Critique my results

[Steve Watts]

Hi,

I have been working with EUR USD historic data downloads from dukascopy and backtesting a system in excel. A typical weeks results look as follows (before any spreads, commisions or fees)

Trades: 647
Profitable: 235
Even: 36
Losses: 11
Best: +0.0011
Worst: -0.0003
Average: +0.00023

The results are consistent but the profit level is small and then there are speads and commissions to consider. I would appreciate opinions on what my next step should be - whether this could be viable and how it could best be executed - I am considering Trade Station or Interactive Brokers and more testing.....

Regards Steve

 

[everyonerich]

test on live account

 

[anley]

Steve

It's 100% pointless looking at any trading strategy unless costs are taken into account.

 

[anley]

never forget that costs play as an important role, if not MORE important, than buying/selling at the right time.

Costs you can control, everything else you can't.

 

[Stewenson133]

2 pips average profit per trades?

 

[Dacamic]

As others have implied, the subject strategy will most likely have a negative expectancy per trade when transactions costs have been properly included; thus, it is not viable.

 

[shadowninja]

Backtesting will give you an idea. Forwardtesting is the only way to determine if it will work. Charts look different when you can't see the right side.

 

[meanreversion]

647 trades in a week, are you nuts? You will develop arthritis within a month. Your platform provider will love you, but your bank manager won't. The bid/ask will remove all profit, along with the cost of replacing your keyboard and mouse on a regular basis. Try moving up to a longer time frame and reducing the number of trades per week to no more than 50.

 

[arabianights]

I don't understand how there are 647 trades, of which 235 were succesful, 36 were b/e and 11 were losses..?

 

[Steve Watts]

Hi all,

Thanks for the feedback.

It is a typo on the trades - not quite sure how why i put 647 but it should have been 282.

I have tightened up and optimised the entry criteria which reduces the number of trades and improves the return, an average week now looks like:

Trades: 57
Profits: 47
Even: 3
Losses: 7
Best: +0.0030
Worst: -0.0005
Average: +0.000612

However the above still does not include spreads and costs, so that is what I will look at next and hopefully trial some live data.

Currently I am using excel to automatically generate long/short signals from the historical price and volume info. Are there brokerages/platforms who can stream data into excel and then take an instruction from it? or will I have to manually enter the data in excel and then the trades on the screen?, or maybe use another program other than excel to inface with the broker?

Any experiences or advice on the possibilities or best ways of achieving this would be appreciated.

Thanks Steve

 

[meanreversion]

Average spread in EUR/USD is 2 pips, so that will remove a third of your profit at least. How can the worst trade be only 5 pips and yet 80 pct of the trades are winners? I guess what I am saying is that the majority of spot trades can easily go offside by 5 pips before recovering, except yours would have been stopped out by then. What is the essence of the trading strategy here, is it RSI/oscillators/Bollinger bands/something else?

 

[meanreversion]

Also, did you plan to trade this manually or through an automated system.. if your stop loss is in the region of 3-5 pips, then this would be a very manual process to operate (e.g. if you use IG index, probably the best platform out there for managing risk, it won't let you put stops as close as 5 pips)

 

[arabianights]

This is clearly a very short term trading strategy, are you sure you can actually act upon your signals quickly enough in a live environment, where every pip matters?

 

[Steve Watts]

Acting on the signal quickly enough may be a problem, I would like to automate the process with the expectation that would help.

The opening signal falls in the 'something else' category, I havent seen it documented, it's not in Kaufman, which I believe covers quite a lot, for example.

 

[Calinor]

I have tightened up and optimised the entry criteria

Did you curve fit the second one? You should be careful with optimising. You can get some wonderful results, which aren't realistic at all.

 

[Steve Watts]

I've tried to avoid overfitting the data, having previously analysed other approaches that gave great results but then didnt hold up on larger data sets.

In this case incremental 'tightening' each of the entry criteria gives a reasonably proportional increase in profit and reduction in trades entered, and fits with the logic of what the method is doing.

The exit is still unsophisticated so maybe there is another 0.0001 to be had there.

 

[shadowninja]

 

[Calinor]

Steve, it is worrying that you are looking for an extra pip. You can tweak your system to get an extra pip, but in real time, price will jump around and when you click to close, that extra pip might turn into a -1 pip. Also despite your reply, I still feel from that reply that you have curve fitted. I have done that myself long ago, and I don't believe it is a good path to take. the backtest results will look amazing

These are also your back test results (dubious), and your live results are likely to be much worse. You've also missed out some important numbers on your results. What is the average win, and what is the average loss. Without that, just stating the best result and worst result is meaningless.

The only way to see if you have anything there, is to take say 50 trades (more would be better, but I think 50 will give you an idea), with MINIMUM stakes. If you can trade 10p a pip, do it. Or do it on demo.

 

[Stewenson133]

you'd better give it away

 

[Steve Watts]

you'd better give it away

Only if you code Java or have experience with FIX CTCI.

Or if you can teach me basic statistics as the other post suggests.