Best Trading System ever: performance limits.
The best trading system is still a model and not the phenomena (i.e. the market) it is trying to describe. This clear frontier should be understood by developers searching for the ultimate limits towards that ubiquitous idea entitled the "best trading system".
As much as in physics, a theory cannot be more than a conceptual-mathematical model to describe the observed phenomena. Analogously, the best trading system ever made will still be an approximation to the efficient frontier that distinguishes the market (the dynamic phenomenon itself) from the trading concept (the model that tries to describe it).
The "perfect trading system" would turn from a system reacting and gauging unknown events and uncertainties to a mathematical model that, for the sake of naming things properly, "know the market": an unfeasible deed, as the amount of possible patterns approaches those generated by a mandelbrot fractal.
A mathematical model to predict stock market targets with certainty is a myth. What exist are models to gauge probabilities and telling what the actual (not the future) trend direction likely is. Prediction is something entirely different from evaluation. The best trading system is a solid model, capable of correctly doing this evaluation, translating the event into a probability, allowing traders a clear access to the current operation, their chances and risks.
Those searching for the perfect timing machine, capable of knowing where prices will be, eventually will find a mirage, a system with impressive performance, just to learn that it only work in a given market, for a very specific timeframe and what is worse, for a season. What they get is just a trading system specialized in a specific market behavior that lasts for an ephemeron period of time, after what they lose synchronicity with the price dynamics failing miserably.
The Perfect Trading System is unfeasible
Those searching for the most profitable trading system in any situation, will practically study history via price data, rapidly in the hope of finding out a generic model that would be able to survive throughout a trading life. It would be the inverse of an specialized model. The final result would be a system very different from the idealized "best trading system". Instead the model would be like a duck, that can fly, swim and walk, but that don't do any of those things as a top performer. The system would at best, survive with a meager performance.
Those searching for the most easily tradable trading system, would be trying to obtain what works best for them as humans trading an inhuman environment, looking for psychological comfort while executing strategies, and perhaps dominating a few setups (identified events that constitutes in trading opportunities for their pattern) with higher chances of a positive outcome, being in the sidelines while similar opportunities fitting in that criteria are not setting up (for instance, building a long position in a certain way that is psychologically comfortable whenever the market finds itself aligned well enough with a small set of 4 or 5 criterions that, throughout history was very profitable, instead of trying to be in the market all the time, what would require insanely complex mathematical models.
There is another approach, which happens to be between trend-following systems and proactive ones.
The ideal trading system does not build exposure before the start of a trend or reversal, because while it is possible to detect when prices are over-extended (and here we should emphasise that accuracy will vary depending on the technical expertise), those situations not always will materialize in a reversal, switching for another trend in the opposite direction. Both timing and trend are key elements: it is useless to correctly timing a top if 6 on 10 times the trader close his long positions, initiate shorts, just to find that this top, while correctly timed, was unable to affect the prevailing trend.
Best trading system or the best system to trade? tradeability is key
The opposite type of problem haunts those who rely solely on trend rasters, waiting for too much confirmation before jumping in the trend when the price is way off turning points. In practice the slower the detection (which many trend-following systems are purposely lagging to filter “noise”), the later the trader will receive a signal, forcing him to practically buy in the middle of the uptrend and selling in the middle of the next downtrend. For some unknown reason, most people don’t realize how rare are long term trends, necessary to allow a slow trend follower to profit.
Why is that, there are nothing between trend-following and “overbought / oversold”? Because building one of those two types of trading systems (or tools, because most are not even trading systems) is easy, while translating the market dynamics into hierarchic equations is an endeavor involving too much resources. Usually a trading system that really work are belong to large institutions (or independent former professionals from the financial industry) who can subsidize serious research in the field.
Those systems have no form. They do not superimpose a pathetic pair of moving average to reality. Do not expect trends after events that anyone can easily track in a chart, as such a breakout. They exploit the situation, not the market. They must be coded to exploit the largest known set of possible situations, which means its library must comprehend a huge database of samples, patterns and its algorithms logics must be capable to correctly use this database to sort and generate signals. It is impossible to have a complete database, and the system must acquire and learn from its own forecasting processes as much as human traders do through the emotional memory (sometimes damage) their bad trades cause.
A neat, ideal trading system is the one capable of anticipating turning points without futile attempts to pinpoint turns: it's aware of overextensions and other extraordinary events that often mark turning points, without falling forit immediately.
Best Trading Systems dealing with risk The dangerous downfall
The best trading system is the one capable of recognizing real patterns, preferably a large collection of trading patterns instead of imposing a frame over the price data, and without falling for the futility of catching the proverbial falling knife, in anticipation of turning points. While the capability to detect a market exceptional status that is likely to eventually result in a reversion is invaluable, this feature is better used to scale out of positions in the actual direction of the "probably" fading trend, than to start building up a position in the hope of a change that is purely based in early symptoms.
Uncertainty related to an outcome is the sole property of reality that limits over exposition. There is no uncertainty on playing the Russian roulette game either with unloaded or fully loaded guns: the outcome is previously known. Similarly, if trading the hypothetical perfect trading system, the exposure due to the lack of risk, would always be the maximum possible, limited only by the market liquidity itself. Bringing this delusion to a more tangible terrain, the best trading system (i.e. a trading system with super accuracy), would theoretically allow for large exposure, just because the loss probability is substantially smaller than the win probability. However, even if 9 in 10 trades are profitable, it is convenient to think it in terms that one will be a loss, and if that loss is large enough to destroy the capital then the best trading system will blow your account. A consistent trading system with such win:loss ratio did never exist and never will, but very good trading systems traded recklessly because of a great performance routinely blow accounts somewhere, particularly accounts of FX, futures and option traders, that are overconfidently and recklessly prone to overleverage.
Best trading systems can still blow it. A sad, not so unusual example
Let's suppose that every time this options trader has a profitable operation he gains 1$ for every 1$ in the trade. On the other hand, when a trade results in a loss, only 70 cents out of 1$ is lost. His potential gain is higher than his potential loss. The signal accuracy of his hypothetical "near perfect trading system" is 80%. That means the mathematical expectancy is very high, making it in one of the most profitable trading systems out there. How can he blow his account? He decide to put all his money (let's say 100k) on the line to “maximize potential profits” in the short term, and be "aggressive". This is 100% exposure, or 1:1 leverage.
In his 1st trade, a win! Wow he turned 100k into 200k, thanks to his 100% exposure. The second one, another win! 200k easily became 400k. What can go wrong? A third win and he is trading 800k. The fourth victory comes, after all with 80% accuracy this is not unlikely... He is trading well over 1 million and now can write books, give interviews and workshops and all that stuff. What if after making 1.6million, he had only two bad outcomes in a row.
The fifth trade, with an exposure of 100%, resulted in a loss of 70% of his total account, and 1.6 million became 480k. Being the sixth also a loss, he is back to 144k.
Now what to do? if he thinks that 44% over his original 100k is a great result despite the insane equity swings, and that his strategy will succeed (after all with a system with 80% of accuracy what are the odds of having a streak of 3 losses in a row?) them he is doomed, because the chances of success or failure of every trade has nothing to do with the outcome of the previous one. So, if he is unlucky, his 144k will become about 43k (a 60% loss over his original capital). And if he is lucky, will still have a few more attempts before acknowledging that, with such levels of exposure to risk, his survival chances lean to zero in the long run. Another loss and he will be out of the business in less than ten trades.
In a simple sentence, every system (okay, only those with less than 100% of accuracy…) may only be traded with a reasoning risk management model.
Everything is the very thing Risk management cannot solve all your problems
Perhaps traumatized by the experience, he concludes that the key is to risk the minimum per trade, so he will have such a large number of opportunities to make his system work, and such a thick buffer zone, that the chances of being expelled from the market are almost zero, while keeping the minimum allowed exposure (one mini contract or something like that). And there he is paying disproportionally high commissions (relatively to the potential gains) and trading to generate small loses even when he lose and nearing no residual profit when he wins, due to slippage.
Beyond that, from a certain point of efficiency, a risk management model should not allow exposures above a quarter (25%) of the total capital (frequently, this number is LESS than 25%), not mattering if backtest results of a trading system with very high accuracy leads to the conclusion that it's performance would be maximized with higher exposure. Why this number and where it came from, is beyond the scope of this article but may be further studied in the x-tudio area, in the Risk Management videos tag (only available for members). And yet, risk control cannot turn a bad system into a profitable. When reading such statement in some risk management specialist book, know that it is a lie. Mathematical expectancy is crucial and should be clearly positive before any risk management become possible. Therefore risk management is no freaking holy grail. You totally need the best trading system you can find and if not develop it yourself.
So, success is the product of multiple indispensible ingredients. Or in a sentence, is about knowing the optimum exposure to trade the best possible trading system, which happens to be a realistic compromise between accuracy, psychological comfort and tradability, and consistency in multiple markets, season after season. We hope to help you the best way we can! Before every trading session remember that probabilities mean no certainties.