The StockRanks - five years of market beating returns and ten essential insights

My feelings:

Quality can't be arbitraged away by the market. If a stock is intrinsically growing its NAV and earnings this must eventually be reflected in the price in proportion to the underlying growth.

Momentum is obviously partly simply the price reflecting the underlying quality but also might contain herd behaviour that will fade away again. In the real world, Quality without Momentum might mean that there is something that the fundamentals are missing but the market has noticed, and Momentum without Quality should give you pause to wonder whether it's just a bubble. But Quality plus Momentum is a very decent combination.

Value is the metric which is most likely to be arbitraged away by an efficient market. Although a high V can protect in a downturn, it can also be a sign that a stock is going to fall even further. In fact V>85 is often a warning sign. High-V stocks have a bad habit of getting worse. On the other hand, when 25<V<75 (or since good stocks tend to be skewed towards expensive, 10<V<60 can work better) the stock in question is fairly priced and well away from the extremes where strange things can happen. As a bonus, middling V usually also means lower volatility. I think of it as being like trying to go straight down the middle of chute rather than bouncing wildly from side to side.

Great stuff which I will have to take some time to properly read through. But to provide an anecdote. I recommended Stockopedia to my uncle as the platform is very easy to use. The second reason is that I think the ranking system provides a good wake up call on questionable stocks.

His feedback was something along the lines of "I see these tips and recommendations in the press and then go to Stockopedia and see they are rated as a load of (*£&$£ [insert expletive]." So I think it has been very helpful for him to not get waylaid by media tips and story stocks (although a few story stocks a good).

The only thing I would say is to improve the ranking system by focusing on risk. Happy to explain how if someone contacts me. For a number of reasons I think volatility is an inadequate measure of risk (backward looking). You could easily have a fourth ranking score with risk. It would include things like the operating margin, capital intensity, financial gearing, long-term track record and you could also put volatility in there perhaps.

This metric WOULD have picked up Conviviality for example which was a low margin business with significant financial gearing. However, the Stockopedia Quality metric had it rated over 90 at one point (not sure how). I would look at how Conviviality could have been as high on the quality metric. But more importantly I would introduce a separate risk ranking using mainly fundamental data.

To be clear, this is not a criticism it is just an idea for one area that things might be improved. Things will go wrong in most systems with stocks not working out. But if there were things that could be improved so that bad stocks don't slip through the net I think they are worth looking at.

Lastly, and perhaps this has been mentioned but if not here goes. If you then have four factors - quality, momentum, risk and valuation - you get the overall Stockrank. However, different investors may feel differently about each factor. So maybe they should be able to weight it themselves to get the Stockrank they would prefer. Somewhat risky for people to do themselves but a possible idea. I.e. so if I am a low valuation and quality investor these are more highly weighted in my overall Stockrank that appears on every page for a stock. Could still keep the default Stockopedia Stockrank on the page too perhaps.

Very interesting but its not always obvious which market environment we’re in!
Perhaps slowdown now as QM High flyers seem to be doing well but is value about to make a comeback?

Another issue is of course JPM are referring the economic cycle. This can both lead or lag movements in the stock markets (although as of late downturns seem to have been caused by stock market crashes)
Gromley you beat me to it and your amendment to the schematic captures the point very well. JPM are not saying that QVM investing doesn't work at certain stages of the cycle but rather that the emphasis on the Q, V, M components should change over time. I had understood that Eds team were doing work on some sort of cyclical adjuster to take this into account.
Also I would contend that the issue of whether or not SRs 'work' in the long run depends on how you use them. I view them as a sophisticated screen rather than a system to be blindly followed.

What I find interesting is that, over the past 5 years, the Momentum Rank has comfortably beaten both the Quality and Value Ranks, but the Momentum Rank has only narrowly beaten the QVM StockRank, which is a blend of the individual Quality, Value and Momentum Ranks with equal weighting to each of those three. So including the relatively poorer performing Quality and Value Ranks in the QVM StockRank's blend has not significantly harmed performance of the latter over a prolonged bull market period which has favoured momentum. This makes me hopeful that if market conditions should change in the future, for example if there is a recession and/or a bear market, so that the Momentum Rank performs relatively poorly compared to the Quality and/or Value Ranks, then the QVM StockRanks blend should still perform reasonably well. That's relative to the market as a whole, of course, which might not be much consolation if everything is falling!

Rob Arnott, who's a factor based manager, has done some research on the long-term trends across growth and value stocks. He did this by going back 50 years to look at the stock valuations then and then tracked forward to see how those valuations panned out in terms of returns to shareholders. He called this Clairvoyant Value.

What the analysis showed, over 50 years of the US markets, was that value stocks and small cap stocks outperformed growth and large cap. But the research threw up two really interesting findings:

• Firstly, that the market had generally done a good job of identifying those stocks which were going to grow quicker - roughly what we would regard as high Q stocks - as opposed to lower growth V stocks.
• Secondly, the market overvalued growth stocks compared to value stocks - that is, over time, although growth stocks grew much faster than value stocks it was value stocks that delivered the best returns, because the market was discounting them too deeply and valuing growth stocks too highly.  Investors roughly paid 50% too much for growth.

Now there are a lot of assumptions in this research and I certainly wouldn't want to hang my hat on it, but in an environment where we have to pay quite heavily for quality and value is being largely discounted I think it would be a very brave assumption to entirely discount it. Paying large multiples of earnings on the basis of future growth is relatively risky, and always has been.

Sadly we'll only know for sure in 50 years time. I won't be around to see that, but maybe some of you will ...

timarr

Re: JPM Quant Investing diagram with StockRank styles overlayed.

Firstly I think there's a slight typo in the StockRank overlays on the diagram.
In the expansion phase, it's low Q, high VM which is classed as a Turnaround in the Taxonomy of StockRanks

Secondly I just don't see the evidence for the JPM argument .
Here's a chart for the S&P 500 since 2004 ...

I'm not sure what you call the uptrend from Spring 2009, is it all Recovery or 50/50 Recovery/Expansion? Maybe the Recovery phase is 2009 up to early 2013 and then the Expansion is 2013 to date (which is the period that the Stockopedia data covers)

JPM would suggest during this bullish trend that ValueTraps or Turnarounds would be outperforming? But that hasn't been the case over that period, well certainly not from 2013 according to the Stockopedia data.

Using the Stockranks performance tool with parameters of > £50m mkt cap and quarterly rebalancing on the UK market on the data set from 2013 to date we see ...

QVM (SuperStock)     +141%
VM (Turnaround)      +138%
QM (High Flyer)         +130%
Value (Value Trap)      +94%

Maybe this isn't an accurate picture as selecting Value as a parameter in the Stockranks performance tool doesn't mean that all the stocks selected would meet the criteria for the ValueTrap label (High V, low QM). So it'd be nice to have a Classification performance tool too.

I wonder if there's any bandwidth in the Stockopedia team to crunch the numbers for us to see if there's anything in the JPM analysis?

Phil

I have some doubt about the labels in that JPM diagram too. They look like the kind of classification you can only clearly delineate after the event when you look back. Then you say "Ah-ha! There is where I should have switched from high momentum to low momentum."

There's an old joke they used to tell when teaching about stochastic processes (essentially what we are dealing with here):

"How do you make perfect toast?"

"Grill it until it burns. Then five seconds less."

The point being that you only know the key bit of info you require too late to be able to act on it.

I can't figure out where we are supposed to be on the JPM graphic. But I want to stay in tune with the market to make maximum returns. To do this I regularly check how the different stock ranks have performed over different recent time periods.
You can see my recent conclusions in a comment that I have just added to my original post entitled "TOP PERFORMING STOCK RANKS"

What concerns me about the Stock Ranks measure is what appears to be their day to day volatility.

Are the graphs above based on an annual average stock rank for any stock or a stock rank measure taken at a particular point in time?

As an individual investor, I don't have access to the former but I do have limited access to the latter i.e those of todays and any previous days I recorded

Hi Edward
Great presentation and I love the site, stockranks and the overall approach. I am from a database/programming and project management background and whilst I know numbers are not everything, they can provide invaluble insights. I do have a question for you, which is around the inflounce of spreads on validity of the returns illustrated, which I will explain in more depth further down this post.

I have also been spending a lot of time reading investment books, focusing mainly on factor investing and behavioral finance, with some popular stock picking books too. What I have found fascinating is that even after really researching investing, and understanding the behavoural traps, I still fall right into them. Turns out I have the same monkey brain as everyone else - who would have thought it!

Hence I am considering a pure factor based approach, using the full range of value/momentum/quality factors in the Stockopedia stockranks and only picking those with an SR of 90-100. I plan on ordering all of these by size (smallest first) as the clear evidence seems to be that small companies outperform, and picking the first thirty, as this means I can invest enough equally in them, without the fixed trading costs becoming too great. (about £3k in each). I plan on rebalancing annually, to reduce trading costs, and expect the latter to roughly equal the dividend yield.

The one extra filter I am planning on using is only choosing shares where the spread is less than 200 basis points, to minimise trading costs, and make sure I am not trading in illiquid stocks. This also seems to rule out most companies under £50m in size, but that is a prettty small size.

The question I have for Ed is if we pick a max level of spread (say 200 bp), and call those within this spread 'tradable stocks' and only including these stocks in the performance analysis, does this affect the historic returns that have been shown.  I suspect it may affect the overall return, as it will eliminate many of the 'micro caps' where there is outsize performance.  I completely understand that any factor based analysis excludes trading costs (and dividends) as they are hard to quantify because they depend on trade size, broker,  taxes etc.  I am more curious about what happens if you only include what I have defined as tradable stocks.

Any comments on what I have proposed or my question for Edward are very welcome, I have only been looking at investment for a couple of years and like learning from others.

I take the JPM chart literally i.e.

• SLOWDOWN = positive GDP growth, but rising at a slower rate
• CONTRACTION = negative GDP growth, falling
• RECOVERY = negative GDP growth, but rising (ie less bad)
• EXPANSION = positive GDP growth, and rising

So looking at the UK ONS data for the last UK recessionary period 2007-2009:

3Q/4Q 2007 : EXPANSION

1Q 2008: SLOWDOWN

2Q,3Q,4Q 2008: CONTRACTION

1Q, 2Q 2009: RECOVERY

3Q 2009 - 2Q 2010: EXPANSION

(NB to answer Nick Ray's valid point re the difficulty of knowing where you are in real time, it is true that there are long lags in the publication of GDP data. Instead, the  monthly PMI data  (manufacturing & services) can be used to reveal acceleration/deceleration in growth rates.

The JP note stretches to 124 pages i.e. it is thorough! It is based on a back-test of 20 years of European Equity data 1997-2016. 

The other thing to note is that they favour looking at factor risk-adjusted returns, not absolute returns. So even if we had Stockopedia data from 1997 for European Equities, we'd still need to calulate the Sharpe ratios assess the quality of JPM's conclusions.

I'm happy to email you the note if you're keen on reading it.

Hi David.

A quick and dirty screen filter suggests that by cutting out stocks with a spread above 200bps you will severely limit the available universe of shares. For example, if you take the AIM all share index, there are 803 companies of which only 123 have a spread of 200bps or less (according to Stocko) , i.e. roughly 1 in 7.

Gus.

I suppose Ed's Nap's approach by having a limited pool of 20 twenty stocks in different sectors presumably in QVM classes may be a way of smoothing out volatility .

Individual stocks I have selected have jumped about in their individual classification. So it a bit like being in a post office queue, you pick the one that will move quickest only to find, that the one next to you is outperforming. However, jumping about only adds cost and market timing issues as a friction.

I note that Ed will be reviewing his own NAPS performance in the not too distant future.

However, surely, a more scientific approach, assuming there are enough QVM qualifications stock available is to run one or more control NAPS selections in parallel to the one he has selected for his own, to baseline to some degree what the results are. This would help screen out some of the noise in his own selections. I appreciate he does this in a broader graphical analysis of QVM performance, but that is very hard to judge in detail and against real results.

If that trends in a similar way, it creeps towards an underlying position that QVM has a higher mean that other categories of selection, and therefore in periods when out performance reverts to mean, you are still optimising overall performance.

As you will appreciate from the above I am not a statistician!

size effect related to liquidity premium (spread)
therefore by choosing small cap with low spread is contradictory in my opinion

Rick: a very good, and important, question.

Can Stockopedia please provide an answer?

Hi Ed,
When will one be able to invest in a StockRanks ETF?
Would save me a lot of time!
Mark

Are the graphs above based on an annual average stock rank for any stock or a stock rank measure taken at a particular point in time?

The graphs are based on an equal weighted portfolio of stocks at a particular point in time that have that particular rank (e.g. QualityRank) between the lower and upper bound (e.g. 90-100).  The performance is tracked in a portfolio for 3 months, and then the portfolio is rebalanced and the same process starts again... equal weighted, held for a quarter, rebalanced... etc etc.

We also track semi-annual and annual rebalancing... which all have the same general trend, though slightly lower performance records.  The benefits of annual rebalancing though are lower transaction costs. 

What concerns me about the Stock Ranks measure is what appears to be their day to day volatility.

There is of course some day to day volatility in the StockRank as it's influence by the stock's price, which changes daily.  But there's far less volatility in the StockRank than in any specific valuation or momentum ratio that is price influenced - e.g. the P/E ratio or Dividend Yield.  The least volatile rank is of course the Quality Rank as it's more stable between reporting periods.

Great analysis thank you.

Some of us try to reduce / balance market risk by having a short portfolio to complement the long.

Understandably there is a large bias on the site to long analysis but i think some punters here would appreciate a little more on the short side. For example does the 5 year record show that the most efficacious long attributes are equally powerful on the short side. Just looking at the bottom quintile chart appears to show this is not the case.

Thanks for a great site.

On the subject of using a long/short system it is quite amusing that the second best performing guru screen is a short screen. The "James Montier Trinity of Risk Screen" which has outperformed nearly every other system, including the stockranks. Probably not the way to go if you are looking for candidates to short.