Put/Call Ratio as Bullish/Bearish Indicator

Research
1

I have been using a signal system in StockCharts that tracks the CBOE Put/Call Ratio ($CPCE). This data measures purchases of options on the S&P 500 Index ($SPX). Traders buy Puts (Bearish) or Calls (Bullish), and this ratio offers a good sentiment indicator for when investors are more Bearish than Bullish and vice-versa.

A smoothed version of the series provides accurate Bullish and Bearish signals for the S&P 500. Signals occur when the CBOE indicator begins rising (BEARISH) and when it is generally falling (BULLISH). A moving average can help smooth the weekly data, and trend identification can be achieved with a Rate-of-Change indicator or a Moving Average Crossover (MAC).

However, I never use single indicators for my ETF-based models, and instead, use composites of multiple uncorrelated indicators to generate weight-of-the-evidence signals that are far more accurate. As you can see from the 16-year chart below, the CBOE indicator offers accurate signals when it trends either higher or lower, which enhances signal composites. I would like to include it (if possible) as part of a signal composite in a new Smart-Money Sentiment Strategy I’m building.

My question: Is there any way to track this series on P123? It’s widely available on many other sites, and I know that the values can be imported each weekend from another site before a strategy’s rebalance/reconstitution, but I find that importing is both a hassle and introduces another point of potential human error (an employee runs our weekend updates). Therefore, if at all possible, I would prefer not to need to do that downloading and uploading each weekend.

If there is no way to get this info currently, is it a series that P123 can add as a technical/sentiment indicator?

Here is a 16-Year weekly chart of the CBOE Put-Call Signal System:

…And here is a 3-Year Chart showing that this indicator would have signaled to reduce exposure BEFORE the COVID CRASH in 2020, then back in for the meat of the post-crash surge. The indicator would have also signaled to exit the market at the turn of the year some seven weeks ago, missing the losses from the downturn that started then:

2

I am not an expert on any one indicator. Not an expert on any indications at all now that I think about it.

BUT I do think that ETFs may have great potential for P123 as far as machine learning and attracting new members. Maybe as much as stock factors in the long run.

FED data, pricing data and data like what ETFOptimize proposes in this thread can be entered into a Random Forest, Boosting program or neural-net VERY EASILY. It is MUCH LESS DATA INTENSIVE for a member and for P123 too, I would think.

To my way of thinking, it would be no different and actually an extension of what people already do: like moving average cross-overs discussed by ETFOptimize above or relative strength as another example.

It is easy to show that a random forest can successfully classify ETFs that are likely to do well over the next month on a laptop. This can be done with pricing data alone and can be shown to be statistically significant. But it is more difficult to backtest this to see what the returns of such a strategy might be. I think P123 could attract a lot of people who may already be doing this (like me) and would like an easier way to backtest the results.

Anyway, if P123 intends to move into machine learning this is something that would not be very computationally intensive (especially relative to stock factors) and would work for the members, I believe. Especially when combined with some mean-variance optimization or other methods of portfolio optimization.

[color=firebrick]If anyone wants to do that now they would probably want to do it at Portfolio Visualizer or somewhere other than P123.[/color]

I think this could generate some quick and easy success for new members. Don’t forget that sims/ports can be entered into this to create a portfolio.

Again one of the main advantages to this is it is easy. It is low-hanging fruit. People could very easily select the ETFs using JASP now. Backtesting the results is the hard part. Something that P123 excels at already.

I see this as a supplement or addition to what people like ETFOptimize are already doing. I will let Chris and others make their case for individual factors. Although is seem like this is a good one if the data can be obtain with reasonable cost and effort.

[color=firebrick]Or just consider it a supplement to pattern recognition that P123 is already trying to implement. And an easier one to code at that.[/color]

Jim

3

Is the historical put/call ratios available for download anywhere?

4

Chaim,

The historical data for $CPCE is available here.

https://www.barchart.com/stocks/quotes/$CPCE/price-history/historical

Regards
James

5

Hi James,

I visited the link you provided for the Put/Call data on Barchart. However, the page says (and dates confirm) it is only three months of Put-Call info offered for free. So fine, I will need to pay to get more data… I checked the Barchart subscription options, and it’s about $200 per year, so not too bad.

However, the deal-killer for me is that after you join, you can only get a maximum of two years of data for Put/Call or any other series. That’s not nearly enough data for running statistically significant backtests.

I’ll check on another site I maintain for analysis - TradeStation. TradeStation has data back to 1900 for most indicators, but the CBOE (the exchange where these Puts/Calls are traded) has only been alive since 1973. Moreover, I don’t know when this ratio of Puts/Calls started being tracked. I can’t find that specific info anywhere.

YUVAL: As the P123 PRODUCT MANAGER, can you address the question of whether this data can be added? I may be willing to pay more if the data is clean and accurate.

6

Chris,

You can also check out this site.

https://ycharts.com/indicators/cboe_equity_put_call_ratio

If you want more data, you can paid and download the historical data directly from CBOE

https://www.cboe.com/us/options/market_statistics/daily/

Regards
James

7

Chris - To answer your question, we don’t have any current plans to offer any options-related data. It’s something we can look into if there’s enough interest.

8

Thanks, James. I will check out both of those sources!

YUVAL - I hadn’t considered the Put-Call Ratio to be “options-related data.” I am no more interested in options or options-related data than you are. However, perhaps I can explain why I have this opinion that probably seems illogical.

The CBOE Put-Call Ratio is considered across the industry to be a high-value, market-based Sentiment Indicator. It may be the best Sentiment Indicator that investors have available to them. Most sentiment data based on self-selected surveys (such as the AAII Individual Investor Sentiment) or one person’s assessment of another person’s sentiments (such as the Investor’s Intelligence Advisor’s Sentiment Report) is questionable at best.

I have been fortunate to have been quoted many times over the last 20-30 years in the Investor’s Intelligence Advisor’s Sentiment report. That well-known report has been published every Wednesday morning since the mid-1960s on the InvestorsIntelligence.com website. I greatly appreciate every time this high-quality publication posts an excerpt from one of my articles.

I consider Investor’s Intelligence current CEO, John Gray, to be a long-time friend, and I certainly don’t want what I’m about to say to be perceived as criticism. However, whoever is assigned to do the Advisor’s Sentiment job is required to categorize and label the content from dozens of advisors reviewed each week as having either a “Bullish,” “Bearish,” or “Correction” opinion about the market. That must be an incredibly challenging job!

However, I’ve found that my missives often get miscategorized, and I’m placed into the wrong group of advisors about half the time. However, the Advisor’s Sentiment product is wholly dependent on one person’s subjective interpretation of another’s writing about a highly complex subject. Everyone’s in a hurry, and I imagine that there’s a lot of skimming of those newsletters and websites.

I like to think that my market-assessment articles are subtle and nuanced and are probably difficult to categorize into a bluntly defined sentiment group. After all, the market itself is infinitely subtle and nuanced, with millions of daily participants having different objectives and time frames, wielding billions of dollars driving changing data points. It’s a fantastic undertaking that is both complex and complicated!

Yet, in the past (I haven’t tuned in to it in years), I regularly saw famous “market gurus” such as CNBC Jim Cramer referring to these subjective Sentiment Reports as if they were gospel handed down from on high.

On the other hand, the CBOE Put-Call Ratio is a mathematically derived numeral determined daily from the records of actual money-based transactions. Investors and traders are voting with billions of hard-won dollars to hedge their long-term investment positions from the overall market turning against them. I believe that this type of definitive, market-based data is far superior to subjective opinions (mine or anyone else’s) and a (possibly incorrect) count of interpretations of those opinions.

For this reason, almost all the content I produce these days involves identifying and explaining changes in critical indicators that are driving changes in my client’s portfolios. I believe that the CBOE Options Equity Put-Call Index could be one of the best indicators that investors have available to determine investor sentiment and periods of increasing market risk. It’s a shame it can’t be available to Portfolio123 users.

Yuval, just to clarify, Portfolio123 has long provided users with “options-based data.” The $VIX Index is another valuable sentiment indicator derived from option-based data from the CBOE. I can’t remember what year Marco added it, but I seem to recall using the VIX indicator in the site’s early years, around 2004-2006.

I’m sure such a highly regarded indicator as the Put-Call Index is available from FactSet or another of the site’s data suppliers, and perhaps at no additional expense. Can I convince you or another P123 staffer to at least investigate it, please?

If it’s added to the data, I’ll even promise to write a tutorial for those interested in the P123 community on using it and provide two or more ways to turn the data into accurate Risk-On/Risk-Off signals.

9

I agree with Chris that the Put-Call Index would be a valuable addition to the P123 data for the reasons advocated by Chris.

I use consumer sentiment as one of my market timers:
Consumer Staples/Discretionary Spending As A Reliable And Profitable Stock Market Timer
https://seekingalpha.com/article/4459097-consumer-staplesdiscretionary-spending-as-a-reliable-and-profitable-stock-market-timer
This timer changed from risk-on to risk-off on Jan-18-2022, which looks like a good call.

Market timing models are no different to stock ranking systems in my opinion. P123 should give us the PIT tools to develop them. One timer is not enough, you need many unrelated timer models for a market timing ranking system.

10

Georg, Chris and All,

This is Georg’s quote above. I agree with this 100% and I think it is VERY EASY TO PROVE THAT WHAT GEORG SUGGESTS WORKS.

In my case, I get many unrelated and uncorrelated models using a Random Forest Classifier: that is what Random Forest Classifiers do and why people use them. Random Forest Classifiers accomplish that, in part, by using bootstrap aggregating (bagging)–something that P123 members have tried to emulate in spreadsheets. Reportedly with some success at P123.

I am not claiming my way of obtaining unrelated and uncorrelated models is better than Georg’s, Chris’ or anyone else’s method. But you can do an awful lot of what Georg suggests above in a short amount of time using a computer to verify that it works.

You can generate 10,000 uncorrelated and unrelated models in 12.57 seconds +/- a few microseconds. I just timed it on an old MacBook Pro. And it is easy to test the statistical significance of a holdout test sample in Python, with Excel or searching for an online statistical calculator with Google (e.g., online Chi Squared calculator for classification).

I can say what Georg said above is 100% correct (specific methods aside).

Four broad points:

  1. P123 has already committed to AI and machine learning.

  2. P123 has already committed to pattern recognition that has VERY LOW POSITIVE PREDICTIVE VALUE: I am all for that btw but there are other things with more bang for the programmer and server buck.

  3. Not everyone wants to mortgage a home and leverage that money on a port that could have a 70% drawdown.

  4. I intent to use ports mixed with ETFs. It is not a competition between the two methods. But everyone has their own risk tolerance and adding ETF models is one way to reduce both market and individual stock risks.

That having been said, it is not about any single indicator. I defer to Chris and Georg on specific indicators and on P123 to determine the feasibility of each indicator.

But if P123 still wants to do AI they should do some easy things that can be show to work as well as continuing to try to do the hard things, like pattern recognition.

Jim