I have read the beginning. I am, however, pretty familiar with several such trading systems, and the anticipated results. There are definitely some spots I see in the chart where this would likely have picked a losing entry point, but the underlying system is still valid. All other things being equal, it will make money, at least for a time.
The issue is more the fact that the bot is A) predictable, and B) prevalent in a small market
In a big market simple ema bots aren't so easy to game, there's much more going on. In a small market like bitcoin, it's not terribly hard to place a trade that causes
these bots to enter, and exploit the known result. (not necessarily to the bots' disadvantage, but often so)
Goomboo addresses B in one of the early posts -
Since the Bitcoin market has so little liquidity, I suggest that new trader try a variant of my system. If a large group of people were to trade the 10/21 crossover, the entries / exits would become crowded and price would move violently around those points, hurting us all! I suggest that people try things such as a 9 / 14 crossover or a 5 / 8 crossover.
A bunch of bots all predictably doing the same thing with not much liquidity is begging someone to exploit them. So often, we're trading in a very tight range, so it's fairly easy to place a trade that will cause a crossover (if you have enough coins to bet on it!)
Goomboo also points out that this is a trend trading system, and as such may have up to 70% losing trades. It only takes a couple bad losses to wipe out many gains in these systems. (as he also points out, the potential drawdown in such a system when used with leverage can be risky)
The basic problem with this type of simple trend system, IMO, is that you wind up trading counter-trend, and in doing so, you set yourself up to be on the wrong side of a trade quite often. The is especially true in low liquidity, low volume markets.
I use simple technical analysis - S/R zones and Fibonacci sequences, but not as Goomboo suggests many traders do - "predictive analysis" - As he said, prediction of a market is not truly possible. Identification of potential entries, and entering on confirmation through price action when it actually occurs is more successful in my experience.
as above, I'm having a look at the bot's code to see how hard it might be to add a few indicators, especially volume, and maybe order depth to determine the real spread. Mostly I'm looking for a couple indicators for when it shouldn't take a trade.
And, looking through the Mt.Gox API, the simplest thing to do might be to take 2weiX's suggestion to make the EMA completely configurable, and to use the VWAP rather than the Open price. I don't know how one might go about backtesting it to see what its profitability and W/L ratio might be.