I first learned of the DV2 Indicator in the paper: “MR Swing.”

Then I looked through the following research to see whether or not the indicator had been successfully applied.

And after, I found the indicator coded for AmiBroker at this Quanting Dutchman post.

I tested the basic system with SPY data from Yahoo! Finance from 1/1/2000 – 7/30/2012. The rules were:

- Buy/Cover: DV2 < 50
- Sell/Short: DV2 > 50

The results are below. I used the close price of the day the trading signal was generated.

The results of this indicator are quite good, especially for such a simple system.

**Performance using a 3x Leveraged ETF:**

Next, I decided to test the system on the 3x Leveraged ETF SSO , while still using the buy/sell signals from SPY.

The results are below. I used the close price of the day the trading signal was generated. The data from was from Yahoo! Finance from 6/22/2006 – 7 /30/2012.

The results of this system seem good from it’s CAGR, but upon closer inspection one can see that all of the returns are derived from 2007, 2008, and 2009 and that the returns these past years have been lackluster (similar to the performance on SPY, but the effect is magnified. The system could be failing as of late, or it could be that the system performs well under certain regimes.

**Performance under Bull/Bear Regimes:**

To classify Bull and Bear Regimes I used the RSRank Indicator described by Jeff Swanson in this post at System Trader Success.

- Bull: RSRank > 0
- Bear: RSRank < 0

Green = Bull Classification

Red = Bear Classification

Then I test going long only or short only (using the rules from above) in these regimes. Same conditions as the test above (data, dates, etc). The performance is:

- Bull & Long: CAGR = 4.52% MDD = 20.44%
- Bull & Short: CAGR = 2.61% MDD = 12.99%
- Bear & Long: CAGR = 8.23% MDD = 20.88%
- Bear & Short: CAGR = 7.05% MDD = 22.59%

Since there are not large discrepancies between the different regimes, it may not be wise to conclude that RSRank is a statistically significant method of classifying regimes or that bull/bear regimes do not affect the performance of the DV2.

**Performance under Trending/MR Regimes:**

**TSI**

The TSI is an indicator created by Frank Hassler from Engineering Returns and David Varadi from CSS Analytics. The formula can be found here. and The classification for Trend Vs MR with the TSI()

- Trending: TSI()>1.6
- MR: TSI()<1.6

To test, I used the same conditions as the test above (data, dates, etc).

- Long & Short When Trending: CAGR = 2.45% MDD = 42.67%
- Long & Short When MR: CAGR = 21.02% MDD = 20.88%

There can be an informal conclusion drawn from this (but not a formal one since this isn’t exactly the most rigorous test). The TSI can separate the market into regimes which the DV2 systems performs well/poorly. Whether or not this has anything to do with a MR or Trending environment is still questionable.

**60-Day ADX**

The classification for Trend Vs MR with the ADX

- MR: ADX(60) > 10
- Trending: ADX(60) < 10

I understand that traditionally, a higher ADX signifies a stronger trend, but for some reason when I looked at the chart of the ADX(60) it seemed that at high values, it signified a range-bound market, so I decided to use high values as a signal for a MR market. To test, I used the same conditions as the test above (data, dates, etc).

- Long & Short When Trending: CAGR = 3.55% MDD = 39.13%
- Long & Short When MR: CAGR = 19.82% MDD = 25.92%

This has results similar to that of the TSI() even though the times that it signifies as Trending of MR as that of the TSI.

**60-Day R-Squared**

The classification for Trend Vs MR with the R-Squared (as a function of the close and time)

- Trending: R-Squared > 0.5
- MR: R-Squared <0.5

Test results:

- Long & Short When Trending: 9.86% MDD = 22.92%
- Long & Short When MR: CAGR = 12.94% MDD = 31.73%

Breaking it down into regimes through R-Squared doesn’t seem to work effectively (there isn’t a significant different between the CAGR or MDD between MR and Trending states).

**Performance under Volatility Regimes:**

I will test performance of this basic system under the following measure of volatility:

- 60-Day Standard Deviation of Daily Returns
- 60-Day Standard Error (Normalized by dividing the standard error by that day’s closing price)

**60-Day Standard Deviation of Daily Returns:**

The classifications for High-Volatility and Low-Volatility with the 60-Day Standard Deviation of Daily Returns

- HV: 60-Day Standard Deviation of Daily Returns > 0.01
- LV: 60-Day Standard Deviation of Daily Returns < 0.01

To test, I used the same conditions as the test above (data, dates, etc).

- Long & Short When HV: CAGR = 17.70% MDD = 26.14%
- Long & Short When LV: CAGR = 5.42% MDD = 32.12%

The discrepancy between CAGR and MDD between these two regimes seem to be significant enough to warrant further analysis of volatility regimes.

**60-Day Standard Error:**

The classifications for High-Volatility and Low-Volatility with the 60-Day Standard Error

- HV: 60-Day Standard Error > 0.02
- LV: 60-Day Standard Error < 0.02

To test, I used the same conditions as the test above (data, dates, etc).

- Long & Short When HV: CAGR = 15.08% MDD = 25.92%
- Long & Short When LV: CAGR = 7.81% MDD = 30.41%

Similar results as using the 60-Day Standard Deviation of the Daily Returns.