Random Portfolio benchmarks

The growth of passive investing has a lot to do with investors migrating to lower-cost investment vehicles, but implicit in the migration to passive frameworks is a presumably lower risk metric. Investors want to be exposed to equities but don’t want to take risks beyond those inherent in the benchmark indexes they are investing in. Additionally, index investors are opting out of the risks of active management because the returns generated have recently not kept pace with the index benchmarks. But indexes are not what investors should be using as benchmarks. They don’t represent a truly passive approach to the market. The only truly passive approach to the markets is one that employs random portfolio construction.

Why random portfolios? Because all indexes are systematically biased by their factor premises. The prevalent index factor is size (market capitalization) but other factor frameworks include sector, industry, value, growth, momentum, and smart beta. Each of these index frameworks employs a systematic weighting of components based on a predetermined valuation that aims to minimize variability of returns based on defined factors.

Take a look at the S&P 500 index. It’s the most traded index in the world - through ETFs like SPY - but more importantly is the primary benchmark for U.S. equities. The performance S&P 500 index guides investors in terms of relative performance of actively managed funds and ultimately is the most broadly used compensation metric of the asset management industry. The fees investors pay these managers and, ultimately, the employment, compensation and rewards these fees fuel depend on the structure of the index.

The S&P 500 index is first and foremost a members club. Stocks are included in the index by decision of a committee. Yes, constituent stocks must meet certain primary criteria - market capitalization (float-adjusted weightings), liquidity, domicile, public float, sector classification, financial viability, and length of time publicly traded and stock exchange - but the criteria is set by a member board. And it’s set with a certain purpose in mind: the index is a gauge of large cap U.S. equities.

So we know that the SPX is a factor-based measure of U.S. equities - it tells us about the price movement of large-cap stocks. That should surprise few people. Implicit in this size factor is the fact that the SPX accounts for about 80% of the entire capitalization of the U.S. stock market. That’s a good chunk of the assets invested in this market.

However, the size factor weighting of the SPX is problematic when it comes to fulfilling the index’s role as a benchmark. The size bias distorts the benchmark performance and concentrates on the largest companies, often adding additional risk in those constituents because many are in the same sector or industry. In fact, the current weighting of the SPX shows that about 28% of the index weighting is in technology stocks. Further, the top 4 stocks by weight in the index are Apple, Microsoft, Amazon, and Facebook. The S&P 500 index has a decidedly tech bias at this time, a considerable sector risk for a benchmark index that is ostensibly supposed to track the overall performance of U.S. equities.

That may be a reasonably correct weighting of big cap U.S. equities, however, it’s not a correct benchmark measure of alpha. Alpha is the intelligence that extracts investment returns above the market performance. A portfolio of stocks constructed with a size bias only tells us about the performance of the bias. It does not represent the performance of a naive portfolio - a portfolio of stocks that has no factor bias. A naive portfolio would not have a selection criteria that restricts to a subset of a given universe of stocks. It would be a portfolio derived from a set of randomly selected stocks.

Random portfolio returns give us an estimate of the returns that are built into the broad market. Irrespective of factors that deliver alpha, a measure of random portfolio returns tells us the returns a given market generates without having any specific intelligence about how to generate those returns. They are the returns that would be generated by an untrained monkey.

What have been the random portfolio returns of the U.S. equity market? How do they compare to the returns of the S&P 500 index? Below is an annual comparison of 52-week returns (%) of the market cap index and the mean (average) return (%) of randomly selected portfolios. The last column shows the differences in returns (%) between the market cap index (SPX) and the random portfolios.

The random portfolio returns are the mean return (%) of 1,000 randomly selected portfolios of 100 common stocks selected at the beginning of the return period. That would be equivalent to 1,000 different monkeys picking a 100 stock portfolio, then taking the average return of those 1,000 portfolios after the 52-week period. The universe of stocks from which these random portfolios are selected includes all NYSE/Nasdaq listed common stocks that have traded for at least 40-weeks trading at a price above $2 and weekly trading volume above 100,000 shares. That would be a universe of about 3,900 stocks.

Annual returns (%) - SPX and random portfolios

This summary tells us that the S&P 500 index has only out-performed the random portfolios in 7 of the last 15 years (2003-2017) and that the sum of the differences in return is -26.3%. The monkey would have outperformed the S&P 500 index over the period by a significant amount. However, we can also see that much of this outperformance comes in 2003 and 2009, both years where the overall market enjoyed exceptional returns after a bear market. Clearly, those were periods where the size bias of the S&P 500 index excluded it from returns that were delivered elsewhere in the market (small-cap stocks, growth stocks, momentum stocks, etc.).

Presently, the S&P 500 index is providing excess returns above the benchmark represented by random portfolios. On the surface that tells us that portfolios weighted toward large-cap - specifically large-cap technology stocks - are outperforming the market. They will until they don’t. Investors should be aware of where their returns are coming from and where their risk is situated. The S&P 500 index is not a passive index. It is not market agnostic.

Stock Trends Slots game!

There is a new Stock Trends learning application installed on the Stock Trends website - a slots game! The Stock Trends Slots game engages users by generating random sets of stocks and matching combinations of their Stock Trends indicators. These matching combinations score points based on the probabilities of the match. Every week a new set of Stock Trends indicators reflect the changing stock market, so the game’s probabilities and rewards change with the distribution of the indicators.

Investors might wonder how a random outcome game applies to the markets. But there are aspects of skill in the game that come with the premium functions subscribers to Stock Trends Weekly Reporter may access. The ability to lock game rows allows users to improve the probabilities of making a match without a corresponding reduction in reward. That means a skillful player who understands the distribution of trend and price momentum indicators in a current market can achieve higher scores.

Each spin of the game creates a random portfolio of five (5) stocks and/or ETFs. This is a useful application of random portfolio generation that also makes for an interesting stock market game. These random portfolios - generated by the Stock Trends Slots game play - can be entered into the Stock Trends Investor Challenge, a weekly stock market competition to see which portfolio has the best return after 4-weeks of actual market activity.

This stock market game helps illustrate how random portfolios perform relative to the market benchmarks. Users have some degree of skills-based intervention in the Stock Trends Investor Challenge as they can either choose to enter their slots game portfolio, or not. The Stock Trends indicators help give guidance on that decision, just as they do for actual stock market analysis.

Give it a try!

Stock Trends Slots

Trend profile - SPDR S&P Metals & Mining ETF $XME

Stock prices tend to move in trending patterns. This is a simple idea that may, or may not be supported by evidence. It really depends on how you frame the question and what time frame is presented. That’s because prices also tend to revert to a mean or average price. This interplay of price trend and price reversion is a fundamental dynamic of the market. It’s also the window dressing for evidence of randomness that underpins market price patterns.

It’s easy to not recognize randomness when we focus so much on trend and reversion. A core mechanism for human understanding is our ability to identify patterns and formulate responses to them. Indeed, much of our learning is dependent upon pattern recognition. Why shouldn’t our understanding of the markets be based on the same formulations?

Many successful trading strategies are based on pattern recognition. Whether fundamental or technical in nature, these systems win when the precepts of their approach match what the market is delivering at any particular span of time. Market is trending: systems based on price trend patterns win. Market is reverting: systems based on price reversion patterns win. A truly intelligent trading system would know how to recognize the difference between the two and when to apply either a price trend system or a price reversion system. (Orpheus Risk Management Indices (RMI) is one such intelligent system http://www.orpheusindices.com/)

However, attempts to predict outcomes in a world of true randomness cannot be absolutely defended, by definition, no matter how intelligent. Looking for patterns in randomness brings us to Chaos theory and fractal mathematics, which explores the transitions between order and disorder in deterministic systems dependent of initial conditions.

This is heady and fascinating stuff that has a growing influence on financial markets analysis, but how does a stockpicker fit in all this? It’s no wonder that the era of the stockpicker is quickly transforming into algorithmic and machine learning systems increasingly favoured by capital markets money managers. That’s all well and good for highly capitalized institutional shops, but what about the little guy?

The Stock Trends Inference Model (STIM) is an attempt to reconcile randomness in the market with evidence of price patterns. It is a simple application of statistics to Stock Trends categorical indicators that answers some basic questions about certain trend characteristics. Examples of these questions include: If the price momentum of a stock is relatively high and it has been in a bullish trend for a relatively long period, will the price momentum continue, and for how long? If a price trend has changed from bearish to bullish, what are expectations for price momentum going forward? If a stock breaks out of a bearish trend, what are the probabilities it will it retreat?

All of these questions are asked with the assumption that the answers provided are independent of the present broad market condition. That is, we want to know return expectations regardless of whether the market is in a bull or bear trend. Why? Because we cannot know whether the present trend of the market will persist. If we make an assumption that it will, then our measurement of the expected returns of an individual market (stock, ETF) will be imprecise.

This is important. When we take a measurement of a particular market condition - as represented in the combination of Stock Trends indicators in each weekly Stock Trends Report for individual stocks and ETFs - the observations of similar market conditions will take place across time periods that span the entire population of observations. Each observation occurs in varying broad market trend phases. In this respect, the Stock Trends Inference Model is broad market agnostic.

For that reason, the base measurement of returns is relative to the historical random returns of stocks, which is about 8% annually, and specifically equate to the following expected returns for each of the relevant periods Stock Trends measures: 0% 4-week return, 2.19% 13-week return, and a 6.45% 40-week return. If a trend condition for a particular stock/ETF does not provide statistical evidence that it can beat these base return expectations, then we cannot say anything definitive about its return expectations. However, if there is a deviation from the base return expectations we can say that the current trend characteristics indicate either over-performance or under-performance projections. This is the objective of the Stock Trends Inference Model.

A good starting point for Stock Trends Weekly Reporter subscribers is a weekly review of the STIM Select stocks report. It shows the stocks and ETFs that have the best statistical trend characteristics. The report is ranked by the 13-week return expectations.

The current NYSE STIM Select report, as an example, includes the SPDR S&P Metals & Mining ETF (XME-N).  The Stock Trends Report for XME shows that the ETF is 7-weeks into a Weak Bullish trend; that it is underperforming the S&P 500 index by 12% over the past 13-weeks and underperformed the broad market index last week  (RSI 88 - ). It’s been in a Bullish category for 54-weeks but has been retreating since February.

The statistical model shows that there have been about 277 observations of stocks and ETFs that have shared these characteristics or have had similar Stock Trends indicator combinations. From this sample we can make inferences about the expected returns of XME over the next 4-week, 13-week, and 40-week periods.

The green sample density plots show the distribution of returns for the three separate periods following the observation. Most generally, these distributions will be centered around the mean random return expected for each period ( 0% 4-week return, 2.19% 13-week return, 6.45% 40-week return). However, certain Stock Trends indicator combinations yield sample distributions that deviate from the expected mean random returns. The sample distribution of returns generated in the XME sample deviate in a positive way.

For the 4-week period 53.8% of returns in the sample are greater than 0%, the expected 4-week return. By employing statistical inference methods to estimate the population mean, we can estimate that the expected (or mean) 4-week return for XME is 1.8%. More importantly, with our assumption of a normal distribution of returns - a defining attribute of randomness - we also can estimate that XME has a 56.5% probability of having a return greater than the expected 4-week return of a randomly selected stock. This in comparison to the 50% probability we would expect from a random stock.

Similarly, the 13-week expected return for XME is 7.2%, with a 60.8% probability of besting the base period expected return of 2.19%, and the 40-week expected return for XME is 21%, with a 64% probability of beating the base period expected return of 6.45%. All better probabilities for beating the returns of a randomly selected stock.

While even a 64% probability is better than a 50% probability implicit in a random selection, it’s still only a 64% probability. There is a 36% probability that it will underperform the expected return of a randomly selected stock. If you know anything about chance, you must know that a 36% chance of being wrong is more than enough to lose your shirt.

However, the Stock Trends Inference Model does tell us that XME is currently in a trend and momentum position that historically has exhibited tendency toward positive returns in the subsequent period. This gives us some confidence in making a directional trade, and can be used as the foundation of a derivatives trade (options) that further improves a trader’s probability of making a profitable trade.   

Stock Trends Slots Game

Stock Trends Slots

The stock market is a game of chance. Try your luck!

Stock Trends reports on weekly price and volume changes for thousands of individual North American stocks. Every eligible stock is assigned a series of Stock Trends indicators which interpret those changes. Match these indicators in the Stock Trends Slots game and learn how the changing market affects random outcomes presented by the game.

This game of chance is a great way to learn about the Stock Trends indicators and how they provide guidance toward probable outcomes.

Every week a new data set is created with the Stock Trends updated indicators. The Stock Trends Slots game data for the current week are the records that have Stock Trends indicator values for listed common stock, exchange traded funds, or income trust units on the following four exchanges - New York, Nasdaq, Amex, and Toronto.

How to Play

Player receives a set of 5 random draws to start the game.

On each draw or play, a random generator fills 5 rows with listings for 5 different stocks, with each record filling the columns in the order of stock symbol, Stock Trends trend indicator, Stock Trends volume indicator, Stock Trends RSI value, and Stock Trends RSI up/down indicator. Learn more about what these indicators mean in theLearn section of the Stock Trends website.

Each draw the player accumulates points and/or free plays with matching columns and rows, as described in the section below. Each week, with the changing market and the new set of Stock Trends indicators published, the probabilities and associated payouts change. See how the market trends affect the game play and how your luck ranks against other players!

At the end of each slots play players may choose to enter a Bonus feature play - the Stock Trends Investor Challenge! Enter your final portfolio of 5 stocks in the rolling 4-week Stock Trends Investor Challenge for a chance to win BONUS awards if your portfolio beats all challengers! Stock Market Investor Challenge winners are awarded every week. (Note: presently we are beta testing the game - no prizes are applied to winners)

Subscribers to Stock Trends Weekly Reporter have the option to lock rows before a spin and improve the odds of making another point scoring match. For instance, if a player has matched two rows, he can select those two rows to be locked in position such that in the following spin only the other three rows are randomly selected. The points payout for matches of additional rows remains the same, but the probability of making those matches improves.

The following rules apply to the locking of rows:

1) once a row match is locked, no points for the locked match are scored on the subsequent spins. That is, there is no double counting of matches when the matches are locked.

2) if the player locks more than two rows, the free spin combinations do not receive free spins. There is a cost to locking more than two rows - the player cannot accumulate additional spins with the free spin matches.

Play to improve your odds and enter your portfolio of stocks 
in the Stock Trends Investor Challenge!

As a subscriber to Stock Trends Weekly Reporter you will have access to additional game features:

• lock game rows to reduce random outcomes and improve your odds of making a match!
• enter your own custom 5-stock portfolio in the Stock Trends Investor Challenge.

GO TO STOCKTRENDS.COM TO LEARN MORE AND PLAY!!

Technical, value-centric filters

Usually when analysts look at the stock market they use a framework that is either market-centric or value-centric. Sometimes these are classified as top-down or bottom-up approaches, and they apply to both fundamental and technical branches of analysis. A market-centric framework looks at the market as a whole, assesses sector strengths and drills down to find specific trading opportunities that fit the best-of-breed theme. A value-centric framework looks at individual securities, finds value (fundamental or technical) opportunities and indentifies sector themes that either support a trade or expose it as a special situation.

Stock Trends is aligned with the bottom-up approach. The screening process employed here is an attempt to isolate trend and momentum trading opportunities. Thematic discussions (which orthodox technicians sometimes cynically refer to as “bedtime stories”) may lend credence to a trade, and certainly they are great for editorial purposes, but a quantitative methodology is more concerned with the mathematical foundation of a trading model. The reasons for a trade are defined by the logic of the model, not its output.

For instance, the current market conditions seem to suggest a rotation toward precious metals stocks. Our filter reports have highlighted some prominent gold plays. Other themes are evident as well, including a rotation to the real estate sector. But the trading models propagated here don’t take methodology beyond the value metric defined by the filter reports or the statistical inference model being introduced here. Patterns evident outside the models are, for lack of a better word, extraneous.

As an investor or an investment professional you may be concerned with context. Perhaps you even need to find your own rationale for a trade that colours beyond the algorithmic pale. This is a human need. It’s why the financial sector is so good at writing bedtime stories, why talking heads on business television sound so convincing when they present such refined tableaus of the economy, of market dynamics, of monetary and fiscal policy. Whether peddling feel good market positivism or the death stare of Dr. Gloom, analysts and commentators are gifted narrators and oracles. They give investors a sense of order in the midst of chaos.

However, as an algorithmic trader it is not necessary to strive to be all-knowing of market order – or even assume that any order exists. It is only necessary to understand the assumptions and methods of your model. In the case of the Stock Trends reports, as well as the inference model elaborated on again below, the output is based on a defined methodology and its implicit assumptions.

The task of a trader to translate the model’s output into a trade setup that executes a trading plan. The Stock Trends model trading strategies are examples, but there are other trade setups that can be employed using alternate plans. At some time we’ll start looking at trade setups that involve stock options, but it is important for investors new to this approach realize that the Stock Trends methodology is more about translating technical analysis into systematic trading.

Our bottom-up value approach (“value” in the world of technical analysis is based on market price and volume parameters, not financial parameters of fundamental analysis most often referred to as “value” by the broader investment business) derives regular weekly output. Let’s look at some of these reports and see how our inference model ranks the performance probabilities of the stocks in these reports.

Below is a heatmap showing the current top Picks of the Week and gold stocks as ranked by the Stock Trends inference model (introduced in recent editorials). Of these filter groups - current Picks of the Week, and gold stocks - these stocks have the best probabilities of beating the expected returns of a randomly selected stock. We'll be moving toward creating new reports that detail these rankings for other groups, as well as reports that show over-all ranking derived from the inference model.

Picks of the Week

 

NYSE

 

1.       Equity Residential (EQR-N)

2.       Noranda Aluminum Holding (NOR-N)

3.       ProShares Ultra Real Estate E.T.F. (URE-N)

4.       Endeavour Silver (EXK-N)

5.       Extra Space Storage (EXR-N)

Nasdaq

1.       VocalTec Communications (CALL-Q)

2.       Myriad Genetics (MYGN-Q)

3.       Commercial Vehicle Group (CVGI-Q)

4.       Cavium, Inc. (CAVM-N)

5.       Flextronics International (FLEX-Q)

TSX

1.       Precision Drilling (PD-T)

2.       MAG Silver Corp. (MAG-T)

Top Gold stocks (price >= $2)

 

 

1.       Franco-Nevada (FNV-N)

2.       Gold Resource (GORO-A)

3.       DRDGOLD Ltd. (DRD-N)

4.       Iamgold Corp. (IAG-N)

5.       Virginia Mines (VGQ-T)

Gold stocks (all):

 

1.       Crocodile Gold (CRK-T)

2.       Banro Corp. (BAA-T)

3.       Gogold Resources (GGD-T)

4.       Starcore International Mines (SAM-T)

5.       Golden Star Resources (GSS-A)