Trading is statistics, generally. No trade is safe beforehand and only hindsight trades look as if they would have been riskless. Most trading systems yield a gain to loss ratio of about fifty percent. If this ratio is much better, chances are the accompanying sizes of losses and gains are malformed.
Forex robots come to mind here. Securing the small gain by closing a trade early, but holding a position that is under water, hoping that it finally turns around and becomes also a small gain, is their standard system.
Autotrading robots are often programmed this way in order to impress the potential customer. Eventually all these stop loss averse robots get hit by the hammer and lose everything in one trade, or more, that they accumulated with their favorably looking win to loss ratio.
Another version of pushing the loss into the future is doubling down the ante and playing the martingale. The day will come you Forex believers when you wake up, still drunken from your sweet dollar dreams, and your nightly autotrading robot smoked your whole trading capital in one final trade.
So, all trading is probabilistic. That is the very nature of trading. What is statistical trading then? It is a kind of a trading with entry and exit points that don’t seem to make sense for a seasoned trader. The signals are generated by a trading system based on a complex algorithm. Instead of looking for the usual suspects like chart patterns or indicator signals, something the human trader does also, perhaps subconsciously, these systems use a trading algorithm that identifies hidden regularities.
Technically they are known as neural networks, assemblies that mimic the linkage of neurons in a brain. Brains are thought to configure their neuron network on a selforganizing basis. The same principle got applied to AI programs with a neural net.
This network of software neurons gets trained, for instance, with the price history of a stock, and adapts itself during the process. After that training phase the assembly of neurons are like a little brain that is able to understand the specific price behavior of this stock. In other words, it is able to forecast the near price future of it.
Of course, this price forecast works only statistical in itself, like the weather forecast. There is no sure thing. But more important, the prognosis this neural net will make, is often hardly believable for humans. If this price forecasting gets used for generating buy and sell signals, these trading signals may look strange at times for a human trader.
Yet, this type of trading works, somewhat. If applied in a trend either directly or with some additional trading rules to make exploiting growth stock trends possible, it works even better.
Have a look at this video of a trading signal generator based on a neural net and look at the screens that show how the signals are distributed. Some look right, but some seem to be outright wrong, eh… interesting.
