The 2-day mean-reversion trade, e.g. RSI2 or DV2 has become very popular. Though the performance of this trade has weakened lately (during 2011). So the twenty thousand dollar question becomes: Is the edge of this trade gone or is it about to resume?
With this post I want to share an observation I’ve made by tracking my Trend Strength Indicator (TSI). TSI is a non-directional trend strength indicator. A lower TSI is indicating a tendency to mean-reversion while a higher TSI is indicating follow-through action. Read more about TSI [here].
Let’s start with some chart porn.
Above you see a SPY chart from 1998 until today (Nov. 2011). Prior 2009 TSI spent most of it’s time bellow 1.6. Post financial crisis 2008 it raised to the highest value ever of >2.0 and it allays maintained the level of > 1.6. However what you also can notice is the trend strength decrease lately. I can also confirm that by simply looking at the performance of my mean reversion strategy over the last few month.
Here are some facts.
This is a simple strategy: long DV2B < 50 and short DV2B > 50. Notice the amazing performance in 2008 and the weak period mid 2011. That’s when a lot of people declared the end of this trade.
The same strategy but with a additional filter: TSI < 1.6
Returns stayed about the same while trades have been reduced. Risk adjusted metrics improved as well. These results are without any trading cost. Considering for commission and slippage the difference would be stronger. There are many ways to improve the effectiveness of the 2-day mean reversion trade. Looking at TSI is definitely a good approach to adapt trading style within trading strategies. Let me know your ways of improving mean-reversion trades – send an email to hassler.blog (at) gmail.com
# Frank
sorry if it is a dumb question…what is DV2B?
Hello Glen,
of course, this is a valid question. Sorry for not explaining that upfront.
DV2b is a percentrank’d DV2. DV2 is a very similar indicator to RSI(2). You can read more about DV2 at David Varadi’s blog http://cssanalytics.wordpress.com/
All the best
Frank
Thanks Frank…that was new to me.
Hi Frank,
My comment actually is actually based on what I’ve studied for the CFA, and what I read not long ago on here. Here goes:
Basically, your results confirm what we all knew/hypothesized. Trending markets are a nightmare for the DV2, at least on daily bars. What I will likely do, and even mail you the results if you’re interested, is test/build a system that trades, maybe 8, non-correlate, highly liquid ETF’s. Using the TSI as a filter, we would select only 3-4 ETF’s from our basket of about 8, and apply the basic 50/50 rules to DV2B. The basic idea would be that the DV2 is obviously capable of extracting gains from instruments other than the SPY (increase number of potentially probable profitable trades), and that by using non correlated assets, we subsequently cut max drawdown as well. Maybe after some preliminary data, we could get into some dynamic ranking strategy too. None the less, it’s just a thought that I might try out. I’ve also would be interested in seeing more testing on individual stocks. For instance, maybe the DV2 on the 50 lowest TSI readings in the S&P100 or NAS100.
What are your thoughts on something like this?
Hi Chris,
have you run this test already? Sounds an interesting little project?
Contact me on hassler.blog (@) gmail.com
Frank
Hello chris,
You should go look at Quanting Dutchmans DVI+DV2 system. He uses a similar concept of ranking a portfolio of etfs, only trading the top 3 etfs with the lowest DV2 score.
MG
Frank.
Great post !!
Can u pls upload an excel file with the SPY TSI data for the period that u analyzed ?
Tx in advance,
Ariel
Hi Ariel,
I don’t use Excel for backtesting, sorry.
Frank
Hi Frank.
That’s ok – I’ve just wanted to compare the TSI values that I have to yours, to make sure I’m calculating the indicator correctly.
Do u think that 1.6 as a threshold works for other ETFs as well?
Ariel
Hello Ariel,
I haven’t done a similar test for other ETFs. I use TSI a lot in combination with stocks and there it works.
Regards,
Frank