Saturday, April 23, 2016

A Statistical Look at the Indicators

Alt title - All you never wanted to know about put/call ratios and more.

Weekly outlook gain/loss. Last week I was expecting weakness and recommended shorting at SPX 2090.  This was hit Mon AM and I recommended a tight stop at 2095 here due higher than expected VXX $ volume and previous indications of a more bullish outcome with the SPX over 2085.  The stop was hit Tues at 2096 on the open, so a weekly loss of 6 pts for a net of +27 the last two weeks. This week may be weak but I expect higher levels before or after the FOMC meeting. No new positions.

Indicator roundup.  I forgot this last week which was a decrease to -9 with 1 up, 10 down, and 5 neutral from the prior weeks -5.  This week there was very little change with a decrease to -10 with Skew moving to neutral for 0 up, 10 down and 6 neutral.  Of note on Tue, there was a very low VIX P/C with over 1.1 M calls - previous similar call volume was seen on Dec 11, two weeks before Dec-Jan slide and three times during the Aug crash.

This week rather than show an indicator update, I wanted to look more closely at the use of put/call ratios and apply the use of a statistical package I found for the testing of the effectiveness of indicators in general.

The put/call data I use comes from the CBOE.  Their major categories are CPC (total), CPCI (index), Equity (CPCE), ETF, SPX, and VIX.  CPC is essentially the sum of Equity, ETF, SPX, and VIX, while CPCI is the sum of SPX and VIX.  You can see this in the table below showing put/call average daily data from two recent months this year using Jan as a down month and Mar as an up month.  The sum of the last four categories is about 98% of the CPC calls and 96% of the CPC puts.  In both cases the VIX calls represent over 20% of the total while the VIX puts are less than 10% of total.  Results for the CPCI with VIX and SPX calls and puts are similar.  For my purpose I use a CPC Rev(ised) leaving out the VIX puts and calls, and the SPX for the CPCI.

To test for the significance of the differences in put/call categories with extension to other indicators, I applied a least squares (regression) approach to measure the correlation between different categories and expected returns as shown in the following table.  Using data from the last two years that the SPX has been trading between roughly 1800 and 2100, April 1 2014 to April 15, 2016, I applied the regression as Returns = b x Indicator + constant. The measure of fit is the correlation coefficient and is shown in the table below as "coef / b". The correlation varies between 0 and 1 with 1 as a perfect fit.  For contrary indicators the "b" should be negative for time 0 (high CPC means low prices), but positive for the future (high CPC means positive future returns).  The returns were constructed as the change in SPX between time 0 and 5,10, 20, 40, and 60 days in the future with 20 trading days per month. To concentrate more on the highs and lows I used a 50 day SMA envelope with a +/- 1.5% spread, omitting points inside the envelop.  This reduced my data set from 520 days to 320 days.  The table below is for the 10 day EMA only.

The results for the CPC show a reasonable correlation between .30 and .40, but the CPCI has a very low correlation with a negative b.  Looking down, however, you can see that the CPC Rev is much higher rising to .50 and .60 correlation in the 10 to 40 day period and the SPX is also much higher than the CPCI, rising almost to .50 by day 40.  The negative effect of including the VIX puts and calls is obvious.  I also included the Smart Beta P/C discussed several weeks ago as ETF puts/Equity calls, and it turns out to be even more effective than the CPC Rev or the combined Equity and ETF P/C.  The other hybrid the BPSPX/CPCE is more effective than the CPC or CPCE shorter term, but less so for 40 and 60 days.

Of the four volatility measures, the VIX term structure for the SPX (VXV/VIX) and VXX $ volume are the most effective with the VXV/VIX less so for longer terms (60 days).  The NDX VIX term structure (VXN/VIX) appears useless and even has a negative b, while the SKEW appears to be somewhat effective, but only up to 20 days.  The two additional components used in the Students Trifecta with the VIX term structure, the overbought/ oversold indicator (SPXA50R/SPXA150R) and the TRIN/BPSPX are both slightly weaker than the VIX term structure short term, but stronger long term.

The various composites, including two new ones, show very similar results with correlations in the .50s at 5 days, .60s at 10 days, .70s for 20 days, then declining to the .50s by 60 days. To summarize with CompositeINT (intermediate term) and CompositeLT (long term):

  • Composite#1, SPX P/C + VIX $ Vol + SKEW
  • Composite#2, CPC Rev + VIX $ Vol + VIX term structure
  • Students Trifecta, VIX term structure + SPXA50R/SPXA150R + TRIN/BPSPX
  • CompositeINT, Smart Beta P/C + VIX $ Vol + VIX term structure
  • CompositeLT, VIX $ Vol + VIX term structure + SPXA50R/SPXA150R
As of next week I will be dropping the VXN/VIX indicator and replacing it with the Smart Beta P/C.

Saturday, April 16, 2016

Sentiment Review - Composites and ETFs

Weekly outlook gain/loss. Last week I started posting a weekly position strategy by projecting a rally through option expiration.  Gain from last weeks close to this weeks close SPX 2048 to 2081 is +33 pts.  Over the next two weeks I am expecting a 2 to 4% pullback from a possible Mon gap up then down into the next FOMC Apr 27/8.  Short from SPX 2090 if hit.

Over the last two weeks I showed how one of the composite indicators, the Students Trifects (VIX term structure, overbought/oversold, and TRIN indicator) gave a sell signal, but I still had reservations based on the neutral position of the volatility measures.  This week I will discuss two of the other composites as well as review the cash flow measures derived from the short/long ETF pairs for the SPX, NDX, RUT, HUI, NBI and TNX.  As a whole these indicators are slightly bearish, indicating a pullback is likely but a final top has not been reached.  As a final note, the VXX $ Vol sentiment did reach a low enough level for a "weak" sell.

The first composite (#1) is the SPX P/C, VXX $ Vol (adj) and SKEW.  I had been expecting something like the sharp decline similar to the Oct 2015 rally, but the current position is more like Feb 2015, supporting more choppy market behavior.  A low SKEW keeps this indicator from being less bearish.

The next composite (#2) is CPC Rev, VXX $ Vol, and VIX term structure.  This composite is more bearish and is at similar levels to both the Feb and Nov 2015 tops.

The next two indicators are for the SPX ETFs for the 2x SDS/SSO and 3x SPXU/UPRO.  Both measures are at positions comparable to the Nov 2015 top, but sentiment has not reached a level extreme enough to expect more than a short term pullback.

The large cap tech sector, NDXrepresented by the 2x QID/QLD and the 3x SQQQ/TQQQ are also at positions comparable to the Nov 2015 top, but sentiment has not reached a level extreme enough to expect more than a short term pullback.

For the small cap RUT index represented by the 3X TZA/TNA ETFs, sentiment is slightly more bearish than at the Nov top.

For the gold miners, represented by the HUI and the 3x DUST/NUGT ETFs, sentiment has reached an extreme normally warranting a sell, but as I pointed out a couple of months ago, the extreme sell level is likely to remain for several months to balance out the six months at the extreme buy level the last half of 2015.  The result is likely to be a trading range.

The biotech sentiment indicator represented by the NBI index and the BIS/BIB ETFs is probable the most surprising result since a 10% rise in price has caused sentiment to drop to neutral.  However, comparing this to the sentiment for HUI, the largest part of the rally for the HUI in early 2015 and 2016 occurred after sentiment dropped below neutral, perhaps that is the "recognition" phase.

Finally, looking at the sentiment for bonds using the TNX as the index with the TBT/TLT ETFs, sentiment remains in the extreme low bearish level.  Since bonds and gold are seen as safe havens, perhaps both will fall when an increase in confidence in the economy is seen. Note that since interest rates and bonds move inversely, low rates equal high bond prices.

Saturday, April 9, 2016

Why I am a Cautious Bear

Alt title - Bears markets only happen when bears are in hibernation.

Last week I made an official sell signal while pointing out that the top may not be in yet. Basically I am expecting sideways chop (up or down) into at least mid-May.  Today I want to show why I am cautious on the short side. Given the sharp declines of last Aug and this Jan, the bears seemed to have focused on the volatility products as their hedging weapons of choice.

So to illustrate this I will show four volatility measures that I track: VXX $ volume (adj) as an intermediate term indicator, VXX/XIV short/long ETF measure as long term, VIX put/call as short term, and the SKEW as long term.  Also, I have finished adding options for longer period EMAs as multiples of the original 5, 10 and 20 day periods, so will be trying those for illustration.

First the VXX $ volume (adjusted for S&P volume).  This is probably the most reliable single indicator I use and currently the pattern more closely resembles the Feb 2015 period than the November 2015 period leading me to slightly favor an upward chop with a target of SPX 2100.  The VXX buyers/bears need to get whacked a few times before giving up.

The following is a two year chart using 12, 25 and 50 day EMAs that shows no significant declines when readings are as neutral as they are today.

As a long term sentiment measure, the VXX/XIV ratio and typically shows extremes at the end of long term moves.

This is seen more clearly in the chart going back to 2013, the earliest reliable data.  Here several cycle are evident.  The pattern from mid-2013 almost looks like an inverted head and shoulder pattern with double left shoulders.  It reminds me of what I would expect if Avi Gilbert's scenario came true where a correction in the SPX to the 1700's was followed by a several year rally to 2500.

The VIX P/C is mainly useful in identifying short term tops as low put/call ratios determined by large number of calls is often indicative of "smart money" hedging positions through buying of VIX calls.  This measure is slightly negative and longer period in the same range would be similar to Nov-Dec of 2105.

The last volatility measure is the SKEW as measured by out of the money option premiums (mostly puts).  I prefer to think of this as a "smart money" indicator where option premiums are set by the writers, but overall this does not show much short term timing effectiveness. Here the axes are reversed since high values are bearish.  The high values from Sept to Dec 2015 did correctly forecast the Jan selloff.

Longer term shows the effectiveness of the SKEW in prior periods which makes the recent low readings somewhat worrisome from the bearish perspective.  Current readings are similar to early 2015.

As a final chart, borrowed from another blogger, is an analog to 1957 which may be an alternative if the SPX retests its 2015 highs.  Here we had a year long trading range followed by a very weak Jan, a rally to mid-year, then a 19% decline.  As a final alternative, I give this one a probability of 20%, but with a breakout over SPX 2085, move to 40%.

Conclusion.  For the indicator roundup this week the readings improved to -5, up from -8 last week, with one positive, 6 negative and nine neutral.  Notably ETF P/C, NYMO, and SSO/SDS moved to neutral with SKEW moving to positive.  Composite#1 moved to neutral as it includes SKEW.  A rally for options expiration seems likely and oil may be strong with another OPEC production meeting next weekend.  A move to SPX 2070-80 will likely provide a good shorting opportunity as OPEC news is likely to be sold.  Probably the best opportunity for a 3-5% correction since late Feb, potential target roughly SPX 2000.

Sunday, April 3, 2016

Maybe This Time is Different

It has been about six months since I completed my sentiment timing model with additions along the way, so I want to start by reviewing the intermediate term signals that have been generated.

My first signal, posted November 3, was following the strong October rally after the late August crash in 2015.  At that the EWers had declared the start of a primary 5 wave that would take the SPX to 2300 to 2400 the first half of 2016.  For technical support they pointed to the Zweig Breadth thrust as confirmation.  However, using the SPX put/call ratio, VXX $ volume and Skew as Composite #1, the sentiment model generated a sell signal here (link to post).  At the time, I noted that momentum would probably carry the market for a few weeks, but that I expected a 10% plus decline in Dec-Jan.  Everybody thought I was crazy.

My second signal, starting Jan 23 when I introduced the Students Tricecta (VXV/VIX), SPXA50R / SPXA150R, and TRIN), I showed that comparison to the 2008-09 bear market indicated a bottom was near, here.  Later, on Feb 6, I noted that the Low VIX Put/Call Ratio indicated that a washout (retest of lows) was likely and that would be followed by a sizable rally, but unlikely to make new highs.  At the time, the EWers had declared a new bear market and were calling for SPX 1600 or lower by March with a longer term target of SPX 1100. Everybody thought I was crazy.

My third signal, posted today is a sell based on various indicators of which I am showing the Students Tricecta today since I was able to back test it for the 2008-09 bear market and I think It may show some insight into the future.  Note that this is an intermediate term signal, and I expect the market to show strength into the middle of May and may possibly retest the highs of SPX 2135.  However, over the past few weeks I have given a target of SPX 2080 and doubt if is exceed by more than 1-2%.  It is interesting that many EWers have now declared a new bull market this time based on the overwhelming strength of the NYMO and NYSI.  I am sure that everyone thinks I am crazy again. Maybe this time is different.

Re-posting the Jan 23 chart for comparison, I want to point out what happened in May of 2008.  After a sharp Dec-Jan decline that took thw DJIA down from 14,000 to 11,500, the DJIA rally back to 13,000 in May and sentiment had fallen to an even lower bearish extreme than it was at the Oct highs.  Given the fact the SPX has been has been in a range of about 1800 to 2100 for two years, I expect to see even greater extreme lows in bearish sentiment before a breakdown of that range.

The rest of this post discusses minor updates to my sentiment model.  I had some difficulty in interpreting the VXX daily $ volume the last week of Dec that kept me from being as bearish as I should have been.  Namely, even though to $ volume met my criteria of 50% of average daily volume, the overall market volume was also about 50% of normal, so the reading was unclear.  Therefore, I have added an option to calculate volume relative to the daily S&P 500 volume.  The results are shown below.


The differences are fairly minor other than the noticeable drop for the end of 2015, but still maintaining the sell level.

Next, I want to introduce a new sentiment measure on a trial basis that I came up after reading a couple articles on smart beta portfolios where portfolio managers may take sizable positions in certain stocks (equity calls)  to raise their alpha but offset the risk by hedging the risk by shorting sectors (ETF puts).  So to duplicate the result I am using the ETF puts/equity calls obtained from the CBOE to create a smart beta put/call ratio.  I am not sure if this adds any new information since I already track the ETF p/c and equity p/c, but it sounded like an interesting idea.  I have also noticed that the number of calls in the equity p/c are about 50% greater than puts, while for ETF and SPX puts are about 50% greater than calls.

This ratio is faster moving than normal put/call ratios so may be better for shorter time periods.  The sharp spike mid-week did warn of an end of week rally.

Looking at the overall indicator scoreboard, the score dropped to a -8 from -6 last week, with 1 up, 9 down, and 6 neutral.  Equity p/c dropped from positive to neutral, while ETF p/c went from negative to positive.  The cash flow measures SDS/SOS and SPXU/UPRO went from neutral to negative.  The volatility measures (VXX $ vol, VIX p/c and Skew) are all neutral, so only moderate volatility is expected short term.