Is it possible to make money with algorithmic trading?

Wherever you spit, an article by a certain Movchan is being discussed everywhere, where he claims that it is impossible to make money with algorithmic trading:

Past me in paper, electronic, verbal and perhaps not tactilely fly, are rushing, float by, dragged through and hobbled back and forth numerous offers to give money for algorithmic trading (anything - stocks, currency, oil, derivatives, etc.).

The proposals are different - illiterate and very neat, with and without a confirmed successful history, for retail and for large clients. In the opposite direction, the opinions of investors fly past me - from "how cool" to "scammers are spamming again". By the nature of my service, I am well aware in general about investment management and in particular about algorithmic strategies - maybe it's time for me to speak out about homeopathy, astrology, investment algorithms.

The investment market is huge and there are a lot of players on it - just like in nature. In relation to real values, investing is a game with a very small positive amount. (formed by the overflow of a part of income from real business to the markets in the form of payment for the capital provided by the markets), in which the participants redistribute mainly that, what was brought to the market, between themselves, not forgetting to pay tribute to banks, brokers, lawyers, tax authorities, scammers, etc.. I.e, translated into butthead language, the overwhelming majority of players simply give their capital to the more skillful and adapted, or - to crooks.

Decades of experience and billions of dollars have certainly given many players the opportunity to adapt to the market environment and adapt markets - as well, how in nature, some have grown teeth, others are claws, still others became very fast, fourth - very large, rest – died. Who are these surviving champions?

These are the insiders. These are major intermediaries, global players, who are able to see the flows and get ahead of them with their actions. These are pirate crews, consisting of top-class professionals, with decades of experience and nerves of iron, who do not even see - feel the quality of this or that investment, just because we have seen something similar on the market more than once. They are monsters, able to invest more than others, conduct on-site analysis by dozens of analysts and experts, agree with those, who makes the policy, organize market manipulations, making the crowd go in the right direction.

Finally these are the ones, who managed to build technology, guaranteeing them ahead of the rest of the players - the most powerful servers, unique processors, programs, noticing arbitration opportunities before anyone else and reacting to them before anyone else. These "technologies" cost hundreds of millions of dollars simply because, that they are constantly getting faster - in this case, the first one gets everything, second - losses. And yet, even all these champions are steadily earning figures that are not impressive to the layman.

Best (if measured on, say, a 10-year horizon) Show 11-12% per annum. Normal, careful and smart - 7-8% per annum, but much more stable. It is quite good if the investor receives and 4-5% per annum - it still wins the market and inflation with a margin. ABOUT, Yes, there are of course those who receive any income, though 1000%, though 1000000%. These are those, who won the jackpot, accidentally hit the bull's eye. Once.

Twice - not excluded by the theory of probability, but not found in nature. And if we talk about stable indicators, then showing 15% per annum on a sane horizon (The same 10 years) - just doesn't exist – with the rare exception of those, who

  Popular trading robots, bots for trading on the cryptocurrency exchange

(and) got a random excess profit 1 once and since then have not eaten it yet (well, say, took Apple with the shoulder at the right moment), or

(b) stood stupidly enough in position, and this position grew (for example if in 2008 took RTS in the fall and lived until the end of 2013). Neither in that, in no other case there is neither art nor technology - there is luck.

What is algorithmic trading, unless it is based on technology worth hundreds of millions of dollars?

Especially if she also brings or promises to bring the notorious "5% per month"? Fraud? Sometimes. But not always. Sometimes it's just & ldquo; survivorship bias & rdquo;.

Guys are going, who have studied the mathematics course of a technical university and have traded on their 5 thousand dollars. shares in BCS. And they decide to cut down on algorithmic trading. Someone believes in their genius from lack of knowledge; someone due to self-confidence normal for a prolonged childhood; someone was lucky while trading at BCS and he believed in his star. They write slow robots (no equipment, channels are ordinary), tuned to simple algorithms (and where can they get the difficult ones in their training and experience) - mainly trade on divergences of pairs with stable covariance, factorial trend recognition, search for simple images, etc.. Groups of such guys gather hundreds of people a year, blessing universities churn out techies and economists, application is not enough for them, and almost every intelligent teenager from a big city 25 years old can program in Russia today, and brokers, ready to connect them to their platform, there are many in Russia and in the world - casinos are always a profitable business.

Their trading strategies are essentially white noise, with a small share of long positions relative to the market, and a sauce of short-term patterns, which they correctly find using regression analysis (only these patterns "creep away" before our eyes). But according to the law of large numbers, their results will be distributed quite randomly., half a plus, half minus. In the first year, half will receive losses outright and for the most part will "merge" from the market.

Thirty percent will make a small profit and decide, that they are on the right track, and will look for new algorithms. Twenty percent will receive a decent profit and believe in his genius. Next year, the ratio will be the same - as a result, in 2 years will remain 4% those, who made huge profits for two years, 6% those, who made a huge profit in the first year and a small one in the second, 6% those, who made a small profit in the first year and a huge one in the second, and finally 9% those, who received a small profit in both years.

After the third year, we will still have about 2% those, someone has made very high profits for all three years, either made a small profit in the first year and very high in the second and third. These will walk with halos and sell themselves right and left completely sincerely.. If in the first year came into play 300 teams, then such great in three years will be neither more nor less 6 teams. Some more will be added to them. 15 teams with more modest, but also good results, they will sell themselves too. If you count, what 10% who entered the game - scammers, then on top of this 21 groups of sincerely mistaken we will have more 30 groups, falsifying their results and approving, that they are fine, and also collecting money. Total each year adds to us conditionally 51 group of algorithmic traders, who sell their services to clients. Pay attention - more 40% of the "successful" really believe in their success.

  Coinbase reports worse than expected. Stock exchange shares fell by 13%

What will happen to these groups a year later (that is - what will happen to your money, if you gave them to any of these groups)? Half of the honest and all the scammers will get you losses - your chance to make money with the team, selling you its three years of successful experience – about 20% (all of them, let me remind you, 51, only half of the 21 teams are not scammers).

Your chance to make big money is about 8% (20% from 21 teams from the total number of 51). Your chance to make big bucks 2 years in a row - already less 2%. Your chance to earn 10 years in a row with such guys - about 1/1024 if we talk about any income and 1/10000000 when it comes to large incomes every year.

And inside the ecosystem of algorithmic traders there is a difficult life, which makes your chances even lower. In particular, some of the "geniuses" are converted into fraudsters upon receipt of their first losses. They cannot come to terms with losses, and therefore for a long time they sell "results for a selected period" or "average over three years", for example +60%, +80% And -90% they do not 1,6*1,8*0,1 = 0,29 (i.e 71% loss), and (0,6+0,8-0,9)/3 = 16,7% per annum, which they present as their sustainable result.

Scammers are improving too: besides simple sampling of the period, fake reports and artificial transactions to change the result, for example, they start two accounts with opposite strategies, and show the official report on that account, who earns this year. Managers are hungry for high commissions and are rather calm about the quick exit of the client, who lost money - during the investment he still paid, and in his place will come another lover of free super profits. Those who use two opposite products at the same time simply divide their assets by two in their minds - one half brings huge commissions and generates new customers, the other half just dumps customers; in the next period they change places.

The question arises - is it possible to earn, transferring money to such a team? The answer is yes. You can earn more than one year. From 1024 teams 1 the team should 10 years in a row to generate profit. If "your guys" 10 years in a row bring you profit - it means that somewhere near at least 1023 investor lost money. What is the probability of making money on 11 year? 50%.

Another question arises - is it really impossible to assume, what suddenly in Moscow (Petersburg, Nizhny Novgorod) there is a genius in the apartment, which will build such an algorithm, well, just a simple algorithm, that he is exactly what will make big money in the markets, and all his clients will be happy, and all not clients are unhappy? The answer is no, and that's why:

Firstly, markets are largely random processes, in which the deterministic component

  • (and) small,
  • (b) Thoroughly studied by thousands of powerful players.

Whatever the algorithm, you can't argue against a random process, that is why all real "algorithms" do not predict the future, but they catch microscopic discrepancies - between the index and the basket, which makes it up, between the cost at different sites, between an asset and a combination of derivatives, which recreates the asset income profile.

  Machine Learning | 88% companies use Data science

These discrepancies are born and die within nanoseconds - because they are awaited and caught., as soon as they appear, hundreds of major players. You don't have mega-equipment - rest, all arbitration opportunities will be taken away a few nanoseconds before, how will you wake up.
But suddenly we were mistaken - and in the markets somewhere there is still a pattern?

Here comes the "second". What is the probability that hundreds (thousands!!!) numerous teams with Nobel laureates in the composition, burdened with expensive equipment and dozens of years of individual experience, did not discover such a pattern, and the genius discovered it? What resources does this genius have?? Where and how does he get the time series of data, which cost hundreds of thousands of dollars to acquire and maintain? What computer does he count them on?? - for the minimum reasonable calculation, mainframes are needed. I do not want to say, that the probability of this is exactly zero, although the number of discoveries in modern science, made on the knee - exactly zero. But even if it is equal to one thousandth, and the probability of making money with random investment is 50%, then i can't tell 50% from 50,1% – if you want to believe in genius, consider that the likelihood of a positive outcome of investing in the product of local algorithmic traders 50,1%. Ouch, do not forget what they will take 2% for management and 20% for income, and the broker's commissions will be from 0,5 to 3%. More profitable (statistically) throw darts into the system screen Bloomberg.

Well, "thirdly". Suddenly the pattern was found and it works. What will happen, if you start using it? On the market, directly on the exchanges, sitting robots analyzing strategies, identifying visible patterns. There are already a lot of them and there will be even more.. A successful strategy will be immediately caught, collect enough data on it, decrypt and copy, and finally - will be used by major players, who are engaged in cultivation and selection of strategies. They will be faster, and eat up your profit from the moment of decryption to zero. Moreover, their actions will change the market., the pattern will stop working - in the market, as in the quantum world, to observe is to change, and investing means changing everything.

Where do happy investors come from?, successful local algorithmic managers, adepts and apologists, hundreds of posts and articles, glorifying successful algorithms? From the same place, where do expensive homeopaths come from, chiropractic, Tibetan medical practices and even folk omens. Their father is human psychology (ability to remember only that, what meets the desires, ability to mistake randomness for regularity, willingness to believe logical reasons, even if they operate on an absurd basis, tendency to ignore facts, going against our desires, etc.).


Yana 95% I agree with his opinion, Partial disagreement is based on the fact that my vision of algorithmic trading, the capital required for it and the amount of profitability does not coincide with the dominant one in the CIS. It is clearly distorted here. Rough, speaking, what I mean — the idea of ​​optimizing the system for some market and starting to trade fifty thousand rubles in the hope of making a million in a year is doomed to failure….. :)

Scroll to Top