Trying to create a "resilient" medians-based QM screen

8th Apr '19 by Nick Ray

17 comments 2288 reads

An examination of H2 2018

Anyone who was invested during 2018 cannot have failed to notice that everything was going really very nicely until about September - and then everything fell off a cliff.

Interestingly, NAPS2018 seemed to start to show weakness from about June 2018, but the rest of us caught up later and eventually the stocks which seemed to have the highest momentum seemed to suffer the most.

I commented previously that what seemed to happen was that the factors we usually rely on seemed to lose their ability to discriminate. So with about six months of distance from September 2018 now, I ran a version of my Bayesian analysis to see whether any of the metrics I track were "resilient" in the sense that they did not lose their discriminatory powers in the second half of last year.

[In the following plots the QVMS ranks at the start of the period are compared with the performance at the end of the period.]

First, here is what the QVMS ranks look like when they are working well:

These plots are showing "likelihood" that a particular value of QVMS will result in above average returns. You can think of them as ~~decimal~~* odds, so a value of 1 is the equivalent of a 50% chance of outperforming the index. [*: thanks to PJ0077 for pointing out that "decimal odds" are defined differently. Here the intended meaning is p/(1-p) where p is the probability.]

Q and M are both quite strong discriminators. Perversely perhaps lower V tends to out-perform higher V because low V tends to be an indication that the market expects better performance in the future. But if you look at the scale on the y-axis you can see that V is a weaker discriminator than QM.

The Stock Rank has a slight droop at higher values. Typically all stocks with a SR >= 80 are roughly as good as each other.

However, look at what happened over the last six months of 2018.

Everything is topsy-turvy. Lower Q and lower M out-performed and high V finally paid off. Even more interesting, V is now a very strong discriminator. On the other hand, SR is really not making much effect. Although it is swinging up and down, the y scale shows that the…

Unlock with Email

Unlock with Google

Already have an account?
Login here

PJ0077

8th Apr '19

These plots are showing "likelihood" that a particular value of QVMS will result in above average returns. You can think of them as decimal odds, so a value of 1 is the equivalent of a 50% chance of outperforming the index

Your graphs & wording are confusing, Nick

Decimal odds are fractional PLUS 1

e.g odds of 4/1 = 5

e.g. odds of 9/2 = 5.5

The attraction to using decimal odds (for me) is the inversion gives the implied probability:

e.g odds of 4/1 = 5 (1/5 = 20% implied probability)

e.g. odds of 9/2 = 5.5 (1/5.5 = 18% implied probability)

Nick Ray

9th Apr '19

Thanks for the correction. I meant odds in the sense of p/(1-p).

I hadn't realised that "decimal odds" was a modified version of what I think of as odds.

iwright7

Nick,

An excellent piece of work that plays to my bias for a high QM investing style.

There is one aspect that is often forgotten in the search for the reason that Momentum generally works? Greed dictates that that there must be considerable insider trading of as yet undisclosed improving company results, which will cause an upward price trend. By default buying into this upward Momentum trend follows the buying activity of these insiders. Combine upward Momentum with Quality companies (that have a high probability of improving earnings) and Hey Presto we have a market edge to.

Cynical I know, but for various reasons your method does tend to select quality companies, with less downside risk in the short and medium term. Ian

rpannell

A very interesting piece of work Nick. Thanks for taking the time to share it. You say:

Over what period have you taken this data to produce these graphs? Excluding the end of 2018, my understanding is that the value rank has been uncorrelated to performance for the past 5-10 years but had correlation prior to that time. Perhaps we are entering a time again when investing on the basis of value could have some traction.

HumourMe

Very interesting work. Thanks for sharing.

You say

I only track a relatively small number of metrics but I did find a handful which seemed to remain reliable predictors even in H2 2018

Which metrics that you track didn't make the screen? Presumably of the un-tracked, those that are similar to these could also be rejected, which might leave a subset of possibles that may or may not be useful.

By the way if you run for Europe, excluding the UK, there are 87 selected. The QM relationship is broadly maintained:

I only subscribe to UK + Europe so can't run anything else.

Over what period have you taken this data to produce these graphs? Excluding the end of 2018, my understanding is that the value rank has been uncorrelated to performance for the past 5-10 years but had correlation prior to that time.

My data only goes back to January 2016. I think there is a similar shape if you scrape the performance data for Value Rank from Stocko's performance charts. (It is hard to see with the cumulative linear plots on the performance page.)

Perhaps we are entering a time again when investing on the basis of value could have some traction.

To be fair the idea here is not to decide when to switch between QM and Value. It is to find metrics which do not "go wrong" in a mild bear market. Most of those metrics turned out to be QM metrics in fact.

I don't really want to get into that level of detail. What I will say is that nearly all the momentum metrics apart from the 52w High one were rejected. And I fully expected that to mean that the screen would fail to pick up on good momentum prospects. But to my surprise that does not seem to be happening. I think there are hints that metrics based on Sales growth are also quite resilient.

But one of the interesting things about screening, and especially screens based on ranking rather than fixed values, is that you should not add rules which are quite weak discriminators. They damage the overall result. You need the smallest number of rules with the best discriminatory powers.

andrewdb

I do not think you can really get any "resilient rules" (many may disagree)

The thing I have been trying to work out is how to detect when the "rules" do change.

mmarkkj777

Great article Nick.

I'm going to take a closer look at this to see if it can add value to my selection process for long term buy-and-holds.

Rusty2

Nick, Looks a good screen, my only query is about growth, Focusrite (LON:TUNE) and RM (LON:RM.) have low EPS growth forecasts for the next 2 years. I would want to see more growth to sustain the share price. I may have a play around with this, thanks for posting it.

Nick, I have taken a closer look at your selection criteria especially in light of your comment about not particularly selecting for Momentum, but ending up with High QM stocks. It would seem that for the period selected (6 months) selecting for your range of >median Quality factors and >medium 52 Wk High Momentum has produced a clutch of High QM stocks, whose share price has been and is still climbing.

When I look at the Relative Strength (RS) numbers for the 23 stocks selected historically:

1M – RS >0 = 78% Beat
3M – RS >0 = 86% Beat
6M – RS >0 = 82% Beat

So around 80% of these stocks have beaten the market over the last 1-6 months with just a >median 52Wk High Momentum filter. Some of these beats are big ones too, far outweighing the market losers. That is super impressive! Ian

10th Apr '19

I've been wondering why these particular metrics might provide some resiliency. What follows is just me speculating wildly, but what are forums for if not to share unjustified opinions with the rest of the world eh?!

% volatility - low volatility is a well-known "anomaly" when discussing efficient markets. It is a sign that there is less price uncertainty which may be an indicator of general happiness with the way a company is performing an being managed. Not a surprise to see it here.

NAV growth % - is the growth of the book value of a company. It acts a bit like a fundamental version of momentum but uses NAV instead of Market Cap. I think this is not a new idea either. There are fundamental indices which try to avoid the herd aspects of Market Cap based indices for example. The link does a decent job of talking about the rationale and also criticisms of the approach.

% vs 52w High - this is a measure of price momentum but it is also sensitive to volatility too. So it might be a poor man's version of the Sharpe Ratio. I might have to check this out further to see if the two ideas are correlated. (Note that the % vs 52w Low metric does not work as well because it does not have the sensitivity to the volatility component.)

Op Mgn % - this one is giving me the hardest time to fully understand. ROE and ROCE are not "resilient" by the test I described in the article - they flip along with everything else. But Op Mgn is. Wildly speculating, maybe ROE is skewed by leverage and ROCE can be flattered by understating book value. Clearly Op Mgn can be manipulated too though. It is possible that CROIC% would also be a good metric but I do not have the data to be able to find out.

Final note: I tried swapping out Piotrovski F score with CROIC% and EPS gwth % instead in the screen and get a similar number of hits. As I mentioned, I am not totally happy with using Piotrowski because it seemed a bit arbitrary and usually I find that F scores >= 4 are usually fine.

Nick,
I see that most of the shares passing this screen have very high Q scores, but that neither high Q score nor M score alone proved resilient over the last 6 months. Could it be that just combining high Q and high M (high QM) would be resilient?

I note that the companies the screen has shortlisted all have low Beta’s as well as low Vol and are highly cash generating (FCF/Sales average of 11). Could it also be that the particular mind set of the of the type of investors buying these type of shares who are not so prone to selling during a correction, so the price holds up better? I am reminded of Buffetts style where he purchases of a slice of a business for its future cash flow, rather than trading shares for short term gain.

Nicowilson

A really nice piece of work. Thoughtful and well constructed. Thank you.

It's high quality writing like this that makes Stockopedia stand head and shoulders above many other discussion boards.

Speculation ahead.

To try to eliminate random fluctuations with some % vs 52 week high based screens I've begun changing %52wh < x to %52wh > - 1yr % volatility. It makes little difference in this case, However, if close to the 52w high then the buyers are ascendant and due to anchoring, potential buyers will still be reacting slowly.

NAV growth represents historic retained earnings. In combination with growing op margin this implies continuation. Also if it is growing factor based academics should be supporting any undervaluation depending on which theory they are following.

Op Mgn is a margin of safety and represents relative efficiency and pricing power. The higher the value, the greater the stability and presumably, the less the volatility. A positive trend can only be good and points towards future NAV/Dividend growth.

Piotroski is good for a quick 'improving FA' approach. Although it is part of the Quality rank here, it was developed I believe for value stocks https://www.investopedia.com/terms/p/piotroski-score.asp To me it indicates that things are generally getting better YonY. However, if things are already good, above a threshold, then YonY loses relevance. A lowish Piotroski with CROIC ranking would seem more relevant in this case and would point towards future NAV improvement.

Personally, I'd have liked to see some form of relative valuation criteria or something that limits paying too much. However, we live in interesting times and applying PEGs etc.. excluded most current candidates.

WarrantStar

13th Apr '19

Hi Nick
Thanks for an interesting piece of work. I tend to invest in High Flyers and would like to make my portfolio more resilient, preferably without sacrificing any performance in the good times. Would you consider the use of Stockopedia Risk Ratings to achieve this? Stockopedia say that "We have designed the RiskRatings to be both a useful predictive measure of future volatility, but also an easy to use measure for accessing the “low volatility anomaly” - the unusual fact in equities, that lower volatility securities tend to outperform high volatility securities over the long term."

Yes indeed. The rule:

1y Volatility % < median

is almost the same as requiring a Risk Rating of Conservative, Balanced, or Adventurous (because according to the definition this adds up to the lower 45% of the volatility range).

However according to the glossary, the Risk Rating is based on 3y volatility and the "1y Volatility %" only looks at the last year. I also use "rank in sector" rather than "rank in market" in the screen itself which tends to be slightly more tolerant (includes more matches). So there are small differences.

The current set of matching stocks only includes one stock which does not have a Risk Rating in the range "Conservative, Balanced or Adventurous" which shows how similar the selection criteria are.

Thanks Nick,
That's a relief! The way that I normally deal with this question is to exclude all shares with a Highly Speculative risk rating from my screens.

Trying to create a "resilient" medians-based QM screen

An examination of H2 2018

Unlock the rest of this article with a 14 day trial

About the Author

Nick Ray

17 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...