Here are a couple screenshots to illustrate the point that autotrading, for me, is the only way. The first is the result of a walk forward test of a strategy I'm working on. The strategy was trained using a stochastic optimiser over 2007-2010 15 minute bars of EURUSD. This found the best parameters for maximum PnL / drawdown (aka smoothest equity curve).
Next I re-ran the test for the full dataset, 2007-2012 with the optimum parameters. Ok the equity curve is still advancing and still smooth. All good I think.
So I zoom in to the chart to see the sort of trades its taking. WTF? Tiny scalps that make no sense. its missing huge moves and taking 20 pips here and there. No way could I personally trade this, but the software says statistically this method is extremely good. Stable for years and slowly climbing the equity curve.
I have yet to run this strategy live - I'm still testing, but I'm encouraged by a walk forward. Whats odd is when I try to add "intuitive improvements", for instance, adding a trailing stoploss, or rules to maximise the profit, it fails. I'm only using a crude stop (very wide) on this to protect against system failure so for all intents and purposes its a no-stop system.
My first and so far best trading strategy for the EUR/USD. Also works on Stock index futures. The probabilities don't lie, even though they don't make sense ...
I agree with the following two points made previously:
1) While the "IT" or programming aspect of developing an automated model may be hard, I find that identifying a solid trading strategy that is based on sound market principles is even harder. When developing quantitative models, it's possible to generate strong backtested results, but if the core principle of the model is not sound, you're at risk for performance drift (or complete breakdown) when trading in real-time. That's why I've shied away from using neural nets, genetic algorithms, etc. although I find the ideas behind them fascinating.
2) "I've always been amazed by how the obvious is so often overlooked by most... Basic concepts, using (often) simple arithmetic, can be quite useful..." The models I use right now in the stock indexes are not rocket science, and several came from books that I adapted to the markets I trade. The fact that several of these are based on market ideas that have been around for ages gives me more confidence to trade them. I have backtested them all as well.
Larger pic of the equity curve. Please note this is the same strategy but with a few tweaks so not the same curve as the other image.
Yes some periods of drawdown, some significant. The red shaded area from 2007-2010 was my training window and green was a walk forward test.
Also note this test traded 10x contracts of EURUSD $1 mini (aka $10/pip). It makes $70,000 in a few years off a $10,000 account. Yeahhh right! I'm not believing this performance, its the increase/volatility of the equity curve I am interested in. In reality I would probably trade live 1x contract off the same account size.
@Fury Developing strategies is definitely the hardest part! The software tools only really help you do this more quickly, but one (successful) trader I know developed his method manually over a period of 3 years. Now executes manually with varying but steady success
Favorite Futures: Futures - bonds, currencies, index
Posts: 288 since Oct 2010
Thanks: 70 given,
That's exactly what I'm thinking. To be honest, if you see a decent equity curve, you usually find there's a problem with something, hence my questions.
I don't just see a potential for losses when live. I see it guaranteed, if I read this right. I don't know exactly how your commissions work in FX, but if it was futures (which is probably similar) then $3.5 per contract average trade is completely wiped out by about $5 commission per contract. In your case it may be $6. That's before we've even started on slippage.
Seriously, don't even bother with your equity curve until your average per contract traded is way bigger (pref much more than treble) than your commission per contract.