Intro This is a personal experiment into the “forum” experience, induced by the “stay at home” policy. And as an experiment in the experiment, I will also post in one or more other forums (or fora, if you are a Latinist.) I do like experiments! I am so clueless regarding forum rules that I am not even sure this is the right section. I trust the forum police to re-direct this as they see fit. I know a similar topic has/is being dealt with in other threads of the forum. From a very quick research (daunting task!) I have found some outstanding contributions. This is my personal one, less than 2 cent, but I guess it’s still ok, as value is not exactly what people seems to be looking for these days! Where do I stand in the debate “algo” vs mechanical vs discretionary trading? Who gives a toss, many will reply, and rightly so. For all the others my answer is…drum roll… I am not quite sure yet. This may not add much to the topic, per se, but I think that the reasons for this answer are more interesting than the answer itself. I have long been fascinated by this debate. I have been reading tons of books, posts, reports for years and there were interesting lessons to be learnt along the way. I am not a programmer, nor a machine learning guru. Worse: I have a financial markets background and even an MBA, so you can imagine how much my inflated ego costed me at the beginning of my independent trader journey! I have been a trader for 15+ years and, as such, I have done all the stupid things one does. These could be broadly grouped into two categories: to think that we possess (or can easily acquire) innate and accurate predictive skills, and to think we are immune from our human brain’s psychological flaws. I have had plenty of evidence of these as a trader, a trading coach/educator (yes, I am into that too…long story) and an observer. To address the former (alleged superior predictive skills) I learnt that I had to objectively establish if I really had them or it was just wishful thinking. As we will see, the only way to do so is to identify the components of these “skills”, isolate them and measure them as objectively as possible. To address the latter (psychological flaws) I have tried a bit of everything; reading Mark Douglas’ and other trading psychology gurus’ books a thousand times, reading NLP, trying meditation, employing performance coaches, drinking matcha tea…. you name it. They all helped, but by nature I am rarely satisfied. The most obvious way to address both was for me to try a system with mechanical rules. This would make the rules measurable and would get rid of the human biases and weaknesses altogether. Neat and clear. As we will see, that is just part of the story. This is a vast topic and probably unbearably boring for most. I am trying to make it even more boring by sounding verbose and pretentious (well, we all have our own talents!) In doing so, I hope to dissuade the vast majority that is looking for excitement, magic signals and shortcut to immense wealth. This thread will probably be more indicated as a week-end read. One nice by-product is that it could be a very efficient cure for insomnia.
Part 1 Definition and characteristics So, some clarification. I would keep the trading side separate from the investing side (another long story.) I should also add that I think there is a significant division between intra – day and daily systems (ditto, perhaps another thread….) I will unashamedly simplify here, apologies to the purists and the Illuminati. Discretional, mechanical and automated systems. I believe that every trader uses some form of system or methodology. The less conscious and less strict the rules, or, put it the other way, the higher the degree of human intervention, the more I call it discretionary. The lower, the more I call it mechanical. When all the components of a system are strictly mechanical and eligible to be coded into an algorithm, I call it an automated system. Spoiler alert: a highly discretionary approach cannot be automated, at least in my experience. The mechanical (automated or not) approach departs from what is considered trading in the conventional sense (not necessarily a bad thing.) One fundamental difference is that in a totally mechanical or automated system a trader does not pick and choose, does not apply minute by minute (hour by hour, whatever) analysis. A trader (or the machine) takes the signals based on the fact that doing so produced on average nice returns in the past. Signals are not questioned. In the middle between discretional and automated, there are the hybrid solutions. If I use a mechanical system, but decide each time which signal to act upon, I am using to various degrees my own discretion. I call these semi-mechanical systems. They can also be semi automated. Having lived in all these camps for many years, I would say that it is hard to declare a clear winner. It depends on the individual goals or preferences. Each approach addresses different problems and creates more issues trying to solve them. Edge A discretionary system is based on the idea that we have superior predictive skills and these give us an edge on the markets. It also assumes that we apply these skills flawlessly. There is a big industry out there that thrives on the idea that these skills are easily achievable with a few books, one course or “the right secret”. I won’t comment on this, but I find the assumption of superior predictive skills quite presumptuous and statistically untenable (there are exceptions of course.) A mechanical system tries to derive its edge from using highly measurable, fixed components (rules) and testing them thoroughly so that they can show a meaningful statistical significance. It derives its edge from statistics. This comes with its own huge problems. It makes it dependant on the quality and objectivity of data it is based on, with the big issues of data collection, data quality, overfitting, etc. It also assumes that the future will pretty much resemble the past. An automated system applies these rules without any sweat and any further input (especially biased humans’ input.) The main advantages of an automated system are: a) Trading becomes a number’s game (wasn’t it always?) Everything is objectively measurable and measured, allowing for statistical analysis (the second best predicting tool after the proverbial coin) and consistent application in all circumstances. b) It frees traders from having to sit at their desks for hours. This is an issue highly underestimated by all traders. c) It avoids human biases and weaknesses (prospect theory, anyone?) d) It allows to use a wide number of markets. This is particularly useful both because we need massive data and because, screening by correlation, we can add diversification, which is a fundamental bonus in my opinion (although it is much more rated in investing than trading.) Obviously there is a flip side: a) It is no “fun”: many people feel they can’t call themselves traders if trades aren’t generated by their superior analysis. Knowing the intricacies of binary languages or simply “buying when it’s green and sell when it’s red” is not everyone’s cup of tea. Looking at charts is more entertaining. The exhilaration of calling the right shots is unmatchable. Commenting on where the Swiss Franc or the S&P is going to be in the next week or month wins hands down the “charisma” game at parties and dinners (remember those?) b) It is rather expensive, in terms of time, study, software, education, equipment, frustration, etc. and one can’t just jump in after reading a book and following a blog by that guy with the Lamborghini (this is a rather good thing though!) c) A more serious issue is that a mechanical/automated system is a “one size fits all” thing. It is usually a pretty blunt instrument. It is not flexible. It is based on a strong assumption: that the future will be like not just the past, but like the past that we observed. This is not always the case. If you already know something about markets, an automated system is far from “perfect”. When I check every single trade the system does, I often find that it entered or exited at the “wrong” place; it ignored very clear support/resistance, trend lines, of Fib levels or whatever one fancies. True, the system will do this consistently, without hesitation. That would imply that if the statistics are still good, our analysis/input/filtering is quite irrelevant. But this is also quite hard to accept. We may see (or think) that a more thorough analysis and filtering of the signals would lead to better results and it is difficult to resist greed or perfectionism. d) There are so many ideas and tweaks to be used in trading that it is very hard to choose. We are spoiled for choice and always agonise that there may be something better. It is easy to jump from idea to idea and system to system, and even from parameter to parameter in the same system. It may also be also an easy short cut to avoid hard work, usually with fatal consequences. Well, if you are still reading, you must be somehow interested, so I will get a bit more personal on my next post.
Part 2 My journey Taking the plunge into automated systems was the start of a long journey. Soon I discovered that simply buying off the shelf automated systems would not work, for a number of reasons we may discuss in other posts. I had to develop my own systems based on some ideas that I would like and accept. Not being a programmer, I was forced to use very simple ideas and techniques, which is a good thing, IMO, and test, test, test. Crucially, I had to do all the testing manually, on excel sheets. While my mates were spending the nights playing Call of Duty (or otherwise), I was pressing keys on my keyboard to move the charts left one bar at the time, all the time thinking that I was wasting my life (a view wholeheartedly shared by my wife, BTW). Surely, had I spent all that time learning Python or C language (or setting up a plumbing business, for that matter!) I would probably be in a better place now, but this was a long time ago and Python at that time was mainly associated to a large snake. This exercise, however, was a fundamental step forward for me and a big lesson. It was vital for learning a great deal about myself, about trading ideas I like and disliked, about market behaviour, about objectivity, about systems. My 10,000 hours to become an “expert”, if you like. And it was, as always, a mixed blessing. On one hand, I could come up with clearer ideas of what kind of trader I was, what I liked, and, occasionally, with something that worked. On the other hand, I could see how what I had learnt from Price Action, from the rhythms of the markets, from all the things we call experience, would regularly beat my original simplified mechanical system. This induced a sense of frustration. Besides, I had not produced a system I would feel comfortable to switch on and let alone without supervision. So, if I had to “be there”, I thought I may as well be there in a more pro-active way and weed out “bad” signals. The bad part was that so many nuances of my “human” analysis were impossible to code (at least for me, but, as I soon learnt, by many “professional programmers”.) So, I would have to apply them “manually”. This has huge consequences (time, biases, mistakes, etc.) and I knew perfectly well that all this might have looked great on paper, but the moment I would have gone live and “overruled” the basic systems with my “superior” analysis, I would fall prey of my human weaknesses.
Part 3 More lessons So, catch 22. My next move was to see how much I could simplify, categorising my human analysis according to “degrees of programmability”, which is the inverse of “degrees of discretionary trading”. For instance, if one rule was: “do not take signals around Pivot Points”, it would be very easy to define what Pivot points are and relatively easy to define what “around” meant: x%? y points? These filters would be quite easy to code. Discretion: basically zero. Others would not be so easy: how do you define a Break-out, or a support/resistance level in a way that makes sense and at the same time a machine can understand? How do you code some more advanced price action information? As I discovered, you basically can’t, at least with the resources I could afford. Trading, like life, is a constant compromise. So I ended up trading mainly semi mechanical, semi-automated systems, where the level of discretion is as low as possible. Rules are fixed and mandatory. Some rule-overriding filters are fixed and less mandatory, some other filters are more discretionary. My discretionary trading part in some systems is well below 10%. Please be aware that 10% discretion is still big enough for your psychological biases to take over! Still, you have to find a compromise that works for you, otherwise you will never be able to follow it. So, this is sub-optimal but much better than gut feeling or shooting blind. The discretionary part, however small, is still an issue though: it is dangerous (it’s a potential trap for disruptive human biases) and it is time consuming (you still need to sit in front of your screens endlessly monitoring charts.) This in turns means that you can’t diversify trading more systems or many instruments. Diversification is my mantra in investing and I do not entirely accept that in trading it is not. It also means that it is highly unlikely that you will replicate the results you would have theoretically achieved with the system (this is another massive topic: some other time.) This second step taught me another vital lesson: the importance of measurement. I do collect and manage my data through journaling religiously all I want (or can) measure. Using a system forces us to define very precisely what our rules are, so they can be measured. If we can’t measure something, we have no meaningful information about how that something is contributing to our bottom line. That means that the less rule-based our trading is, i.e. the more discretionary, the more uncertainty we have on how it has really performed in the past (and why) and how it will in the future. Not only that: if we can’t objectively measure it, we will use your own memory to decide if something has worked or not. Our memory is a really bad adviser: we will remember the things we want to remember (confirmation bias, endowment bias and the lot.) So, the more material we monitor, the more information we have, the better we and our system will become. If I have learnt anything about this experience, it is the importance of keeping a proper journal. For every system I use, I journal all the trades a system would signal and measure every filter’s behaviour. I write down all the things I want to measure, from time of the day to different targets achieved according to different trade management rules, etc.. I have appreciated how vital this is not only for myself, but also when I coached traders. When I could show, in black and white, with numbers, what impact a particular behaviour (ignoring rules, creating exceptions, getting out too early, procrastinating because one is “not sure enough”, etc.) has on the results, the awareness of someone’s issues raises dramatically. It makes it easier to accept that these issues are real and it’s the first step to address them. This is usually a big game changer.
Part 4 …and?.... Well, that’s it. That’s what I wanted to say. Ok, I guess the next question is whether I have created any system (automated or semi automated) that works. This is where I should add a picture of a young me sitting in my convertible Ferrari drinking champagne surrounded by girls in bikini and a link to my website where I teach you how to do it for $9.99. Sorry to disappoint you, I won’t. Seriously, the answer is complex and it provides another big lesson. Yes, I have created some systems that seem to work, for quite a long time now, but that does not mean much. I may have found something that works reasonably well for me, my ideas, my risk tolerance, my idiosyncrasies. What I find acceptable may look totally wrong for others. Do not underestimate this. At the beginning of my journey “teaching” traders, I thought it would be pretty much teaching a system and basically cloning myself. I soon learnt that the personal, psychological framework of each individual is very much a fixed factor virtually impossible to change. You can “teach” about a break-out, trend following system to somebody who is by nature a contrarian, but he/she will have a terrible experience trying to follow it, as it would not fit his/her personality and beliefs. If you don’t believe me, here is another example. One of my best systems has by far the best performance if used in conjunction with a 2:1 Risk/Reward ratio (yes, you risk 1 to get half: how dumb is that?) The numbers over thousands of trades do not lie, but this is so hard to accept that virtually all the traders I know have not been able to stomach it. And that’s OK. You need to find what works for you. I am happy to report that all this gargantuan work has finally paid off. Whether it was worth it or not is a subject of a much more personal assessment, but I have been trading some of my systems for years, my main source of income comes from trading and investing, and I am constantly researching. Now, the obligatory part where I say I am not here to sell you any system or to defend the merits of this or that methodology. I have seen many promising threads (I am not claiming this is one!) destroyed by quarrelling and bickering nonsense. And please do not start asking for equity curves and broker’s statements. Having said that, and for transparency, if some people, despite the boredom and lack of appeal of the subject really insist to pay me a substantial sum to have a piece of my “extensive knowledge”, or would like to invite me for a lecture in an exotic place (after the lockdown of course), or would like to show a material appreciation for helping them curing their insomnia, I will be happy to put them in touch with my accountants. After all, we are all in here for the money, aren’t we? Also, please do not take this as an invite to submit your or others’ great systems. There are billions out there. This is not the place. I am interested in the discussion, the experience and (who knows?) development….(I would really do with that exotic place…) Well, just scraping the surface of a complex subject; make what you wish of this. Thank you and congratulations if you have managed to read so far.
Respect to you for putting so much thinking into this and your wife for holding on For me, moving into mechanical strategies eventually lead to using more simple stuff and repeating it more often. Lower expectancy there is typical and compensated by frequency. An attitude change is required as lots of things such systems will pick will not look or feel "right". It really becomes thinking about next 1000 trades outcomes and ranges of distribution. Another attitude thing with mechanical system is letting them be, especially with larger accounts. Seeing +/-20k$ daily is hard. But it is a statistical noise for my size. I learned not to look. My journal - Fully Automated Stocks Trading. Val
I read your full post in length. Interesting stuff - everyone has a story of their journey and some are more exciting than others. I don't have anything else to add. I just wanted to post that I saw your thread at forexfactory too: https://www.forexfactory.com/thread...vs-mechanical-vs-discretionary-trading-debate
%% Good points. I think most that profit with automated systems; those automated traders have pretty good discretion . No telling how many discretionary traders/investors use plans + sometimes automated sell or buy orders.......... Good reads.
Automated trading probably have higher average but lower spread/variance in results relative to discretionary trading (not talking about chart reading day traders). In discretionary trading, you can create and execute trade plans with outsized profits potential (talking about hundred to few thousand %) like John Paulson, Michael Burry type.