Why is it hard to make an Automated Forex Trading System to Work?

That's right.   It is hard to devise sound trading system that works on the long term, especially systems designed to work with one forex pair only. The markets seems outsmart any possible indicator. It doesn't seem to be any fixed rule about each indicator, like for instance 'enter at a bollinger band cross'.  Every time the indicator can double cross you, or even if it doesn't , there is no proper indication of where the take profit or the stop loss levels should be. For example bollinger band breakout:

- the market can walk outside the bollinger band cross for a while, then fall back into;
- the market can walk along the bollinger band cross, intersecting seldom, then fall back;
- the market can stop anytime and turn back swiftly.

Conversely, if you want to trade fall backs, the market can continue to trend ; or if you wait for a fall back behind the bollinger after a trend, the volatility can jump so much that when moving average cross produces the result will be random. And these, of course, if you coded correctly the criteria for such a fall back.

Forex has a powerful random walk, bringing chaos to indicator signals.  The strangest thing about it is that the random walk is not arbitrageable white noise. There is no actual meaning of Forex, as with stocks. There is no "growth" in Forex. At least the stock holder can exit the market if he doesn't want to short and wait for better conditions to go long (perhaps this was the origin of "long" and "short" - in the roaring 20s stocks were soaring for long time and falling for short time), but the forex trader doesn't have this luxury - pairs can be interpreted both ways.

There are also indicator less systems and trading strategies, but they are not problem or risk free.  You can still take multiple losses in a row or spend a lot of time until they become profitable , and also they may require low leverage to allow high drawdowns.

A system must have a surefire bet inside. And you can't find a surefire bet on one Forex pair - unless you trade it on accounts with two brokers . There must be something that the system can rely on - either arbitrage with synthetic Forex rates, Forex to futures misalignments caused by latency difference, option strategies based on volatility that require also trading the underlying.  At least futures as CFDs were implemented until now, but back testing multi-asset systems is still very difficult . I have hopes for this from Meta Trader 5. I am also waiting for options - because I'm sure that the sound strategies that we all read about in finance books are back testable and applicable - gamma scalping, [almost] costless collars - strategies that the Forex traders don't know because the proper assets were not offered - due to stations or regulatory requirements.  It doesn't matter if Forex systems can work or not in the long term. It matters to have access to markets and tools that allow to backtest and trade the whole spectrum of markets - Forex, Forex Futures, commodities, stocks, options, or even bonds and credit derivatives - that's when opportunities arise for automated systems.