stocks they have substantial “volatility” but typically they have less actual risk in most market conditions (with the exception of bear markets). The predictability of momentum stocks makes them easier to trade. The riskiest stocks are in fact are volatile stocks with little relative strength. If you were to buy and hold the most volatile stocks, you would in fact lose money and suffer much lower trade winning percentages than if you bought and held stocks with low volatility. That said, highly volatile stocks can be effectively traded with short term timing strategies. CXO Advisory recently reviewed an academic paper on this same topic: http://www.cxoadvisory.com/volatility-effects/combining-realized-volatility-and-simple-moving-averages/. The caveat here is that the volatile stocks that perform well with moving averages are not the ones with low momentum in relation to volatility. Furthermore, correlation also is a major mitigating factor as I will present in future research. Nonetheless, this academic paper indirectly lends support to the thesis that high LTR stocks tend to trend—and this effect is robust across stocks, not just for large caps.
[A full Livermore Trend Ranking report for both OEX and NDX stocks is provided weekly exclusively to DV_indis Research Subscribers]
In the last post we looked at trading DV2 at extremes in rising or falling 50dma environments. Our goal was to perform an intermediate-term time frame analysis using the 50 day moving average to represent the trend. Today I thought we would look at capturing something different: trading DV2 extremes within different 30-day volatility regimes using DV Percentile Rank Volatility (DVPV) and multiple holding periods. In terms of settings, we will define low volatility as DVPV<0.5 and high volatility as DVPV>=0.5. This gives us a larger sample size to draw from to make conclusions. First we shall examine the results from long trades (DV2 < 20).
Here we see it definitely pays to invest in those high volatility periods associated with high uncertainty vs. those that are relatively calm. Throughout the entire holding period spectrum the high volatility environment beats low volatility for long trades and only begins to converge at the long 20 day holding period. The high volatility trades exhibit a consistent uptrend in average gains per trade as the holding period increases. The low volatility curve is similar but only increases significantly at holding periods higher than 7 days. Both are consistently profitable but the low volatility trades may not be robust to transaction costs. Thus when volatility is low, it may be best to avoid using the DV2, and focus on a more intermediate-term indicator such as the Super-Smoothed Double Stochastic (DVDS).
Taking a look at the risk adjusted measure, the DVR, we find outperformance of trades in the high volatility environment for holding periods under 10 days by a significant margin. However there is a clear deterioration in the DVR for the high volatility curve as holding periods increase while the low volatility trades show the opposite effect with DVR’s improving with longer holding periods. The deterioration for the high volatility trades is likely due to the fact higher volatility also tends to coincide with the SPY being below the 200ma. In contrast the observation that low volatility trades improve over time is likely due to the fact that this environment also tends to coincide with the SPY being above the 200ma.
Next we shall examine the results from short trades (DV2 > 80).
Here we see very different results in comparison to the long side, with generally much less consistency in edge. Shorts in low volatility environments tend to exhibit consitently low positive returns across the holding period spectrum with no discernible trend in P/L, with the edge peaking out at the 4 and 10 day holding periods with an average trade of 0.32%. In high volatility the average gains are higher for the most part except at the long end where holding periods longer than 15 days show negative returns. As a whole, short trades remain still much less profitable than the long trades we looked at earlier. Lets take a look at the DVR to see what insights we can gain:
Looking at risk, we can see that the low volatility trades exhibit a poor reward/risk throughout the entire spectrum, with the optimal result falling within shorter time periods under 5 days. The high volatility trades start off with a high reward/risk for the 1-2 day hold backtests, but decay in an oscillating fashion as the holding period gets longer until the measure goes negative for holding periods greater than 15 days. It is very interesting to note the divergence in the pattern of risk-adjusted versus absolute returns for the high volatility environment. Effectively it may indicate that since volatility is highly cyclic it is not efficient to harvest it over longer time frames.
As we reverse the analysis from the last post, we take a peek at the short side. Looking at the SPY again, here are entries into overbought conditions (DV2>80) relative to the rising or falling 50dma, a gauge of intermediate term trend conditions with a lag.
Here we see again it pays to invest with the trend, as short positions initiated within a falling 50dma did better for all holding periods and stayed positive with gains. The shorts with a rising 50dma were still profitable for most of the holding periods except for past 10 days and in general the return for shorts within a falling 50dma increased with holding period while the other subset did the opposite.
Turning to the risk adjusted outlook, the DVR also shows the same effect with the shorts within a falling 50dma having a higher reward/risk than their rising 50dma counterparts for the entire holding period spectrum. The metric also goes negative on holding periods far out if you’re going countertrend, so you better think twice! Interestingly though, the risk deteriorates through time for both curves.
Once again looking back at some of the previous results and comparing the 50dma to the 200dma subsets, we see clear outperformance in the longer time frame from a return standpoint across the spectrum of holding periods. However, when we look at the risk/DVR metric we find that the falling 50dma subset beats the close below the 200dma group for extremely short holding periods. The difference probably isn’t enough to be statistically significant, but it’s still an interesting observation. Also, while the return increases with holding period, the risk actually deteriorates for both timeframes. Long or Short, it pays to trade in the direction of the intermediate or long term trend.
Now we turn to look at trading in the direction of the trend on a shorter time frame , looking at oversold entries (DV2<20) relative to the 50dma instead of the 200dma to see if the same conclusions hold true. In the last decade, markets have become increasingly volatile and as a result moving averages have become less reliable, so for this test we’re going to use a rising 50 day moving average instead of the close being over the 50dma to define the trend.
While the returns from long trades within a falling 50dma environment are slightly higher for shorter holding periods, you can clearly see that trades with a rising 50dma are much higher for longer holding periods. There is also a clear progression in profitability as the holding period increases for the rising 50dma group.
Looking at our risk metric, the DVR, we can see that for all but the shortest of holding periods the rising trend environment beats the falling trend for long trades. It is also very clear that the risk/reward increases for increasing holding periods when you aligned with the trend while when you trade against it there is a clear deterioration.
Looking back at some of the results from the earlier blog post involving the test within the context of a longer timeframe, we see clear outperformance in the longer time frame from both a return and risk standpoint across the spectrum of holding periods. Once again we find that it pays to trade in the direction of the longer term trend for long positions. But will the same hold true for shorts using the 50dma?
Last post we looked at the absolute and risk adjusted returns for long positions above and below the 200dma, initiated by extreme oversold conditions in the oscillator (DV2<20). This time we’re going to flip it around and look at short positions after overbought readings (DV2>80). Let’s take a look and see what can find out….
No surprise here, average short trades above the 200dma are flat for short holding periods and go negative if you hold on too long. When the trend is your friend below the 200dma, profitability skyrockets up to an optimal holding period of 7 days for absolute returns. Next let’s look at the risk metrics…
Looking at the DVR, we see a clear winner with significant out-performance across the spectrum for the short trades below the 200dma versus those above. Interestingly, both curves deteriorate as the holding period increases, an effect that I would speculate is caused by the fierce rallies seen in bear markets.
Bringing up the rear is the Win/Loss ratio for trades. Here the picture is less clear although short trades below the 200dma still have the upper hand for the most part. This is why we prefer to look at the DVR, a superior way of distinguishing the reward to risk in trades! In the end, trading within the context of the larger trend was shown to be superior in absolute and risk adjusted terms when looking at the short side. But what happens when the definition of trend is changed? How about if we filter by the 50dma? To be continued……
Markets go through different phases, bulls rule one month, bears the next. There are many different strategies to navigating the markets but following the long term trend in terms of the 200dma is a favourite that is helpful in maximizing profitability. Waiting for an oversold condition to enter positions is one way of getting on board. Here we are looking at long trade entries above and below the 200dma using extremes in the DV2 oscillator, initiating on entries below 20. But what should be our exit criteria? Let’s look at the effects of time based exists, holding periods of 1-5,7,10,15 and 20 days. Hat tip to Rob Hanna of Quantifiable Edges (http://quantifiableedges.blogspot.com/) for inspiring the time based analysis and tables!
Interestingly enough the long trades below the 200dma outperform those above the 200dma for short holding periods, presumably because of the high volatility and compressed cycle length associated with market conditions under the 200dma/in a bear market (see http://cssanalytics.wordpress.com/2010/04/07/a-simple-adaptive-change-to-daily-follow-through-to-capture-cycle-changes/). However holding on to those entries above the 200dma leads to much higher gains per trade when holding above 7 days. The old Livermore saying “It never was my thinking that made the big money for me. It always was my sitting.” comes to mind, but only when the trend in a higher time frame is your friend.
Looking into risk adjusted details we see a fascinating picture emerge: as the holding period above the 200dma increases, the DVR, a measure of reward to risk which specifically is the linearly adjusted Sharpe ratio, increases up to an optimal holding period of 10 days while below the 200dma the DVR peaks at 2 days and deteriorates rapidly. This is exactly what you would expect trading countertrend.
Looking at another measure of reward to risk, the Win/Loss ratio we see the same effect, although in a less pronounced fashion than with looking at the DVR.
In the next post we’ll look at the same type of time based exit analysis only with short positions, stay tuned….