STATS4TRADE
  • Company
  • Application
  • Performance
    • Dow Jones 30
    • NASDAQ 100
    • Euro Stoxx 50
    • CAC 40
    • DAX 30
  • Media
  • Blog
  • Contact

Machines and Humans and Trading

3/4/2017

3 Comments

 
Last week in Handelsblatt we discovered a rather interesting and succint article titled "Battle for a Training Edge" about artificial intelligence and trading. The article emphasizes several critical points about algorithmic trading, which we strongly believe investors both private and professional should always keep in mind when considering machine-learning approaches like STATS4TRADE.
Picture
No Machine is Perfect
One of our core tenants at STATS4TRADE is to always be open and transparent. As a result the first point in the article is salient – namely no algorithm can ever claim risk-free trading with only positive returns and no losses. In fact the article prominently quotes Matthew Griffen, CEO and founder of the 311 Institute, who correctly states that no model is perfect. We fully agree – our models are certainly far from perfect. Moreover and as we have stated multiple times before, if you expect to "get rich quick" with our approach, you will be sorely disappointed.

So what is the value of algorithmic approaches such as STATS4TRADE, if they are imperfect? Well as in much of life, the goal is not to be perfect but simply better –  in our case statistically better than pure chance (e.g., greater than 50% probability) at picking winning stocks in the short-term and better than benchmark indices with respect to both return and risk over the long-term. We do this by using machine-learning engines to make data-driven forecasts about upcoming price movements that eliminate subjective yet very common human biases*.

Of course our forecasts are not always right and we incur losses – but the key is that we are right more than we are wrong! In fact if you consider the results for any market on our Peformance pages, we typically have at least a 60:40 win-loss ratio. Compare this result to a casino's probability of taking a customer's money with roulette – almost a 53:47 win-loss ratio on an even/odd bet for the American version! In this case we perform significantly better than even a roulette table in a casino**.

Emotional Humans
The second critical point in the article concerns human emotions and their effect on markets and ultimately prices. Markets are not logical inanimate objects; instead markets are social communities with emotions ranging from joy to fear and every feeling in between. Investors often behave irrationally with suprising regularity. Therefore and to the probable chagrin of some financial academicians, markets are not fully efficient. As a result we can tease obsure statistical patterns of behavior out of long-term historical price-data. In turn this enables us to raise our probability of making a winning trade above chance.

However and as Greyfeather Capital's Matthew Sandretto accurately notes in the article, the trick for any algorithm is the balance between illogical behavior and logical conclusions. Again no algorithm is perfect at drawing conclusions from irrational human behavior – especially over short timespans. Therefore the models from our machine-learning engines strive to balance accuracy with consistency over varying timespans – all verified by rigorous testing with both training- as well as test-data. In effect we statistically optimize our models for both return and risk over the long-term instead of the short-term.

Machines on the March...
Now comes the third and fourth points. Clearly trading algorithms are on the march with much potential for growth. As a result the financial services industry faces drastic changes in transitioning from human-centric to machine-centric decision-making processes. Just like managers in industries such as transportation and manufacturing, fund managers now face a stark choice: either adapt to the oncoming disruptions or likely perish.

Bluntly stated – in comparison to highly-paid analysts and traders, machines are reliable, consistent and cheap. Moreover and as Alessandro Di Soccio of A.I. Machines cites in the article, when expenses are considered, only about 1% of today's asset managers actually outpeform benchmark indices while 25% underperform. No wonder investors are frustrated and flock to passive index funds! As a result it will be increasingly difficult for any fund to justify high fees and mediocre performance in the future. 

...but Humans Still Needed
Nonetheless this trend towards machines in trading does not mean that humans will become completely superfluous. As Andrej Rusakov of Capital Data Management mentions in the article, it will always be the responsibility of a human – say a fund-manager – to set the parameters for any automated investment strategy. At STATS4TRADE we use the analogy of an automatic pilot. In modern aircraft a computer normally pilots the plane; however if an emergency occurs, a human pilot is always available to intervene and control the aircraft herself. It will be no different in the future for "financial pilots".

No Magic Wands
The article's ends with an important quote from Alessandro Di Soccio of A.I. Machines. He states about artifical intelligence (AI):
People think of AI as a sort of magic that can solve everything. That is not the case.
We cannot agree more! We cannot wave our magic wand and create models that are perfectly accurate. However we can offer investors inexpensive data-driven applications that allow investors the opportunity to design their own portfolios, which increase the odds of a postive trade above chance and consistently beat benchmarks. Can well-known retail and hedge funds say the same? You decide.

Of course we do not ask that you fully accept our analysis or conclusions. If you have different ideas, please do not hesitate to post a comment or contact us via email at contact@stats4trade.com. As always we look forward to hearing from you.

*See for example the article "​Why We Think We're Better Investors Than We Are", G. Belsky, New York Times, 25.March 2016.
**Why anyone would ever place such an irrational bet with losing odds is a question for human psychology.

3 Comments

Free Trading with Robinhood

27/3/2017

2 Comments

 
We mention on our Application and Performance pages that we use Robinhood to trade live with our own capital and buy/sell signals from our beta release for private investors. However we have noticed that Robinhood is still relatively unkown among our user community – even with glowing articles like this one in the New York Times. Therefore and in this blogpost we thought that it would be a good idea to introduce you to Robinhood through the lens of our trading experiences over the past months.

However before we begin please note the following important point. Namely we have no business relationship with Robinhood – or any other brokerage. Therefore our views are unbiased and based purely on our experiences with our own capital and buy/sell signals from our application over the past months.

Quick Introduction
Robinhood is a relatively new, online brokerage that no charges no fees to trade in US-based equities. Yes, you read the last sentence correctly – no fees to trade stocks! Of course there are caveats associated with its no-fee service (see details below). Nonetheless the fact remains that Robinhood is the only no-fee brokerage available today.

The ability to offer brokerage services at no cost is groundbreaking. For the first time private investors have direct access to markets in the United States without paying a "gatekeeper". In turn the elimination of fees presents investors new opportunities to trade actively instead of resigning themselves to buy-and-hold strategies that minimize turnover and hence trading costs.

We take advantage of this new no-fee environment and offer private investors a low-cost and automated approach to actively trading in equities. Indeed if not for no-fee brokerages like Robinhood our ambition to democratize investing would be much more challenging indeed.

And how does Robinhood earn money, if it does not charge fees? Well Robinhood invests any cash from customers' accounts in risk-free and liquid US Treasury securities. Now these risk-free investments do not yield much return. Therefore Robinhood relies on customer volume because its margins are rather thin.

Account Set-Up
The process for creating an account is easy and starts on Robinhood's website. Simply click the "Sign Up" button. Afterwards Robinhood walks you through the entire procedure for US residents including approval, downloading the application to your mobile device (IOS or Android), and ultimately transferring funds from your bank to your new Robinhood account.

​Note that Robinhood requires no minimum deposit – the minimum deposit is only limited by the price of the stock(s) that you wish to buy. Moreover one can begin trading immediately for deposits up to US$1000. However for transfers greater than US$1000 allow about five business-days until the full amount of your deposit is available for trading. In Robinhood parlance any cash amount actually available to buy stocks is called "buying power".

Buying and Selling
To purchase a stock you simply search for the stock symbol – for example JNJ for Johnson & Johnson. Then you enter the amount of shares that you wish to buy – say thirteen shares per our lastest buy signal for JNJ (from our standard portfolio composition for private investors on 3 March). Next click "Done" and swipe downward to initiate the trade. Robinhood informs you from the application as well as via email that the order has been received and then executed. After you buy a stock you can track your performance in real-time via Robinhood's application.  

To sell a stock is just as easy. You select the position – again JNJ – and click the "Sell" button. You then select the number of shares to sell and swipe downward to initiate the trade. As when you buy, you receive confirmation of your sale from the application as well as via email.

However there is one large caveat when you sell.  Namely you must wait about five business-days to access cash from your sale – either to transfer back into your bank account or buy additional stocks. Of course though any gains (or losses) are booked immediately on our account.

For many investors the time restriction on access to cash from any sales is frustrating. However one should always remember that trading with Robinhood is free. Instead Robinhood earns money from its short-term risk-free investments in US Treasury securities. So this is the true price of trading with Robinhood – you essentially give Robinhood a one-week free loan with any sales. Furthermore you can always upgrade to Robinhood Gold, which charges fees but also allows immediate access to cash from sales.
​
Overall Experience
So what is our overall experience using Robinhood over the past months? Well we are happy to report that we are quite pleased with Robinhood. Sure there are time restrictions associated with deposits and sales but these restrictions are not blocking factors for typical investors. Moreover we can absolutely state that trading via Robinhood's mobile application is both quick and easy.
Picture
New Markets and Potential Competitors
For now Robinhood is restricted to the United States. However Robinhood soon plans to roll-out a similar no-fee service in Australia and eventually China. Unfortunately Robinhood has no current plans for Europe.

Nonetheless there is a potential competitor to Robinhood in Europe called DEZIRO. According to its website DEZIRO soon plans to offer no-fee trading in Europe but the exact availability date is unknown.

However DEZIRO's parent company DEGIRO offers no-fee access to US markets from Europe. Therefore and like investors in the United States with Robinhood, European investors can trade in US markets without fees via DEGIRO. In the meantime DEGIRO offers traditional fee-based brokerage services throughout Europe.

If you have any questions about Robinhood, how we use Robinhood to trade in the United States with our buy/sell signals, or DEGIRO in Europe, please do not hesitate to contact us anytime at contact@stats4trade.com. We look forward to hearing from you!
2 Comments

Be Very Careful About "Sure Bets"

22/3/2017

0 Comments

 
Well it has happened again. Yet another large well-known hedge fund has suffered a huge loss. This time it is William Ackermann's Pershing Square Capital Management. As detailed in the New York Times by Gretchen Morgenson and Geraldine Fabrikant, Pershing Square reported an incredible US$4 billion loss on 13 March – all caused by a "sure bet" on one company, Valeant Pharmaceuticals.
Picture
This huge loss is a perfect example of how risk is greatly amplified when one subjectively bets large amounts on one stock. There are no absolutes in markets or any other type of human behavior – Mr. Ackermann should know this already! And if he does not, then why he is leading a well-known hedge fund? Or even more salient, why do investors invest in his fund with such risk?

We at STATS4TRADE adamantly believe that it is a fool's errand to predict price movements with absolute certainty for "sure bets" never exist. Morever it is extremely risky to bet everthing on just one stock.

​Instead we fervently believe that markets are complex with stochastic characteristics that lend themselves to statistical approaches like ours – approaches that increase the probability of picking stocks with positive returns while enforcing diversification to reduce risk. Sure we might miss the stellar returns that a bet on just one stock can yield – say Apple at the moment. But we also minimize the risk of "putting too many eggs in one basket" like Mr. Ackermann.

We ask that you think about Mr. Ackermann's investment process and ask yourselves – why would anyone pay his fund's fees considering such an astounding yet preventable loss? Investors expect and indeed demand better performance. We do too.
0 Comments

Warren Buffet's Indictment of the Hedge-Fund Industry

28/2/2017

1 Comment

 
Another article just caught our attention in today's New York Times. Andrew Ross Sorkin reports that Warren Buffet has condemned the poor performance and high fees of hedge-funds in his annual letter to Berkshire Hathaway shareholders.

We fully agree with Mr. Buffet's harsh criticism of the hedge-fund industry and invite you to read the article via the link below.
Picture
1 Comment

Exactly the Problem

17/2/2017

0 Comments

 
We are in the midst of a beta rollout of our application for private investors and extending our website. However something caught our eye today in the financial press that we just cannot overlook. The Wall Street Journal reports that Catalyst Hedged Futures Strategy Fund lost $600 million or 15% of its value in just one normal week.
Picture
To put it mildly this news is absolutely astounding not to mention troubling. First Catalyst Hedged Futures Strategy touts itself as an active-fund that protects capital. A 15% loss in one week without any catastrophic event is protecting capital – especially for small investors? Clearly by any measure the fund's managers have completely failed in this regard.

Second according to Morningstar the fund's annual management fee alone is 1.75%. And the fees do not stop here. The fund nominally charges an upfront sales fee of 5.75% of initial value (e.g., a sales load) as well as various additional annual fees such as marketing that bring the total annual fee to 2.17% in 2016.

So let us summarize. Catalyst Hedged Futures Strategy touts it capital protection strategy yet loses 15% in one normal week and charges high fees. Is this the peformance that investors expect? We think not.

This is exactly the problem today. Actively-managed funds are not meeting investors' expectations. As private investors ourselves we demand transparency with consistent performance in terms of both return and risk as well as fair fees that eliminate excessive annual management fees and any sales/redemption loads.

In view of this most recent example we encourage our user community to continually question their funds' performance and fees – including pension fund arrangements such as tax-deferred 401(k) structures in the United States. Furthermore and like our tagline "Democratize Investing" states, we support an on-going democratization of the investment process by giving private investors the necessary tools to bypass expensive underperforming funds and actively manage their own investments.
0 Comments

New Markets and Website

21/1/2017

0 Comments

 
We have been diligently working over the past months to add more markets and revamp our website with new metrics for portfolios tailored for both private- and professional-investors.  In addition we are almost ready to debut our trading application for private-investors.  

There are many new details to explain to our user-community – in fact too much for one blog-post. As a result we intend to introduce a series of upcoming blog-posts that explore our new metrics, approach and applications in more detail. So please stay tuned!

In the meantime and as always please contact us anytime if you have any questions or feedback. We look forward to hearing from you.
0 Comments

Bug Fix for Dow Jones Forecast-Data

22/11/2016

2 Comments

 
Late last week we noticed an anomaly with our Dow Jones forecast data.  In particular the daily list of buy-recommendations has apparently for some time only listed the ISIN code for the Dow Jones index itself. Moreover the charts reflected a corresponding and unusual lack of movement.

We spent the last few days investigating the root-cause and ultimately discovered a subtle software bug, which we fixed today. Now the charts and list of buy-recommendations are again accurate.

We sincerely apologize for the bug, which was associated with the daily synchronization between our external price-database for US stocks and internal software environment.

If you have any questions, please do not hesitate to contact us via the contact formula on this page or email at contact@stats4trade.com.  
2 Comments

New Forecast Format

1/9/2016

 
Well it has finally arrived – our new forecast-data format is now available.  However the new format is quite different from our prevous format. Therefore we decided to provide our user community an introduction in this blog-post.  More detailed information can also be found on our new Help page.

Motivation for Change
First a little background as to why we changed the forecast format.  Over the past few months since the release of our initial forecast format in early April, we came to the realization that our initial format was unwieldy in terms of both useability and flexibilty. This conclusion was based not only on our opinions but also early feedback from our user community. As a result we embarked on a development effort aimed at improving the format with respect to ease-of-use and options for customization.

More Useability and Flexibility
Compared to the old format, the new format is much simpler and therefore both easier to understand and implement.  For example before we listed raw probabilities for price movements relative to ceilings and floors.  Now however we have condensed these probabilities and levels into a simple buy recommendation – updated daily and valid for twenty trading-days.

In addition the new format offers more opportunities for customization than before.  Whereas the old format only offered forecast data for the CAC 40 and S&P 500 indices themselves, we now provide forecast data for all stocks in the CAC 40 and Dow Jones 30 indices (depending on interest among our user community, we plan to add stocks in the S&P 500 at a later date).  Therefore one now has access to a much wider range of invidividual securities.  In turn this allows the development of unique investment strategies focused on specific stocks or groups of stocks across markets.


Key Features
Our new forecast format incorporates some key features, which we believe you will find useful for your investment activities.

As mentioned above we now provide daily buy recommendations for all individual stocks in the
CAC 40 and Dow Jones 30 indices.  The recommendations are based on forecasts of price-increases from our machine-learning algorithms and are valid for the next twenty trading-days.

For each index we now provide charts of portfolio performance for a simple
investment strategy based on our buy recommendations. Namely each day we buy all the recommended stocks in our forecast.  Then for each stock we buy, we sell it twenty trading-days later. The charts are interactive and continuously compare our relative performance since 1.January.2014 with the performance of a buy-and-hold strategy for the underlying index.

The same charts of portfolio performance also allow one to track the amount of what we call "
capital invested" – that is the amount of capital invested in our portfolio of buy recommendations at any moment in time. As you will see with our portfolio, one is not rarely 100% invested in the market like traditional buy-and-wait strategies. This fact offers several advantages including reduced volatility and thus risk as well as freeing capital over extended timespans.

Each time we sell a open investment-line after twenty trading-days we incur a return – hopefully positive! We feel it is important to show you these returns.  Therefore we provide pie-charts that show the relative frequency of returns for investment-lines based on our current investment strategy

Additional Information
We have addressed only the highlights of our new format in this blog-post. For additional information, again please see our new
Help page. In addition we intend to publish several follow-up blog-posts that expand on specifc details of our new format.

Overall we hope that you find our new forecast format and portfolio summaries not only more useful but also more intuitive for your investement needs.  In the meantime we thank you for your patience.  And if you have any questions, please do not hesitate to contact us – simply use the contact-form in the right-hand margin. We very much look forward to hearing from you.


Changes Have Arrived!

26/8/2016

 
As we mentioned in our previous blog-post, changes are coming.  Now these changes have arrived! 

The first change occurs this weekend with the end of our current forecast-data format over four-days with ceilings and floors for the CAC 40 and S&P 500 indices. In the coming days we will replace this format with a new one consisting of specific buy-recommendations over the next twenty-days for individual stocks in the CAC 40 and Dow Jones 30 indices. 

Since April we have also issued a summary of our forecast results each weekend. This weekend we also cease publishing this summary after twenty-one weeks.  Instead we will soon provide new portfolio summaries – updated daily and in an interactive chart-format – for investment strategies based on our new forecast format.

Again we hope that you find our new forecast format and portfolio summaries not only more useful but also more intuitive in terms of your investement needs.  In the meantime we thank you for your patience. And if you have any questions, please do not hesitate to contact us.  We sincerely look forward to hearing from you.

Changes Are Coming

9/8/2016

 
If you have visited our website over the past weeks, you have noticed a few additions like new pages.  These additions reflect much internal change – change that will very soon be reflected on our website. 

New Forecast-Data Format
The biggest change is the format of our forecast data.  Presently we publish a four-day forecast with ceiling and floor probabilities for the CAC 40 and S&P 500 indices. However very soon we will switch to a twenty-day forecast for individual stocks in the CAC 40 and Dow Jones indices. Stocks in other major indices will follow later. 

The new forecast format will contain two significant improvements in terms of both useability and flexibility. First we will provide daily buy-recommendations for stocks in each index. The buy-recommendations will be straightforward and include a list of stocks, whose prices are forecast to rise in the next twenty trading-days with a corresponding probability. Second we will publish easy-to-understand performance summaries that compare our on-going performance to a buy-and-wait strategy for the corresponding index.


More User-Friendly and Flexible
After much testing we are now convinced that our new approach is more user-friendly than our previous approach. Moreover the new approach allows our users more flexibility in developing their own investment strategies across groups of stocks. For example one can now tailor an investment strategy to exclude traditional energy companies and only include environmentally-sustainable stocks.

Of course we intend to publish upcoming blog-posts that explain our new forecast approach in detail. Nonetheless these posts will come after we change our forecast format in the next two weeks.  Therefore we wanted to give you advance notice of our near-term change in this interim blog-post.

If you have any questions, please do not hesitate to contact us.  As always we look forward to hearing from you. In the meantime sit tight and get ready for our new forecast format!

How Do We Do It?

23/6/2016

 
We just showed you an objective approach to investing with our forecast data that yields both decent and consistent returns. But at this point you might have some questions about our forecast data. For example how do we generate our data? What it is our accuracy? Are there any important considerations to keep in mind? In this blog-post we take a step back from finance to answer these and other questions about our forecast data.

Identifying Market Patterns
Simply stated our goal is to identify market patterns for a specific security or a group of securities like an index. However markets are extremely complex, often illogical and highly random environments. Therefore it is impossible to create a solely deterministic model of market behavior. A better approach is to recognize the inherent randomness of markets and adopt a statistical approach that models markets in terms of statistical parameters over time. This approach implies the use of some type of statistical model, from which we can identify statistical patterns over time and ultimately forecast the likelihood of particular events occurring – such as a price decrease in the next four days.


The first step in generating our forecast data begins with a statistical model for a particular security or group of securities. The process for creating a statistical model is shown in the schematic below.
Picture
Everything begins with a machine-learning algorithm that constructs, trains, evaluates and tunes a statistical model using so-called "training-data". In our case the training-data are historical price-data for a particular security, basket of securities, or an index like the CAC 40. During the training process the machine-learning algorithm attempts to find relationships among statistical parameters in the underlying price-data over different timespans. As you can imagine, this process is quite complex mathematically and relies on robust machine-learning tools that abstract much of the complexity away from the end-user while still yielding valid results. Nonetheless our end-goal always remains the same no matter the complexity: the creation of a statistical model that identifies statistically-significant patterns in market behavior.

For machine-learning algorithms we rely on BigML, a leading US-based provider of such tools with worldwide presence. BigML's software environment and training materials along with its deep expertise in machine-learning allow us to not only create and train statistical models but also iteratively improve results. As a result and in close collaboration with BigML, we are able to optimize our statistical models for both accuracy and consistency over long timespans with various securities and indices.

Model Verification is Key
Just because a statistical model for a specific security or index has been created does not mean one is finished. Actually far from it!  Instead the tedious yet necessary process of testing the model now begins. This step entails testing the model results against historical prices over many timespans.  We want to ensure both statistical accuracy and consistency against the actual price-record regardless the timespan.

Of course we must be very careful during model verification so as to avoid only testing timespans, during which we train the model. Why? Well this would induce a bias into the verification process. Essentially we would be testing a model using the same price-data, which we originally used to train the model.  In other words, "training-data in, training-data out". Therefore we also test against price-data, which was not used to train the model. This ensures that we truly vet the underlying statistical model and avoid fooling ourselves with biased test results.

How Have Our Models Performed?
Each week from April to August 2016 we published a summary of the forecasts from the previous week for an earlier forecast format. These summaries also included a running tally of the accuracy of our forecasts since we began publishing forecast data in early April. But how have we performed over longer periods of time than just six months? Let us take a look at our historical performance with the CAC 40 index from our previous blog-post.

First consider the ceiling forecast data shown in the chart below from 1990-2015. Along the horizontal axis are our three confidence intervals:  50%, 60% and 70%. The vertical axis shows the actual success rates for each of these confidence intervals. The sample sizes are listed in each data column (e.g., n = 1097 for the 50% interval).
Picture
Over time with ever increasing sample sizes, our success rates should converge to the given confidence intervals. Indeed this is the case as each of the success rates is within (or almost within) its respective confidence interval. For example when we state there is a 70% chance that the price goes above the ceiling in the next four days, on average we are correct 70% of the time.

Second consider the floor forecast-data shown in the next chart.
Picture
Like the ceiling the success rates for the floor converge to their corresponding confidence intervals. As a result a forecast of "70% stay above" really means that about 70% of the time the price remains above the floor.

However the results in the previous two charts contain significant timespans with training data – that is data used to train the statistical model. Therefore we must also consider timespans without training data.  For example let us now compare the timespan 2013-2014 without training-data against the timespan 1990-2013 with training data. We begin with comparisons for the ceiling forecasts in the following two charts. The first chart is for the "go above" forecasts; the second chart is for the "stay below" forecasts.
Picture
Picture
As you can see the results still converge to the expected confidence intervals with little significant difference between time-spans with and without training-data. This indicates that the underlying statistical model is insensitive to timespans with and without training data.

Now consider the floor comparisons with and without training-data in the two charts below. The first chart is for "go below" forecasts while the second chart is for the "stay above" forecasts.
Picture
Picture
Just like the ceiling results, the floor results converge to the same confidence intervals. Exactly what we want and indeed expect to see – assuming of course that we have done our job correctly!

Important Considerations to Keep in Mind
The first and foremost consideration involves the potential bias from too few data samples – either for model training or during result interpretation. For example just consider our weekly summaries from April to August 2016. Although we collected twenty-one weeks of data, we still have only about ten data samples for each confidence interval.

Unfortunately too few data samples yield significant uncertainty in our results. Therefore when we train or verify models, we use at least two years’ data (although it is usually much more than just two years). Typically two years provides about sixty or more data samples. This is exactly the reason why we clearly state: if one is expecting to become rich quickly with our forecast data, she will be sorely disappointed. Instead one needs about two years to begin seeing decent returns relative to traditional investment strategies like buy-and-hold.

The second consideration to keep in mind involves the trade-off between model accuracy and consistency over different timespans both short and long. Sure one can have higher probabilities up to 90%. However, consistency will suffer. Nothing in life is without compromise – our models are no exception. Therefore we always aim to create models that balance accuracy over any given timespan. Such an approach demands a continuous balance between accuracy and consistency with neither characteristic being optimized alone but always in combination.

The final consideration to remember involves yet another bias – this time associated with training-data. Imagine for example that you have training-data for a timespan in which the market continually trends upward. Such a trend indeed occurred from about 1995-2000. Now suppose that you train your statistical model with data only from this timespan. What would happen? Well if you then used the model to forecast performance after 2000, you would be rather disappointed for the years immediately after 2000 that are characterized by significant market declines across a wide-range of securities. Therefore it is critically important that we use training-data from long timespans that contain realistic market behavior with both gains and declines. Likewise it is important that we test the models with and without training-data over a range of timespans with realistic market behavior.

Generating Value with Our Forecast Data
As you can see, the process of generating our forecast data with statistical models incorporates several steps and involves state-of-the art tools from BigML. However the overall goal always remains the same. Namely we aim to identify underlying statistical patterns for a given security, basket of securities, or an index and thereby accurately forecast upcoming movements in price.


The process of generating both accurate and consistent forecast data underpins our focus on creating sustainable value for investors in the form of decent yet stable returns over a number of years instead of decades. As a result we offer small investors a better path to investing. A path characterized by data-driven trading decisions within an automated and configurable software framework – all without relying solely on subjective information, error-prone emotions or expensive financial advisors.

In closing we again emphasize the same point that we made several times before in our posts. Please do not trade with our public forecast data! At this point we are simply trying to educate you about our approaches by being as transparent as possible. Nonetheless if you insist on trading with our data, then you assume full responsibility for any outcomes.

Next Blog-Post
In our next post we will introduce new forecast-data types and associated investment strategies, which we have been testing in parallel over the past months. These new data-types and strategies yield higher, more consistent returns than the simple strategy outlined in the last post with the current four-day forecast data.


As always let us know if you have any questions or feedback. We are always happy to hear from you.  And stay tuned for upcoming blog posts!

Path to a Better Way

2/6/2016

 
It has been several weeks since our last blog post.  In the meantime we have been busy confirming several investment strategies with extensive back-testing as well as improving our models and developing new statistics. We apologize for the delay. However we wanted to ensure the validity of our data. Now we are ready to release some initial results.

As we mentioned in the last post, our raison d'être is using science – namely machine-learning tools and statistics – to help small investors achieve decent returns with low volatility. OK, great. But what does this really mean? Well in this post we show what is possible with a simple strategy, which uses our public data and strives to beat a traditional buy-and-wait strategy for a difficult index such as the CAC 40. What is important to remember in this example is the underlying message of how statistics can help you make consistent, data-based decisions to achieve modest yet stable returns. All without passively waiting decades, using expensive financial advisors, or relying on subjective "gut-feelings".

So what exactly is our simple strategy? Well as you know each night in Paris just after markets close in New York, we publish a forecast over the next four days for the CAC 40 (we also publich a similar forecast for the S&P 500). Each forecast contains two probabilities:  one probability for the ceiling, which is above the last closing price; and one probability for the floor, which is below the last closing price. These two probabilities are important because they serve as buy-triggers for our simple strategy as follows.

Buy-Triggers

▪ Go above ceiling:  50%, 60%, or 70%
▪ Stay above floor:  70%


We also have two sell-triggers. If our forecast is correct, then we should be able to sell at the end of four days with a gain. However there is always a chance that our forecast is incorrect. Therefore in those cases in which the price falls below the floor contrary to the forecast, we sell at the floor before the end of four days as outlined below.

Sell-Triggers

▪ Sell if price goes below floor
▪ Otherwise sell at end of four days


The process is straightforward. If the next day's forecast for the CAC 40 satisfies the buy-triggers, we buy shares in a CAC 40 index fund the next day immediately after market-open. Then we track the position over the next four days. If the price ever drops below the floor, then we sell the position. Otherwise we sell the position after four days. We repeat this process every time there is a buy-trigger.

Seems simple enough, right? So how did we perform in back-testing against a traditional buy-and-wait strategy for the CAC 40 over the past twenty-five years since 1990?


CAC 40 Results
In the following chart we list our annual returns* and variances** without dividends as well as the returns and variances from a buy-and-wait strategy without dividends for the CAC 40 (see our previous blog entry). The timespan is 1990-2015.
Picture
Across the entire timespan from 1990-2015 our simple strategy for the CAC 40 delivers a higher return with lower variance than a traditional buy-and-wait strategy. In particular over twenty-five years we deliver an average annual return of 8.5% while a buy-and-wait strategy delivers an average annual return of 3.1%. Furthermore our returns have about 7% lower variance.

Now let us focus on the results from individual decades as shown in the table below. Note that the identical returns from 1990-1999 and 2010-2015 for our simple strategy (labeled STATS4TRADE) are coincidental and only due to rounding.
Picture
First consider 1990-1999, the latter half of which was characterized by a speculative run-up in prices. During this decade we realize a 12.8% return versus 11.3% for the build-and-wait strategy.  Moreover we have almost 9% less variance.

Now consider the following decade from 2000-2009, which witnessed two large declines at the beginning and towards the end during the global financial crisis in 2008. During this decade we yield almost 6% more return than the buy-and-wait strategy:  1.7% compared to -4.2%. Again our variance is about 8% less.

Last consider the period from 2010-2015, during which the trend from the previous decade continues. We achieve a 12.8% return versus 2.7% for the build-and-wait strategy. As during the previous two decades we still have less variance but the difference this time is less at about 3%.


What About Trading Fees?
Until recently small investors have not been able to avoid trading fees. However just like in other service sectors such as transportation (Uber and Lyft come to mind), new companies are beginning to disrupt and challenge the traditional fee structures in financial services. One good example is Robinhood, a California-based discount-broker, which charges no fees to trade in the United States.

According to its website, Robinhood is presently expanding into Australia and China. Therefore it is in the realm of possibilities that a European version of Robinhood will be launched soon. Or perhaps DEZIRO will actually launch? If so then one could avoid trading fees for the CAC 40 as well.

Unfortunately however trading fees are still a fact of life today in France and the rest of Europe. Therefore we must now show you the drastic effect that trading fees have on our returns. In this case we have assumed a 0.2% fee each time we trade (0.2% subtracted from each individual return). See the following chart.
Picture
Over the twenty-five years from 1990-2015 we now deliver only 2.4% instead of 8.5% without fees and versus the unchanged 3.1% for a buy-and-wait strategy (we have neglected the relatively minuscule fees for a buy-and-wait strategy). The only consolation is the fact that our variance is still about 7% lower.

Like before let us look at the results with fees from individual decades as shown in the table below.

Picture
First from 1990-1999 we now realize only 6.5% return instead of 12.8% without fees and the unchanged 11.3% for the buy-and-wait strategy. Again our consolation is about 9% less variance.

Second during the decade from 2000-2009 we now yield -4.1% instead of 1.7% without fees and the unchanged -4.2% for the buy-and-wait strategy. However we still have about 8.5% less variance.

And last from 2010-2015 we now only achieve 6.5% return instead of 12.8% without fees and versus the unchanged 2.7% for the buy-and-wait strategy. The difference in variance is unchanged at about 2.5%.

Wrapping It All Up
As you have just seen, even a simple data-based approach for making trading decisions can yield decent, stable returns compared to a traditional buy-and-wait strategy. However the simple strategy outlined in this post is certainly not without its limitations – namely trading fees, which essentially strip away any excess returns versus a buy-and-wait strategy. Nonetheless the message is clear. A data-based approach works. Moreover with the advent fee-free trading portals like Robinhood, trading fees will decrease even in Europe. It is just a question of time. Therefore data-driven approaches like ours will become increasingly attractive to small investors across the world as companies like Robinhood proliferate.

At this point we must emphasize a point that we made in an earlier post. Namely please do not trade with our public data using any strategy including the simple example outlined in this post. We have not yet clearly outlined all the limitations associated with our data (trading fees are just one example). These limitations embody significant risk especially over the short-term. We are working now to reduce these risks with improved approaches, which we outline in coming posts. 
However if you insist on trading with our data then we must be clear: you assume full responsibility for any outcomes. Therefore please just wait for now!

Next Steps
Presently we are back-testing improved strategies with other types of statistical data and forecasts longer than four days. We are certain that these efforts not only yield better results over shorter timespans but are also more useful for the small investor. We plan to publish these approaches and their results in upcoming blog posts.

However our next blog post is not about improved strategies or even finance.  Instead we take a step back and show you process of generating our stastical data as well as historical success rates from our back-testing. Statistical data underpin all of our processes; hence an understanding of these data – including limitations – is essential for understanding our overall approach.

As always if you have any questions, please do not hesitate to contact us anytime. Just fill out the comment form next to this text with your name, email address and question. Please do not be shy – we sincerely look forward to hearing from you!
*The returns for our simple strategy are arithmetically calculated without dividends on an annualized basis. We use an arithmetic average because we are often unable compound – buy-triggers often occur on successive days, which prevents straightforward compounding. On the other hand returns for the traditional buy-and-wait strategy were calculated geometrically with compounding on an annualized basis again without dividends.

**Variance is calculated using the standard method.  More specifically we caclulate the arithmetic mean of the individual returns within a year, sum the square of the difference of each return with the mean, and normalize the sum by dividing by the number of returns.

Can You Really Make Money by Buying and Waiting?

17/4/2016

 
Our raison d'être is using science to help risk-averse small private investors achieve decent returns with as little volatility as possible. But do private investors actually need our help in the first place? For example consider putting your money into a diversified index fund and waiting decades. Would not such a traditional buy-and-hold approach yield decent returns with low volatility? Normally yes – but beware. This strategy can yield poor results with rather high volatility for some indices. For example just consider the performance of French CAC 40 over the past two decades.

The CAC 40 index is a diverse weighted stock-price average of France's forty largest public companies including such internationally famous names as Airbus, L'Oreal and Michelin. As a result the CAC 40 should serve as an ideal index for the risk-averse private investor in France with a buy-and-hold strategy over the long-run. But what is the peformance of the CAC 40 over the last two decades? Well consider the following graph, which shows the performance of the CAC 40 between 1988 and early 2016.
Picture
Copyright © 2016 New York Times.  All rights reserved.
At first glance two things stand out in the graph. First note the three-fold increase between the years 1996 and 2000.  Second note the large gyrations between 3000 and 6000 from the year 2000 until the present.

Now consider the hard numbers shown in the chart below. Between 1990 and 2015 the average annual return for the CAC 40 is 3.3% without dividends and 6.5% with dividends. And between 2000 and 2015?  The average annual return is -1.6% without dividends and 1.7% with dividends. Moreover both of these returns came with rather high volatility. For example as a conservative estimate of volatility between 1990 and 2015 the standard-devations of annual returns without and with dividends are 22% and 26%, respectively.
Picture
Source:  Euronext.  *Annual averages are geometric with compounding.
So what does all of this mean? Overall it is clear. A buy-and-hold strategy for the CAC 40 over the past twenty-five years and particularly since 2000 yields low returns with rather high volatility.
Of course however we do not proclaim that a buy-and-hold strategy with diversified indices is poor.  Quite the contrary! Just consider the performance of the S&P 500 as represented by the yellow line in the comparison graph below. Between 1990 and 2015 the S&P 500 yielded average annual return of 7% without dividends and 9.3% with dividends as well as lower volatility with a standard-deviation of 18% over the same time period. So the performance of the S&P 500 is clearly better than the CAC 40 in terms of a buy-and-hold strategy over the past two decades and particularly since 2000.
Picture
Copyright © 2016 New York Times.  All rights reserved.
Overall we strongly believe that buy-and-hold strategy is absolutely valid for the risk-averse private investor – especially if one considers trading fees and the cost of financial services. However we also recognize that relying on such a strategy alone is not without potentially bad surprises depending on the index. The CAC 40 is a good example (actually the Nikkei 225 is an even better example). Moreover even with the S&P 500 one needs a lot of patience with a buy-and-hold strategy over decades.

So the fundamental question remains: how can the risk-averse private investor earn decent returns – say in the range of 6-9% including all fees – with low volatility over shorter time periods than decades? Glad you asked because this is the basis of our business, which we introduce in more detail in the next post of this two-part series.

In the meantime if you have any questions about this topic, please do not hesitate to contact us via email. As always we look forward to hearing from you.

What Is Your Tolerance for Risk?

4/4/2016

 
We will use our forecast data for our own trading activities based on different investment strategies.  In turn our strategies are driven by our financial goals – namely decent stable returns over several years with risk as low as possible as measured by volatility.  However your financial goals and tolerance for risk might be different.  Therefore if you want to use our investment strategies and forecast data to help you reach your financial goals, you should also know your particular tolerance for risk before you begin.

So what risk profile do you have?  Are you cautious like us and always try to preserve capital and dampen volatility?  If so then you are very careful and typically invest smaller amounts across a wide-range of securities and indices to maximize diversification and lower overall volatility*.  On the other hand are you a gambler?  If so then you are prepared to invest larger amounts in a fewer number of securities, and perhaps even consider short-selling.  Or are you somewhere in between?

Most of the time your tolerance for risk will depend on how much you can lose at any one time without seriously damaging current liquidity or long-term stability.  For example most of us can afford to lose €1000 (or its equivalent).  But what about €10000?  Or €100000?  At some point such losses start to seriously threaten our financial security both now and in the future.  Therefore and before executing any strategy it is imperative to determine how much you are willing to lose without risking serious harm to your financial health.

We are undoubtedly risk-averse.  Therefore we believe in conservative investment strategies, which minimize volatility and large losses over time.  We strongly believe and indeed intend to prove that periodic, objective, data-driven trading with our strategies and forecast data yields decent returns with less volatility over reasonable time spans of a few years – even in very difficult markets such as the CAC40 over the past two years.  In coming blog entries, we intend to outline our investment strategies, which will provide insights into our methodologies and hopefully assist your own investment activities.

As always, if you have any questions or feedback, feel free to contact us anytime. In the meantime stay tuned for upcoming posts about our investment strategies and how our strategies can help you too.


*Note that risk is mathematically measured by volatility, which in turn is simply the standard-deviation of returns over time.

Setting Initial Expectations

13/3/2016

 
Like any long-term relationship trust is the absolute basis for our success.  Therefore we feel it is absolutely essential to clearly set some basic yet important expectations among our user community from the very beginning.

Estimating Outcomes
Someone once said that predicting the future is a fool's errand.  We agree.  However one can still use stastitics to estimate the likelihood of future events based on past data and an underlying statistical model.  In fact statistical methods have been used extensively for years in activities like consumer research, weather forecasting and of course finance.  In our case we use powerful machine-learning methods to estimate the probability of short-term price movements of selected securities and indices, currencies and commodities.

Statistics Implies Uncertainty
Statistics implies a stochaistic world with random events.  As a result uncertainty is ever present in any statistical data.  For example a high probability of a price increase in the coming days is no guarantee that a price increase will indeed occur.  Moreover there are always "black swan" events like natural catastrophes or economic turmoil that can never be predicted using statistical methods.  Therefore please always keep this inherent uncertainty in mind when interpreting any statistical data such as probabilities.

Know Your Risk Appetite
Risk is aways associated with uncertainty.  For example the probability may be reasonably high (say above 70%) that a stock price increases in the next few days.  As a result you decide to buy the stock now.  However there is still a real chance that the price actually decreases and you lose money. It all depends on price volatility and the amount you invested.

​Therefore it is absolutely essential that you know your individual appetite for risk.  Are you risk-adverse like us and tend to invest small amounts in a diversified range of securities with less volatility?  Or do you embrace risk by investing larger amounts in less securities with more volatility?  It all depends on you so it is absolutely imperative that know your particular appetite for risk before you begin.

Our Initial Recommendation
During our initial alpha and beta phases we are introducing both investment strategies and forecast data to the general public.  The investment strategies will have been back-tested but not yet live-tested.  Therefore we ask that you refrain from using our investment recipes and public data for any trading purposes until our alpha and beta phases end and we have had a chance to conduct live testing.

Why do we emphasize this important point again?  Well we want to again emphasize mutual success for us as well as you!  As a result please wait until we have had a chance to fully validate our recipes and prove legitimacy over the next few months.  We will certainly inform you of any changes to our policy via this blog.

Legal Disclaimer
Now we must add the standard legal dislaimer if you ever decide to use our investment strategies and trade with our forecast data.  If you use our investment strategies and forecast data we cannot accept responsibility for your results including any success or lack thereof.  We hope that you understand that in today's legal climate we must mention this fact. 

As always if you have any questions, please feel free to contact us anytime.  We will be more than happy to help you or answer any question. In the meantime we look forward to hearing from you and sharing our thoughts.

Ready, Start, Go!

13/3/2016

 
Welcome to our inaugural blog entry – our first step in building a lasting relationship with small investors just like you.  As with any successful relationship we are adamant that our relationship with our community always be based on trust, respect and mutual success with the goal of not only providing investment strategies and forecast data to our users and customers but also educating small investors so they can actively invest with more confidence and independence than ever before.  As a result and before we begin, we want to introduce and emphasize our values and goals while outlining our next steps.

Core Values
Our work is based on three fundamental values that guide our daily work and interactions with all of our stakeholders both external and internal.  First and foremost we believe that trust is the absolute bedrock of any sustainable relationship whether it be professional or personal.  In turn and for us trust implies clear and transparent communication, delivering on our committments, and accepting full responsibility for our actions including openly recognizing and correcting any mistakes while ensuring that they never occur again.

Second we believe that social committment is essential for any modern enterprise.  Therefore we are committed to providing our user community not only commercial but also social value.  In particular we are committed to helping small investors decisions attain realistic financial returns with lower volatility over shorter timespans.  In doing so we are "democratizing" investment decisions such that small investors have access to the same information, which professional investors have long possessed.

Third we believe in mutual success for our entire community.  In other words we constantly seek "win-win" situations, in which all of our stakeholders – users and customers, investors, suppliers and colleagues – together profit from our work.  Only when we strive for mutual success among all of our stakeholders can we enable our own success and thereby ensure a sustainable long-term venture.

Longterm Goal
Our longterm goal is simple.  We intend to help small investors augment standard investment approaches with active approaches that yield benchmark returns or better with lower volatility over timespans ranging over years instead of decades.

Next Steps
Our first major step in realizing our longterm goal is the execution of our alpha and beta phases over the next few months.  During these two initial phases we will gradually introduce investment strategies and forecast data to the general public via this website and Twitter.  In addition we will begin live-testing our investment strategies using our forecaset data.  Overall at the end of these two phases in the summer of 2016, we aim to prove the legitimacy of our investment strategies and forecast data for small investors by proving that we can deliver benchmark or better returns with lower volatility over a comparatively short timespans.

And what is planned beyond the conclusion of our initial alpha/beta phases in the summer of 2016? Well we will concentrate on officially incorporating our company, consulting with investors, enlarging our user community, fine-tuning our investment strategies, and extending our forecast data.  However we are adamant that we first prove legitimacy.  Therefore and for the moment our focus is on our initial two phases over the next few months.

We intend to not only provide you upcoming information about our investment recipes and forecast data.  As a result please regularly check this blog for both educational information and updates.  In addtion we encourage all comments and feedback.  Therefore please do not be shy!  We very much look forward to hearing from you and begin developing lasting relationships with our user community.

    Categories

    All
    Algorithmic Trading
    DEGIRO
    DEZIRO
    Forecast Data
    Fund Industry
    General
    Investment Strategies
    No-Fee Brokerages
    Return
    Risk
    Robinhood
    Software Bug Notices
    Statistics

      Contact Us

      The personal data entered into this form are mandatory and will be subjected to computer processing. By clicking on the “Submit” button below, you acknowledge and agree that STATS4TRADE.com may store and process your personal data outside of the European Union. Your personal data are processed to allow STATS4TRADE.com to answer to your request. According to the French Data Protection Act, you have at any time a right to access, and oppose for legitimate reasons, the processing of your personal data as well as the right to update, correct or delete personal data collected by STATS4TRADE that would be incomplete, inaccurate, outdated data, or those forbidden by law, by sending your request via email at contact@stats4trade.com. For more information, please refer to our Legal Notices.
      The personal data entered into this form are mandatory and will be subjected to computer processing. By clicking on the “Submit” button below, you acknowledge and agree that STATS4TRADE.com may store and process your personal data outside of the European Union. Your personal data are processed to allow STATS4TRADE.com to answer to your request. According to the French Data Protection Act, you have at any time a right to access, and oppose for legitimate reasons, the processing of your personal data as well as the right to update, correct or delete personal data collected by STATS4TRADE that would be incomplete, inaccurate, outdated data, or those forbidden by law, by sending your request via email at contact@stats4trade.com. For more information, please refer to our Legal Notices.
      The personal data entered into this form are mandatory and will be subjected to computer processing. By clicking on the “Submit” button below, you acknowledge and agree that STATS4TRADE.com may store and process your personal data outside of the European Union. Your personal data are processed to allow STATS4TRADE.com to answer to your request. According to the French Data Protection Act, you have at any time a right to access, and oppose for legitimate reasons, the processing of your personal data as well as the right to update, correct or delete personal data collected by STATS4TRADE that would be incomplete, inaccurate, outdated data, or those forbidden by law, by sending your request via email at contact@stats4trade.com. For more information, please refer to our Legal Notices.
    Agree + Submit

    The financial data and interpretation published in the blogposts are provided without any representations or warranties, express or implied and shall not be regarded as a recommendation and/or an advice to invest from STATS4TRADE. For more information, please see the notices on our Legal page..

    Authors

    Jean-Marc Guillard and Mike Stafne are technical professionals and avid albeit cautious private investors – perhaps just like you! We enjoy providing our community tips, tricks and insights to help you interpret our public data. If you have any questions, just contact us anytime.
The content and data on this website are provided without any representations or warranties – either express or implied, and shall not be construed as financial recommendations and/or investment advice from STATS4TRADE or its partners. For additional information and important details please refer to our Legal page.

Copyright © 2017 STATS4TRADE.  All rights reserved.

  • Company
  • Application
  • Performance
    • Dow Jones 30
    • NASDAQ 100
    • Euro Stoxx 50
    • CAC 40
    • DAX 30
  • Media
  • Blog
  • Contact