Successful FX Algo Trading (the hows and whys)

[waza1960]

I have been successfully autotrading FX live using MT4 for over 3.5 months now. So I feel it is time to post some results and look forward to debating some of my approaches with the community.

So my results are from trading 10c/pip as I regard the trading to date as (live Testing) So the results are not about the money but the pips earned and consistancy achieved.I am willing for the moderators to check my live account for authenticity if required.

The main metrics which I am concerned with is avg win:avg loss,drawdown and then % successful trades in that order.I also want a good frequency of trades so that I can arrive at meaningful results for analysis as quickly as possible.
I am trading short time frames 240mins and less and my strategies work on different pairs and different time frames.
I am using some of the following in my strategies:
Advanced money management including initial stops/trailing stops/indicator exits/Break Evens
I am using negative progression ( a form of martingale)
I use a basic portfolio hedging strategy.
I don't believe in Walk Forward testing in these short time frames.




so here are some results to date.................................

So in summary this report is from 01/04/2012 until today (15 weeks of trading) 140 trades ...1790 pips net
avg profit:avg loss is positive ,40% successful trades these results are within 10% of backtesting results.

[tech/a]

Well done WAZA

How do you find meta language?

Seems your method is fine Short but struggles long -- would that. A fair assessment ?

Good pips but fairly flat equity curve.

[waza1960]

Mql4 (metaLanguage) is no harder than say c# as used in ninja trader but I don't do all coding myself I use overseas coders for the more complicated stuff.

Seems your method is fine Short but struggles long -- would that. A fair assessment ?

Yes on those stats the shorts are performing better but my strategies can take long or short positions depending on entry conditions so just think its a reflection on market dynamics atm but its one of the questions that will be answered in the future after more trading.
Also in May I made a mistake with the negative progression setting (it was double what it should have been) so it caused a net $ loss for the month but still came out pips positive for that month.They were long trades so affected the results including the equity curve(see circled trades in screenshot).
June and this month are the first two months that I haven't had any errors on my part in setting parameters etc.
So the strategies would have performed a lot better in the first two months if it wasn't for my errors.

[sinner]

Hey waza, interesting stuff, I'll be following along.

Is the algo running 24 hours a day or are you running during preferred 'trading hours'?

Personally I can see how this kind of approach is very tempting and see a lot of the "hedge" systems posted on forexfactory, not sure why they're called hedge really though. I notice that that daily ranges on FX pairs are generally 50-60% of what they were a year ago, it might be worth thinking about how this sort of strategy will perform when volatility picks up. Even when EURUSD dropped 1.32 down to 1.24 in May ATR was running less than half of Sep/Oct 2011.

[waza1960]

Here's my July results............Still going well IMO
I had a bit of a down period towards the end of the month.
I am actually noticing a possible tendency regarding this.
It is heartening to see win:loss ratio maintaining consistancy around 2:1.
Obviously needing this when Profitable trades are 30--40%.
Its funny when posting these results that human nature dictates that most will gloss over the results..
Will focus on $ amounts and not look at the important bits such as win:loss ratio and Expectancy.
The other thing that is worth mentioning is that this account has achieved a profit in June
based on the starting balance ($200) of 44% and July of 35%.

[waza1960]

Is the algo running 24 hours a day or are you running during preferred 'trading hours'?

The algo is running 24 hrs a day .I have the ability to stop certain hours being traded but am resisting due to
possibility of curve fitting.

I notice that that daily ranges on FX pairs are generally 50-60% of what they were a year ago, it might be worth thinking about how this sort of strategy will perform when volatility picks up. Even when EURUSD dropped 1.32 down to 1.24 in May ATR was running less than half of Sep/Oct 2011.

Actually the more volatility the better this particular systems goes (it has an ATR filter) and my backtesting period
does cover the period you have mentioned . I'm more concerned with how it will go when volatility dies down and we enter a trending market although I doubt if that will be a problem for a while.

[skc]

I am using negative progression ( a form of martingale)

The stats look nice, the thing that jumps out is the maximum loss which I am guessing it is due to the "negative progression" sizing method.

Can you explain this "negative progression" in a bit more detail? How important is this to the overall strategy? If you are using this with a larger account, will it start to affect fills?

My layperson view of any martingale-style sizing is that, the larger your sample size (or the longer you trade), the more likely you will experience some outlying event (like 20 reds in a row on the roulette table). So one needs to do something to make sure that doesn't knock you out of the game...

Do you need to implement an absolute cap on the trade size at some stage?

Since you have a decent win%, win/loss ratio and frequent-enough activity level, may be capping your negative progression limit may actually improve your expectancy?

[sinner]

The algo is running 24 hrs a day .I have the ability to stop certain hours being traded but am resisting due to
possibility of curve fitting.



Actually the more volatility the better this particular systems goes (it has an ATR filter) and my backtesting period
does cover the period you have mentioned . I'm more concerned with how it will go when volatility dies down and we enter a trending market although I doubt if that will be a problem for a while.

Thanks for the response. Gave the "good" answer to my first q

Not sure I understand your second q answer though, your backtesting covers the last year or longer? If only the last year the vols have steadily been trending down?

An indicator like this one might help with your concerns

http://engineering-returns.com/tsi/

It's not the only one of that sort I've seen, but quite like it compared to others.

[waza1960]

The stats look nice, the thing that jumps out is the maximum loss which I am guessing it is due to the "negative progression" sizing method.

Can you explain this "negative progression" in a bit more detail? How important is this to the overall strategy? If you are using this with a larger account, will it start to affect fills?

Negative Progression is increasing position size after a losing trade (or series of losses) forgive me if I don't give exact details here.
Martingale is a form of Negative progression where you double the position size after each loss (my neg prog is less agressive than that).
My strategies do not rely on Neg Prog to be successful and I backtest/optimise with neg prog set to false...
As far as trade size and fills go I will worry about that when the time comes just chasing consistancy ATM.

My layperson view of any martingale-style sizing is that, the larger your sample size (or the longer you trade), the more likely you will experience some outlying event (like 20 reds in a row on the roulette table). So one needs to do something to make sure that doesn't knock you out of the game...

Yes this is true but this isn't gambling (or at least i don't think it is LOL) or a mathematical exercise in my view.
I have backtested systems with a certain number of stopouts .If in live trading they exceed their stop outs by a predetermined amount then they will be stopped. Also the aforementioned portfolio hedging assists here.

See below from another forum ,my strategies have been built around the statement on the last paragraph below

Consider this....when speaking in binomial 50/50 equivalents.....

a strategy that's 50% win rate, needs a 1:1 Risk Ratio (techinically better with slippage and fees) or better
a strategy that's 33% win rate, needs a 2:1 Risk Ratio or better
a strategy that's 66% win rate, needs a 1:2 Risk Ratio or better


Given that a system with a 50% win rate can expect a 1.5% chance of six losers in a row, so a sample size of 100 trades....

the bottom line is that the larger your sample size, the more chance you have of longer strings of winners and losers.

In order to reduce the consecutive losers, you could try to increase the R/R ratio....but usually that comes at a cost of win rate (it's very difficult to increase your profit targets and decrease your stops and still maintain the same edge)

Since you have a decent win%, win/loss ratio and frequent-enough activity level, may be capping your negative progression limit may actually improve your expectancy?

I am running a spreadsheet on Negative Progression Analysis of the trades so far there are some trends emerging in the Data but need more trades to confirm.
For example it may be beneficial to have no progression on the first losing trade and then activate.
BTW Thanks Sinner I always enjoy your links