Tools being used by your trading competition

[howardbandy]

... developing AGI, artificial general intelligence. Interesting times....especially to see how it develops in the trading world.

AGI == the singularity. Some believe it will happen in the very near future, others a few years in the future. There is pretty high agreement that it will be achieved by the end of the 21st century.

The articles describing the increase in use of AI in trading were the motivation for beginning this thread.

Best, Howard

 

[Valued]

That's pretty amazing if that's legit. I want to see a Libratus vs Libratus! Maybe the world will cave in on itself?

The way this stuff is advancing poses some questions about our future, they're talking there about how it's going to be used in fields like cybersecurity, medical diagnosis, in a couple years it looks like we won't be driving anymore(no delivery/truck drivers eventually?), factory jobs are out because of robots. I gather one day AI will be writing the actual code themselves too? What are we all going to be doing in the future? In what areas do humans have an advantage over AI? Creativity? On the spot reactions I guess, although in that article they're also talking about developing AGI, artificial general intelligence. Interesting times....especially to see how it develops in the trading world.

The poker pros they have that computer playing arn't the best headsup players in the world though. The three best players in the world right now at headsup no limit holdem specifically (the game the computer is playing) are likely Doug Polk, Ike Haxton and Ben Sulsky and maybe Daniel Cates fits in the top 3 it's a bit hard to tell. I know Doug Polk played a computer for 8 hours and won. We arn't really there yet. The computers have a hard time too with 6 handed and 9 handed games because of how much more complex they are. The edge humans have is they adapt faster to changes in the rules. In poker if you change the rules it invalidates everything the AI has learned up to that point. For example, Omaha is just like texas holdem but you draw four cards and not two and this changes the game so much that a new bot has to be built to play Omaha. You can also change the blind structure etc.

Maybe the same could be said with trading. If traditional correlations in fundamental data or even just between two different asset classes break down because humans change the rules (new inventions making some commodities obsolete or large structural changes or even changes in laws and how markets work) then a computer that learns based on the fundamentals may not be able to adapt very fast.

 

[howardbandy]

Maybe the same could be said with trading. If traditional correlations in fundamental data or even just between two different asset classes break down because humans change the rules (new inventions making some commodities obsolete or large structural changes or even changes in laws and how markets work) then a computer that learns based on the fundamentals may not be able to adapt very fast.

The primary use of computers -- where they have an advantage over humans -- is trading rather than investing. By trading I mean frequent changes in positions based on recent price history with short holding periods; and by investing I mean infrequent changes in position based on use of fundamental information and long holding periods.

It is my opinion that market efficiency limits the use of fundamental information and that it is of near-zero value to individuals other than those with a seat in the boardroom and their best friends. I have published my thoughts on use of fundamental information several times -- most recently here:
http://www.blueowlpress.com/wp-content/uploads/2016/10/FT-Fundamental-Analysis-Appendix-A.pdf

In trading, automated techniques such as model design, signal generation, trade selection, timing, order management, and risk management are superior to those of human traders. It is those aspects where I recommend that humans who want to compete do so by broadening their skills -- as the video referenced in post number 1 of this thread suggests many large trading firms are already doing.

Thanks for listening, Howard

 

[tech/a]

Howard

Thank you for sharing your view.
Anyone having a good read and using nothing more than common sense.
can see that fundamental analysis has as many flaws as any other form of analysis.

To the point as you say of being meaning less unless you are on the inside looking out.

Happy to stick with price action.

 

[Klogg]

It is my opinion that market efficiency limits the use of fundamental information and that it is of near-zero value to individuals other than those with a seat in the boardroom and their best friends.

It limits it yes, but stating it is of 'near-zero value' is very mis-leading.

I'm not a professional and this 'near-zero' value has lead to returns of more than double any Australian index of your choice. I have no friends with a seat in any boardroom, nor do I have any.

I accept that there are markets that are very efficient, but blanket statements like that are very misleading.

Even taking the first statement of your publication:
"In order to be valuable, any data series or indicator, company data or economic series, must be:
• Timely
• Accurate
• Predictive"

This is massively wrong. I don't need very accurate data to make money from an investment. As Buffett says, "you don't need to know an man's exact weight to know he's obese".
And backing that up is Keynes' famous quote "it is better to be roughly right than precisely wrong".

Likewise, it doesn't need to be 'timely' for it to be useful.

Perhaps you meant that the fundamental data is of near-zero value in your particular type of analysis. That I can agree with.

EDIT: I should mention that the idea that markets are completely efficient has clearly been killed. It works in theory and creates nice mathematical equations for all, but in practice it clearly doesn't exist.

 

[Valued]

@howard by your definition are swing traders who hold positions for a few weeks at a time traders or investors? What about someone who buys USD for the intention of holding it for the next 3 - 6 months because they think the US will rise interest rates? Are they traders or investors? What is infrequently?

I don't think it follows that if markets are efficient fundamental data is useless. Efficient markets should reflect the underlying fundamentals. If the price of oil is rising this should have a follow on effect eventually. You should see changes in the Baltic Dry Index (indicating shipping goods by sea is becoming more expensive) and then you should see a gradual rise in other commodities and from there inflation and potential interest rate rises. If you get on board when this starts and put your money in commodities then because you know the leading fundamental indicators that predict commodity price rises an efficient market will make you a lot of money. To make a lot of money you have to be with the trend and with everyone else anyway.

I am not convinced markets are that efficient in the short term but even if they are there is nothing magical about this that prevents you making money knowing the fundamentals support your position and everyone else agrees and you ride the trend upwards for 6 months or whatever it may be.

@Tech for break out trading price action is most important but fundamentals are vital in other asset classes. Commodities without fundamentals is going in blind.

 

[howardbandy]

It limits it yes, but stating it is of 'near-zero value' is very mis-leading.

Hi Klogg --

I stand firmly with the material I published in that article.

Between delays, revisions, granularity of data, unknowable biases, and general agreement that the efficient market hypothesis is not usable, what remains is broad market timing and subject sector selection. I do not see how the scientific method can help.

If you have found techniques that work for you, ignore me and continue as you were.

What happens in market downturns will sort us all out.

Best regards, Howard

 

[Klogg]

I stand firmly with the material I published in that article.

I didn't expect you to change your stance from my one post. You've committed so much to publishing that piece (a few hours as a minimum), so I would expect some resistance before writing that off as a lost cause - and that's before considering the fact that you've opened it up to the public, so you'd want to save face and remain consistent (we all do this to a degree).
My post was to mitigate the impact on others who are new and to create discussion.

Another problem with this stance is it completely ignores the success of firms like Berkshire Hathaway, Oaktree Capital, Markel (I can go on and on...), as well as many who post here.

In my view, it's really a form of cognitive dissonance - you can see evidence of the EMH not holding true, yet you use EMH to form a view on fundamental analysis. It confuses me a little.
Perhaps you know a way in which these investors (listed above) are successful whilst EMH holds true - if that's so, please do tell me. I'm genuinely interested.

 

[Valued]

I don't understand why the EMH applying means fundamental analysis is any less valid. I disagree with the EMH but assume markets are efficient, say CBA shares are trading at 60 dollars and the market is efficient so they are actually worth 60 dollars, if you do some fundamental analysis of that stock and come to the view that CBA is a good company, it will keep growing, it has a competitive edge in the market, it is coming out with exciting products, and you want to invest and get some dividends and watch your capital grow with the company, and then over time the company does grow and becomes more profitable and dividends increase and this raises the share price, you have made money. I would say an efficient market rewards good stock selection since to make money you need to invest in a good company and in fact an efficient market means you're not overpaying too and you're likely to get the benefits of the growth.

Richard Farleigh, an Australian investor, believes markets are much more efficient than people give them credit for but also notes you can just ride a trend as an investor and make money. If then the conclusion is that fundamental analysis is only valid for investments or longer term trading then obviously that is true. However, that has nothing to do with whether markets are efficient or not.

So how does market efficiency impact on the validity of using fundamental analysis? It does not. Presumably, we want to use fundamental analysis to pick the best performing stocks in the best performing sectors who have competitive edges and who are going to grow and generate more profits and the shareholders then benefit from that.

 

[Klogg]

So how does market efficiency impact on the validity of using fundamental analysis? It does not. Presumably, we want to use fundamental analysis to pick the best performing stocks in the best performing sectors who have competitive edges and who are going to grow and generate more profits and the shareholders then benefit from that.

Perhaps you've misunderstood EMH. From investopedia:

The efficient market hypothesis (EMH) is an investment theory that states it is impossible to "beat the market" because stock market efficiency causes existing share prices to always incorporate and reflect all relevant information.

EMH http://www.investopedia.com/terms/e/efficientmarkethypothesis.asp#ixzz4WvdMHJgY

The bolded section says it all. Hence, if people are beating the market, it cannot be 100% efficient.


Yes, some markets are more efficient than others, but that doesn't mean they reflect ALL available information (that is, not all markets are 100% efficient, as EMH suggests) - hence, fundamental analysis should (and does) work.

 

[Triathlete]

Yes, some markets are more efficient than others, but that doesn't mean they reflect ALL available information (that is, not all markets are 100% efficient, as EMH suggests) - hence, fundamental analysis should (and does) work.

I have no doubt that Fundamental analysis works although in my case I will also use Technical analysis for my entries and exits.

One of the issues with Fundamentals is how does an investor stay on top of it since the information is usually in a quarterly,half yearly ,yearly report or a special announcement.

The reason I bring this up is what if in between these reports the company starts to experience difficulties and no information is forthcoming early from the company or is delayed in its announcements investors are left holding the bag.

I do use Fundamentals to choose financially strong companies but use Technicalls for entries and exits.

My belief is that a technical person may have and advantage because they may start to see something changing on their charts that alerts them to something that does not seem right and so will take steps to protect their positions.

I mean there could be times that information has gotten out to big players that has not got to the general public and the only way we can see this early is on a chart and not in a company report that are months apart.

The last two companies that come to mind are SGH and now Bellamys.

Does anyone have any comments on how just being a Fundamental investor would have helped these investors..?????

 

[Valued]

Perhaps you've misunderstood EMH. From investopedia:

The efficient market hypothesis (EMH) is an investment theory that states it is impossible to "beat the market" because stock market efficiency causes existing share prices to always incorporate and reflect all relevant information.

EMH http://www.investopedia.com/terms/e/efficientmarkethypothesis.asp#ixzz4WvdMHJgY

The bolded section says it all. Hence, if people are beating the market, it cannot be 100% efficient.


Yes, some markets are more efficient than others, but that doesn't mean they reflect ALL available information (that is, not all markets are 100% efficient, as EMH suggests) - hence, fundamental analysis should (and does) work.

I don't think it follows that if you beat the market it invalidates the EMH. Again, I don't think the EMH is right but the fact that someone beats the market (which I assume means the same as beating the index) does not in itself invalidate the EMH. The market is an average. Not every stock grows at the same rate as the index. Some stocks out perform the index and some stocks under perform the index. Some businesses are just better than others. The fact that an efficient market reflects all available information does not mean you can't beat the index.

 

[Klogg]

The fact that an efficient market reflects all available information does not mean you can't beat the index.

My take on it is that if it were efficient, every stock would give the same return.
Let's agree to disagree, as I've derailed the thread more than enough... my apologies

 

[Valued]

Does anyone have any comments on how just being a Fundamental investor would have helped these investors..?????

For SGH I believe Warren Buffet for a long time warned to be very cautious of a company who grows by acquisitions since acquisitions may on paper grow the company but in reality may negatively affect its value. Growing by acquiring is exactly what SGH is doing. I actually chose not to invest because of how they kept acquiring other firms, especially firms overseas that operated in a legal jurisdiction that most SGH management would have no experience in. I also knew a lot of the PI claims they were acquiring were very likely to be rubbish but that's because I have inside knowledge of the industry. That same knowledge to acquire for a retail trader would likely cost about 15k - 30k or so which isn't really realistic for most investors simply conducting some due diligence.

 

[Valued]

My take on it is that if it were efficient, every stock would give the same return.
Let's agree to disagree, as I've derailed the thread more than enough... my apologies

That makes no sense.

 

[skc]

Perhaps you've misunderstood EMH. From investopedia:

The efficient market hypothesis (EMH) is an investment theory that states it is impossible to "beat the market" because stock market efficiency causes existing share prices to always incorporate and reflect all relevant information.

EMH http://www.investopedia.com/terms/e/efficientmarkethypothesis.asp#ixzz4WvdMHJgY

I can never quite understand the EMH. It's not like there's actually an independent being called "the market". The market is the collective result of traders / investors / computers etc etc.

If one player has correctly discounted all the relevant information (and is hence being efficient), then by definition the counterparty (there's always one) who took the opposite side of the trade has been inefficient. So clearly... some market players are more efficient than others.

So to say that market is efficient so no player should try seems a bit irrational.

One of the issues with Fundamentals is how does an investor stay on top of it since the information is usually in a quarterly,half yearly ,yearly report or a special announcement.

I think one needs to define fundamental analysis as being much broader than just company financials and announcements. To me it includes all information other than price action. There are lots of information out there that impacts a company and are released periodically, or can be researched/collected by those who has the willingness and the ability to do so. For example, data like retail sales, housing starts, webpage impressions, news by peers in other industries, regulatory announcements or even just social media and regular news (look at TNE on Wednesday). Some of these are easily accessible by all market participants.

Then there are others that are harder to access for the general public. For example, the hedge fund who shorted BAL used supermarket scan data (which revealed supermarket sales were falling way before any announcement was made) as part of their trade rationale. Analysis of satellite images is also a good example, anything from new cars arriving/leaving the port, number of flares in a gas field, green-ness of certain crops etc can inform how commodities or companies are travelling.

This is a field where big data/machine learning could be very useful... e.g. what's the correlation between temperature and coke sale, or clothing sale, or salmon prices? Or between rainfall and almond production or wine glut? Nonetheless, I think this will most likely provide a temporary edge until everyone else copies it, especially if it proves profitable. Part of the reason for Renaissance Technologies enduring success is probably in no small part due to the secrecy behind its quantitative methods...

Now there's no doubt that, those who possess the fundamental information edge (note they are not necessarily insiders) will make their move before any price action tells them to do so. Indeed, their actions are probably what created the price actions in the first place, followed by the technical traders. That's why, in the absence of anyone being absolutely sure that they have an information edge (which probably applies to most retail traders), price action should to be respected.

Another interesting aspect worth mentioning relates to how price actions in one market feeds into another. For example, price action in US bond yields' feeds into the performance of yield stocks on the ASX. I am not sure whether that should be categorised as technical or fundamental information... but it is the collection of all these information that forms the "context" for an astute trader/investor. I do believe that a machine may be able to paint the context picture in a more quantitative manner than a trader, in a steady-state kind of situation. Though I remain to be convinced if the machine has a definite edge when the context changed.

 

[Triathlete]

I also knew a lot of the PI claims they were acquiring were very likely to be rubbish but that's because I have inside knowledge of the industry. That same knowledge to acquire for a retail trader would likely cost about 15k - 30k or so which isn't really realistic for most investors simply conducting some due diligence.

Thanks for your comments Valued........Your comment is exactly my point,the only way I stand a chance of not getting caught is following my Fundamental information with a chart since the information is instant and I would not have inside information.


Now there's no doubt that, those who possess the fundamental information edge (note they are not necessarily insiders) will make their move before any price action tells them to do so. Indeed, their actions are probably what created the price actions in the first place, followed by the technical traders. That's why, in the absence of anyone being absolutely sure that they have an information edge (which probably applies to most retail traders), price action should to be respected.

[/QUOTE]

SKC......I think your last comment is also what I mean and if your latest information is saying one thing and the price action is saying something different than you really need to take notice.........

 

[howardbandy]

Greetings --

Please understand that my perspective is development and use of trading systems that are based on the scientific method. That is: observe data, postulate rules, validate the rules using data not previously used. The metric I use is risk-normalized profit potential -- CAR25.

The points I am making regarding fundamental analysis are that no information usable in creating scientifically-based trading models that have reasonable risk-normalized profit potential can pass through the filters of granularity, revision, and bias.

If a person or organization feels something in the fundamental data is helpful and increases their confidence in taking a position in the issue, they should continue to do that. Similarly for Gann, Fibonacci, divergences, crossover of the 50 and 200 day moving averages, etc. I have found no predictive value in any of those, but many people use them and defend their use. I have no argument against doing that.

Generation of trades is separate from system health and position size. That is, the system that generates trading signals is separate from the system that determines position size. Any set of trades, no matter how they are generated, can be used in the trading management system. It is there that risk-normalized profit potential is measured.

I have found no way to use fundamental data to form rules that generate trades that have reasonable risk-normalized profit potential. That is, the CAR25 value of every set of trades that I can generate using a set of rules derived from fundamental information is too low to be worth trading.

Prove me wrong.

Post a system that uses fundamental data from, say 2000 through 2010 (or a period of your choice that you have found to be stationary), to form a discoverable set of rules to buy and sell some tradable issue. Test those rules on previously unused data that follows the data mined to discover the rules. Define a statement of risk tolerance and analyse the out-of-sample trades. Determine the profit potential. If you prefer a different metric, state it and defend it as preferable to CAR25. Post a report describing the whole experiment.

-----------------------

Regarding Warren Buffett and Berkshire Hathaway. From my perspective, Mr. Buffett is the CEO of a multi-faceted conglomerate corporation. BH buys very large stakes in companies, expecting to hold them a very long time. They influence the management of those companies. That is a corporate view, rather than a trader's view.

My personal risk tolerance is to hold risk of drawdown to a small chance of a drawdown greater than 20%. Many money managers view even that as too risky. When BH held through drawdowns in excess of 50% in 2009, that was an indication to me that they were managing corporate subsidiaries, not trades.

My comparison would include Mr. Buffett in the same group as Jack Welsh (General Electric) and Ken Lay (Enron) -- CEOs of conglomerate corporations. A group of CEOs of trading organizations would include David Shaw and James Simons.

Best regards, Howard

 

[Triathlete]

Greetings --

Please understand that my perspective is development and use of trading systems that are based on the scientific method. That is: observe data, postulate rules, validate the rules using data not previously used. The metric I use is risk-normalized profit potential -- CAR25.

The points I am making regarding fundamental analysis are that no information usable in creating scientifically-based trading models that have reasonable risk-normalized profit potential can pass through the filters of granularity, revision, and bias.

If a person or organization feels something in the fundamental data is helpful and increases their confidence in taking a position in the issue, they should continue to do that. Similarly for Gann, Fibonacci, divergences, crossover of the 50 and 200 day moving averages, etc. I have found no predictive value in any of those, but many people use them and defend their use. I have no argument against doing that.

Generation of trades is separate from system health and position size. That is, the system that generates trading signals is separate from the system that determines position size. Any set of trades, no matter how they are generated, can be used in the trading management system. It is there that risk-normalized profit potential is measured.

I have found no way to use fundamental data to form rules that generate trades that have reasonable risk-normalized profit potential. That is, the CAR25 value of every set of trades that I can generate using a set of rules derived from fundamental information is too low to be worth trading.

Prove me wrong.

Post a system that uses fundamental data from, say 2000 through 2010 (or a period of your choice that you have found to be stationary), to form a discoverable set of rules to buy and sell some tradable issue. Test those rules on previously unused data that follows the data mined to discover the rules. Define a statement of risk tolerance and analyse the out-of-sample trades. Determine the profit potential. If you prefer a different metric, state it and defend it as preferable to CAR25. Post a report describing the whole experiment.

-----------------------

Regarding Warren Buffett and Berkshire Hathaway. From my perspective, Mr. Buffett is the CEO of a multi-faceted conglomerate corporation. BH buys very large stakes in companies, expecting to hold them a very long time. They influence the management of those companies. That is a corporate view, rather than a trader's view.

My personal risk tolerance is to hold risk of drawdown to a small chance of a drawdown greater than 20%. Many money managers view even that as too risky. When BH held through drawdowns in excess of 50% in 2009, that was an indication to me that they were managing corporate subsidiaries, not trades.

My comparison would include Mr. Buffett in the same group as Jack Welsh (General Electric) and Ken Lay (Enron) -- CEOs of conglomerate corporations. A group of CEOs of trading organizations would include David Shaw and James Simons.

Best regards, Howard

Howard with your systems is there a specific group of companies that you concentrate on eg top 50 or 300 to generate your trades or do you also use maybe speculative companies that are included as well...?????

 

[howardbandy]

Howard with your systems is there a specific group of companies that you concentrate on eg top 50 or 300 to generate your trades or do you also use maybe speculative companies that are included as well...?????

Greetings --

Company selection is a system in itself.

I recommend Not applying a set of trading rules to 3000 tickers, finding that results for 30 of them seem to be good, and trading those 30. That is "optimizing / curve fitting the symbol space" and is a poor idea.

Rather, begin by looking for issues that have the potential of being tradable -- adequate profit with acceptable risk. There is a technique described in both the "Foundations" book and the "Quantitative Technical Analysis" book called the "The Prospector." It is possible to determine the risk of trading any price series even before a set of rules to generate buy and sell signals is applied. We want several characteristics:
1. Enough volatility so there is profit available by trading.
2. Not so much volatility that whatever model is applied results in too much risk.
3. High liquidity (and low bid-ask spread).
4. Given that the issue passes the first three filters, then see if there is a trading model that can detect persistent patterns that precede profitable trades.

Use the accuracy / holding period charts to help focus on high CAR25. The sweet spot is high accuracy and short holding period. Again, description of the technique is in the chapter on risk in both books.

The list of issues that pass is fairly short. The major sector exchange traded funds (XLB, XLE, XLF, ...) work, as do a few common stocks. With the stocks, be careful to distinguish between a rising market and persistent patterns. If you have the "Foundations" book, discussion of "The Prospector" begins on page 73 in Chapter 2 -- Risk. In "QTA," it begins of page 124.

I recommend working with a single issue long / flat to begin. If / when that works, extend to long / flat / short, and / or add additional systems, each a model and a single data stream. The Bayesian sequential learning that is the basis for Dynamic Position Sizing ("QTA" beginning on page 385) works to switch between alternative systems. Since CAR25 is a Dominant metric, the highest risk-normalized return comes from trading the best and switching as conditions change.

Best regards, Howard