Why mechanical trading systems fail
For two main reasons. Either the market refuses to cooperate or the system developer refuses to accept the trading reality. It is the latter case that is much more serious and can do more damage to your account than the former.
Yes, the market can refuse to cooperate and it does so time and again, but if your trading system is simple and robust and you adopt a proper time perspective you should be able to withstand the market changes and come ahead as a winner when all is said and done. It may happen though that the system goes through a prolonged period of flatness when its equity curve refuses to grow and instead remains stuck in a range. This can obviously be frustrating and can cause some to abandon their systems prematurely. That's why it is important to start in a drawdown of at least 50% of the previous maximum drawdown and allow oneself a 150% drawdown reserve in case the market decides to continue digging in your equity curve. When you do this and stick to a 6-12 month time frame, you stand a very good chance to make money. No one can predict how much though and so sometimes when the year of trading the system comes to a close you may feel rather disappointed if the previous years prepared you to expect a better performance. It has been my experience that simple, robust (non-optimized) mechanical trading systems rarely completely fail, they might though test your mental strength more than you anticipated it. It is probably best to have no expectations, but simply set a reasonable time horizon and take what the market gives you.
The other failure mentioned above is of human nature and, unlike the previous one, it can be avoided. The failure of this kind usually happens for two reasons: either the system overtrades or its slippage is not properly accounted for. Obviously, both of these situations can take place too and when this happens you have a recipe for a disaster. While frequent trading per se does not have to be detrimental to the system performance, in reality it is as it leads to more trading mistakes: the more often you trade, the more likely you are prone to various life imperfections and the actual performance suffers as a rule. It is however the other factor, the slippage that is not properly accounted for, that can easily kill your account, particularly if overtrading also takes place. While the slippage for the market and stop orders (the so-called "regular slippage") can be easily estimated and taken into account, this is not so for the systems that use limit orders. (To be quite frank with you, I have seen vendors touting their systems as the true Holy Grail that would turn out to be only mediocre when regular slippage was taken into account. Their performance looked great simply because they traded frequently, but with slippage taken into account this was not so at all any longer.)
The non-fill slippage (for limit orders) is harder to take into account and when ignored, it can easily lead to very inflated profits that will never materialize in actual trading. Obviously, the more often the system trades the bigger this kind of slippage is. The net result is a great looking equity curve that serves as a disguise for a losing system! It is only one way to make sure that something like that does not happen and that is to eliminate from your testing all the trades that only touched the entry but did not manage to penetrate it by at least one tick. I do this routinely to the systems I trade and if they survive such cleansing, I know that they are robust and can be trusted. Alas, this approach is rarely ever practiced by other vendors.
So who designs these kind of systems? Well, either really incompetent developers or the people of the mentality of used car salesmen, always eager to sell you their latest lemon. Because that's what it is, the lemon! Don't assume that systems like that do not exist on the market. They do, some vendors even seem to specialize in them, and they can do a lot of damage to your account rather fast. Your losses can be faster than if the market refused to cooperate, so whenever you see a great looking equity curve that belongs to a system that uses limit orders, do not assume immediately that this curve has much to do with reality as this may cost you a lot.
It's silly to blame the market for the lack of cooperation and you cannot expect it to deliver. But you have the right to expect that the vendor provides you with reliable information about the realistic system performance. Since this does not always happen, there is no substitute for solid due diligence and the age-old saying about a fool and his money applies here as much as anywhere else.
Needless to say, the above comments apply to both (e-mini) futures trading systems and stock trading systems, although it's the former that this site is primarily concerned with.