A Trader Journal

We hear often about high correlations in stock market but not much about ETFs as a driver of high correlations. One would probably come across more media/blog bytes on Risk On-Risk Off etc than ETFs impact on correlations. 

ETFs had $1.2 trillion in assets under management in early 2012 and is one of the fastest growing segments. So it is likely that ETFs continue to accumulate more assets under management and along with that increased impact on underlying asset prices as well. 


When shifts happen, some adapt while others fight it . I think one way to check whether a methodology is fighting or floating with this ETF tide is to check for things like - (a) are the new opportunities sparse/decreasing relative to past? (b) does the methodology require lot more complexity to accomplish same thing that in past was simple? and (c) are the profits more harder to come by relative to past? If answer is Yes then I would imagine the   methodology and trader are fighting the tide. Please feel free to disagree/comment.
 
Recently I came across an interesting paper on ETFs and Asset return correlations. Following are some highlights from the paper. 

Why ETF's drive the asset correlations?

• ETFs have a greater potential to affect asset correlations than mutual funds for several reasons. First, traditional mutual funds have some leeway on where to invest their money, and must typically keep some cash on hand for redemption. ETFs, on the other hand, are created in units which must contain the appropriate portfolio. Each time a unit is created or destroyed, the stocks in that ETF portfolio potentially trade together.
• The second reason that ETFs can drive correlations is the arbitrage that they make possible between the price of the ETF and the price of the underlying basket of shares. Arbitrageurs are likely to favor ETFs because, unlike mutual funds, they are easy to short and quick to trade. Now when the basket of shares is bought or sold together for arbitrage purposes, this places demand on all of the stocks together, which in turn increases correlations.
• ETFs, by making it easier to trade stocks with similar characteristics for investors, they acerbate co-movement among stocks that share similar characteristics. By similar characteristics, I mean like size based (small cap, large cap...) or style based etc.

Findings:

• The paper finds that the more an ETF owns the market capitalization of stocks in its portfolio, the more the stocks in that ETF portfolio tend to move together in the subsequent month. 
• An ETF's turnover is another strong determining factor in driving the correlated movement of stocks that make up its portfolio.
• Another finding from the paper is the more a stocks market cap is owned by ETF's, the more that stock co-moves with the market in the subsequent month.
•  Similarly the weighted average turnover of ETF's that owns the stock is a strong factor in driving the co-moves of the stock with the market.

I would imagine some would agree with above views and others won't. We learn more when our views differ. So please feel free to let me know your views. I have not yet figured on how to create a quant test as well as capitalize on these findings. My one gripe is the paper could have chosen better metrics for results and also presented in a more reader friendly manner.

If you are interested in reading the full paper, following is the link to the academic paper - ETFs and Asset Return Correlations

Wish you all good health and good trading!

I came across this viewpoint which was scheduled to appear in Winter 2013 issue of Journal of Portfolio Management. You can find the link to article at the end of the post.

The core idea of the article is to see portfolio risk as a seamless continuous curve composed of 4 distinct regimes. These risk regimes are identified by the potential size of the losses a portfolio can incur. For example, one way to define these 4 risk regimes is (0% to -5%), (-5% to -15%), (-15% to -35%) and (-35% to above).  If we approach our portfolio risk management this way then a logical next step is to adjust/use appropriate risk management strategies as the portfolio risk transitions from one regime to next regime. 

For example, for smallest market fluctuations (i.e., 0% to -5%), one approach to manage portfolio risk is by doing dynamic balancing like volatility targeting i.e., regularly balance the asset (stocks, bonds, cash...) allocation in portfolio such that total portfolio volatility is within a predetermined target.

Four modes of risk management

Now to protect portfolio against losses in 2nd regime (say -5% to -15% losses), one effective approach seems to be finding alternative assets to re-align portfolio exposure. What I understood is basically expand the portfolio to have more diversification in terms of strategies and alternative exposures. Feel free to correct if I got it wrong.

To protect portfolio against even deeper losses i.e., 3rd regime (say -15% to -35%), an effective approach is to explicitly hedge the tail risk via option-like strategies. That makes sense as basically by using option like strategies, one is out-sourcing the risk. 

Also it makes sense to use in this regime and not in earlier regimes as portfolio incurs cost in implementing this approach.  Note: The risk management strategies to handle earlier regimes also has costs like whipsaws/giving up on gains for additional diversification in regime-2 or jump costs when doing re-balancing in regime-1.

One challenge is losses don't really care about our regimes definition and can seamless move from one regime to another making all above approaches to manage risk useless. Another big challenge is future is unknown. The reason I found this article interesting is it provides a framework to consider portfolio risk and to craft ahead a plan on how one can go about managing the portfolio risk and surprises. Your thoughts?

Link: Four Modes of Practical Risk Management

Markets are always changing. Some key questions every trader has to answer irrespective of their approach is - how is my trading strategy doing currently? Is the market currently conducive to the strategy for pressing the edge or to reduce the exposure? Under what environments  my method will shine and in which environments will it run into rough seas? Is the edge gone permanently or is it just a temporary draw down?

Interestingly we don't hear much in trading literature/blogs about this aspect and techniques one can use. If you have heard, then please let me know. I like to read. 

Anyway, in my opinion, a strategy performance is dependent broadly on three factors:

1. Suitability of current market environment to the method.
2. Suitability of the money management algorithm being used to the method.
3. Suitability of the trader (nature/personality) to the method.

This post focus is on first item i.e., getting an idea on whether the current market environment is conducive to the strategy and in what environments will it do well. Similarly when to pull the plug on the system and shelve it or revisit its core logic. As an example, I am using the strategy we covered in prior post. If interested, you can find more details of the system here and here.

In my opinion, having a simple and objective process/rules for doing this analysis makes a big difference both to the account and to trader's health. Often the aspects of trading that causes stress are those areas that are not simple and crisply defined to follow repeatedly.

I am sure there are an alphabetical soup of quant/statistical tests (i.e., A - Z tests) one can perform to determine how the system is doing currently. Similarly another alphabetical soup of adaptive approaches to side step this problem. I could be wrong but IMO the problem with adaptive approaches is they will satisfy intellect more than the account, adds lot of complexity to strategy and then somehow magically hit the one case we forgot to consider. Do you know of any adaptive strategies that withstood last decade and performed well?

The approach I use is fairly simple, effective and objective. I don't know of any quant tests that can do better than the approach I use currently. Doesn't mean there are no better approaches out there nor this approach is the best. I am sure there are and look forward to investigate. I welcome readers to share their thoughts and techniques.

I think there is wealth of information one can gain from a simple performance summary chart. To get better mileage for you and for me, I recommend readers to take a deep look at the 1st image for couple minutes and note down what comes to their mind about the system character. Then look at the 2nd image which is heavily annotated with my observations.  Then please let me know where our observations differ or things I overlooked/mistaken. If enough readers do, it will be beneficial to all.

Profiting from emotions price action strategy performance

Note: Read the annotations in the order they are numbered. These will set the path to the final question i.e., how is the trading strategy doing currently? is the edge still there? when can one press the edge for this system? and finally what are the the red flags to watch for that will let me know the system edge is in danger. 

Profiting from emotions price action strategy performance with annotations

Please feel free to share your thoughts, any techniques you found useful and also any inconsistencies in my analysis. We learn most when our views and ideas differ. I hope readers got some useful takeaway from the post.

Wish you all good health and good trading!

If your platform is not Ninjatrader then probably you can skip this post. All the quant stuff on this blog are done using Ninjatrader & R. For last few days I was trying to get Ninjatrader software handle following scenario. I feel the scenario is fairly common and the solution will be useful others. So posting it on the blog.

Scenario:
I have couple strategies that works off daily and weekly charts. Now I would like to use the strategies for live trading. But I am NOT comfortable with software placing orders automatically. Also I would like to use some level of discretion. So my desired workflow is:

• Each day wait for US markets to close for the day. Then start Ninjatrader and connect to an EOD data feed (Example: Yahoo/Kinetic etc).
• Select my strategy and run on a pre-defined watch list in Strategy Analyzer.
• The strategy executes on the watchlist and generates list of orders (buy/sell/short/cover) with details. I either take a printout (or save in spreadsheet) to trade manually the following day/week.
• Repeat the above step for other production strategies & watch lists.

Last post covers a simple price action strategy to profit from crowd emotions and some stats on it by market regimes. You can find the post here. It is easy to either skip that or move on to something else after quick read because it is too simple.

This simple strategy had beaten buy-n-hold by a wide margin overall in last 17 years. CAGR of 10% is good especially given the short time the strategy spends in the market. IMO time is one of the safest risk control a strategy can have. You can see for yourself the results, consistency and other ratios etc in the following two images along with my annotations.

Strategy - Equity Curve, Weekly Returns, Drawdowns

Strategy - Performance Analytics

Currently the max drawdown of this method is 20%.  That is too high for my comfort. One of my reasons for sharing this method on the blog is to hear your thoughts and suggestions on ways one can reduce the draw down of this strategy. Any suggestions?

Wish you all good health and good trading!

Many people approach the market with assumption that only ideas that are complex or arcane can provide an edge in the markets. Unfortunately most trading books and vendors promote this assumption to sell their own services. Would you buy/subscribe otherwise? Another culprit is the notion that the more complex and intellectual a method or concept is, the better it is. Typical justification, otherwise everyone would have figured it. Another reason - intellectual addiction.
 
Sometimes simple things can provide an edge. Partly because they tap into simple basic emotions we all have as humans. For example, remember the last time market (SP500) had a strong sell off and closed near its lows. If you are visual, think of a large big red bar. What was your thought pattern and reaction? What sense did you get from financial media and popular blogs?

My guess is in both cases, it is a negative emotion followed by intellectual/logical reasons to justify market sell off and continuation of it. We know majority people are bad at timing. So how about taking other side of the crowd? This is where the price action research of this post comes in.The purpose of the study is to see what are the market returns when we follow the crowd vs when we takes opposite side to the crowd. 

Using SP500 (SPY) market as an example, one simple way to define crowd behavior for the day is based on the market close with respect to its high, low and close of the day. There are multiple ways to identify but for this test how about we define crowd behavior as follows -

• The location of the market close with respect to its high of the day provides the emotional temperature of the crowd for the day. The farther the close from high of the day, the stronger the emotional temperature of the crowd.
• The location of the market close with respect to its open provides the emotional sign of the the crowd for the day i.e., down day = negative emotion, up day = positive emotion.
• To quantify above, let's divide the day's range (i.e., high - low) into 4 quartiles to identify in which quartile the market closed for the day. Following image provides a pictorial description of what I mean.

Definitions:

• Bull Market - Market is above 200 day moving average. 
• Bear Market - Market is below 200 day moving average.  
• Graphs Annotation - In below graph, on X-axis, label 25 means 0%-25%. Similarly label 50 (25%-50%) , label 75 (50%-75%) and label 100 (75%-100%).

Close_Test:

• Compute the quartile for today's SPY market close i.e., (0-25)  or (25-50) or (50-75) or (75-100)
• Buy market @ open the next day.
• Sell market @ open the following day.
• Calculate the performance stats.

Note: Another interesting test would be to buy @ close today near end of the day (instead of buying @ open the following day). This can help one to get in on overnight action. I assume most of this blog readers are likely end of day traders. So for the test, I am using buying @ open the next day.

Bull_Market_Test:

• SPY market close is above its 200 day moving average.
• Compute the quartile for today's SPY market close i.e., (0-25)  or (25-50) or (50-75) or (75-100)
• Buy market @ open the next day.
• Sell market @ open the following day
• Calculate the performance stats. 

Bear_Market_Test:

• SPY market close is below its 200 day moving average.
• Compute the quartile for today's SPY market close i.e., (0-25)  or (25-50) or (50-75) or (75-100)
• Buy market @ open the next day.
• Sell market @ open the following day
• Calculate the performance stats. 

Misc:

• Test Duration: 1996 t0 Current
• Friction less results i.e., no commissions, no slippage.
• Long only trades.

Fading vs Following Crowd Emotions

Results Analysis:
My observation is in general it takes lot more time to write blog and generate plots compared to time it takes to create a test or analyze market. I already spent lot of time on this post to provide something worthwhile and results are fairly clear cut. Look at the annotations on the graph - 

• Specifically the results of 75%-100% quartile for all three tests. 
• Results of this top quartile under bull and bear market regimes.

Note:
The above is not a system nor it is a recommendation. Just a research into of one of the market characteristics. Feel free to let me know your conclusions from results. Also feel free to agree/disagree. 

Wish you all good health and good trading!

"There is just one life for each of us........ Be yourself!"

This study is about bull and bear markets (S&P 500) performance by VIX regime. For the test, I used SPY etf as the proxy for S&P 500. 

Definitions:

• Bull Market Phase - Market is above 200 day simple moving average. 
• Bear Market  Phase - Market is below 200 day simple moving average. 
• VIX Regimes:  0-15 (low volatility), 15-30, 30-45, 45-60 (high volatility)

Bull Market Test

• Market is in Bull Market phase.
• Go long when market transitions from previous VIX regime to new regime.
• Exit long when market transitions from current VIX regime to next regime.

Bear Market Test

• Market is in Bear Market phase.
• Go short when market transitions from previous VIX regime to new regime.
• Exit short when market transitions from current VIX regime to next regime.

Misc

• Test Duration: 1996 t0 Current
•  Friction less results i.e., no commissions, no slippage.
• Long only trades.

Results Analysis - Bull Markets: 

1. Profitable in all 3 VIX regimes. The Trades category provides an idea of how many times the market entered into a particular volatility range.    
2. The Win% is highest in VIX range 30-45. But the number of time market entered into that VIX range is relatively less.  
3. Average Win/Loss Ratio and Average trade returns are highest in VIX range 0-15. I wonder if it is because of low volatility anomaly in markets?

Results Analysis - Bear Markets:  

1. As expected, results show bear markets are more volatile than bull markets. Unlike bull markets, bear markets entered VIX 45-60 range multiple times. 
2. Long trades are profitable in the volatility regime 30-45. Not sure Why?  
3. Another is high average win/loss ratio in volatility regime 0-15. Why? 

 Feel free to let me know if your conclusions from results is different from above. Also I am curious to hear your thoughts on above 3 questions.
 
Note: The above is not a system nor it is a recommendation. Just a study of one of the market characteristics.

Today while scanning through WallStreetCurrents site, I came across a post on new Volatility ETF (VIXH). What caught my attention in that ETF prospectus was its rules based on VIX levels for buying VIX options. Thought will check out how those VIX rules would fare if I apply it on SPY.

Now rather than blindly buying SPY at each VIX level, thought I will combine with another study I am checking currently i.e., week of the day effect on SPY. (Note: If any readers are interested in pure VIX level based entries test then please let me know. I will do in one of the future posts).

Test:

• Divide VIX range into 4 levels : 0-15, 15-30, 30-45, 45-60. (Note: My levels are slightly different from ETF but that shouldn't make much difference).
• Buy @ market next day open and sell after 2 days. Note: Only one position at a time. Next position is opened after the current position is closed. I think this condition is more realistic.
• Finally tabulate the performance metrics categorized by VIX level and Week of the day.
• Test Duration - 1995 to 2012 Current. Caveats - Results are frictionless i.e., no slippage & no commission.

SPY ETF - Week of the day profile by VIX regime

Results:
Some takeaways

• Poor performance of longs when VIX level is above 45.
• Low performance of longs when VIX level is below 15. But draw downs are also low. So may be risk parity approach to increase the returns.
• The sweet spot seems to be to go long on SPY only when VIX level is between 15-45. 

In the ETF prospectus, rules related to VIX level are as follows:

• VIX futures less than or equal to 15, no VIX calls are purchased
• VIX futures above 15 and less than or equal to 30, 1% of portfolio in VIX calls
• VIX futures above 30 and less than or equal to 50, 0.50% of portfolio in VIX calls
• VIX futures above 50, no VIX calls are purchased

Your thoughts?