Ah, what a fascinating topic! Can’t help dipping in.
I am sure many of you will know my position on the subject
. For the very few still in the dark it goes a bit like this:
Mechanical (algo) good, but usually not great, as many nuances of Price movement are too difficult to code. (I can see the storm coming on this!
)
I won’t go into the technicalities (the perils of optimisations, the dark matter of back testing, the demo vs. real, etc. etc. etc.)
I would like to comment on general philosophy, if I can. Trivialising a bit: on one hand you have the coders and on the other the traders.
The coders (usually a rather arrogant bunch, in my experience
) think that everything can be coded, including my dog’s farts, provided they have the “right input”, which is usually a mathematical formula. If I do not give them the exact rules for my dog’s farting, it’s my fault and my problem. Fair enough.
The traders (usually a rather arrogant bunch, in my experience) think that they are applying an obvious set of "rules" that makes a lot of sense to them. They can “tell” when their dog is about to fart. They can see it from the position of the ears, or from that guilty look…whatever. And can’t understand why the coders “do not get them” and instead go about asking for formulas for dog’s farting! How weird!
My opinion is that they are talking about different ideas of what trading is about.
One group is acting on data and mathematical formulas. This has many advantages. It gives a scientific rigour to the results, as these can be properly analysed via statistical methods and give more confidence on consistency, provided they know what they are doing, and that one assumes that the future will mimic the past (a very big assumption, but, still, better than nothing or chicken entrails.) This “scientific” approach also prefers “universal rules”: a catchy E=mc^2 that encompasses everything. Writing billions of formulas to “explain” something is counterintuitive. The problem is that that “something” (price movements) is usually rather complex.
The other group is applying some form or “art”, developed over months and years of observation, study, experience that form its convictions, sensitivity and adaptation to the cycle of trading. Of course it uses rules, but they are a sort of “guidelines” and can be by-passed by someone who “knows better” in this or that circumstance. This group usually can’t explain everything they do in order to take a trade. They are good at driving their cars. They are hopeless at explaining, in rule terms, what they do when they drive their cars. Also, paraphrasing what Shouma and others pointed out, traders do not drive 24/7. And if they need to drive more than 7 hours they start behaving differently, and in the rain they drive differently, etc. Algos can’t cope with all these variables, but at least they are not subject to the vagaries of human biases and weaknesses that cripple traders’ performance.
When these 2 groups meet there is always tension.
So, where does this leave us? I do not know. Perhaps they should keep talking and find a compromise that works for both?
And I leave you with one interesting thought: the super famous Medallion and Renaissance funds’ Jim Simon and his group were not traders at all. They were scientists who knew nothing about trading. Still, at some point (they already made several millions, though) they tried to incorporate some traders' views. A good omen for a compromise? Perhaps, but I would not bet my house on it.
I am sure many of you will know my position on the subject
. For the very few still in the dark it goes a bit like this:Mechanical (algo) good, but usually not great, as many nuances of Price movement are too difficult to code. (I can see the storm coming on this!
)I won’t go into the technicalities (the perils of optimisations, the dark matter of back testing, the demo vs. real, etc. etc. etc.)
I would like to comment on general philosophy, if I can. Trivialising a bit: on one hand you have the coders and on the other the traders.
The coders (usually a rather arrogant bunch, in my experience
) think that everything can be coded, including my dog’s farts, provided they have the “right input”, which is usually a mathematical formula. If I do not give them the exact rules for my dog’s farting, it’s my fault and my problem. Fair enough.The traders (usually a rather arrogant bunch, in my experience
) think that they are applying an obvious set of "rules" that makes a lot of sense to them. They can “tell” when their dog is about to fart. They can see it from the position of the ears, or from that guilty look…whatever. And can’t understand why the coders “do not get them” and instead go about asking for formulas for dog’s farting! How weird!My opinion is that they are talking about different ideas of what trading is about.
One group is acting on data and mathematical formulas. This has many advantages. It gives a scientific rigour to the results, as these can be properly analysed via statistical methods and give more confidence on consistency, provided they know what they are doing, and that one assumes that the future will mimic the past (a very big assumption, but, still, better than nothing or chicken entrails.) This “scientific” approach also prefers “universal rules”: a catchy E=mc^2 that encompasses everything. Writing billions of formulas to “explain” something is counterintuitive. The problem is that that “something” (price movements) is usually rather complex.
The other group is applying some form or “art”, developed over months and years of observation, study, experience that form its convictions, sensitivity and adaptation to the cycle of trading. Of course it uses rules, but they are a sort of “guidelines” and can be by-passed by someone who “knows better” in this or that circumstance. This group usually can’t explain everything they do in order to take a trade. They are good at driving their cars. They are hopeless at explaining, in rule terms, what they do when they drive their cars. Also, paraphrasing what Shouma and others pointed out, traders do not drive 24/7. And if they need to drive more than 7 hours they start behaving differently, and in the rain they drive differently, etc. Algos can’t cope with all these variables, but at least they are not subject to the vagaries of human biases and weaknesses that cripple traders’ performance.
When these 2 groups meet there is always tension.
So, where does this leave us? I do not know. Perhaps they should keep talking and find a compromise that works for both?
And I leave you with one interesting thought: the super famous Medallion and Renaissance funds’ Jim Simon and his group were not traders at all. They were scientists who knew nothing about trading. Still, at some point (they already made several millions, though) they tried to incorporate some traders' views. A good omen for a compromise? Perhaps, but I would not bet my house on it.
