Heart of the Matter
Now onto what I really do - or better said, what I am really meant to be doing when I'm not trying to give NinjaTrader the kiss of life or get IB to connect or BT not to be so bollox.
I'd almost forgotten, but I'm actually trying to morph the 3 Ducks into a profitable system across all the markets I point it at.
I had it working profitably on 10 forex pairs back in September which was a good month for trend following, although it got a bit hairy because my choice of markets was very dollar-centric - I ended up going into non-farm payrolls short 1.2 mega$.
Secondly there was a fearsome bug in the code with kept me in at least one trade by missing the exit signal - there may have been more - although in retrospect those losses are nothing compared to my own self-inflicted errors.
Plus it was hopeless going short. Made no money. I hadn't realised this in backtesting at all. I figure now I have to optimise long and short sides seperately and if I can't make money going short, I need to understand why.
Plus when I tried to diversify into non-dollar pairs, the results were rubbish.
So I pulled the system out of production and replaced it with another, and have spent the odd hour here and there since September trying to improve it.
I didn't realise I'd got so little done. I need a new deadline. How about June next year? Deadline June! Thanks for suggesting that MeanReversion. It gives me way too much leeway to faff around for six months or for my other trading system to blow up and sink the whole operation.
I do have some progress on it though - I've managed to make it work on most of the 20 forex pairs I have data for, but I haven't worked out why it doesn't like going short.
I played with every single imaginable combination of entry trigger and exit mechanism using 3 MAs and pivot points, but I ditched the pivot points because I don't think they work on the lesser crosses like AUDCAD. All the techies who make pivot points a self-fulfilling reality are too busy watching the Euro, the Yen and the Pound to bother seeing where today's pivots are for AUD and CHF.
I picked up a couple of tips bizarrely from a blog entry on a fraud prevention blog
Catch-a-thief. The author promoted the Guppy MMA which is just a set of 30 MAs all drawn on the screen together and coloured differently. Pretty much like having 3 Ducks really, except 10 times as many.
The author also chucks in a volatility channel filter, a trend line slope indicator, MACD, RSI and if the 30 MAs wasn't enough, then a pair of exponential MAs as well.
But having said that, the backtest results without any optimising yet are a vast improvement on the rubbish results I was getting first up as I tried to use the pivot points as entry points, and secondly as I tried to tie in some indicator divergence to filter the entries when the 3 MAs were all in the same direction.
I've also had a constant stream of improvement ideas for the system that's in action right now. Some of the trades it takes are inspired, but some are absolute lunacy. It will be interesting to see at what point adding in improvements will turn into curve-fitting.
I can predict the outcome already - even if I find two different improvements to make to the system which backtest with significantly better results, I bet they will adversely affect each other when combined.