KISS MA systems that work on longer time frames

Hello everyone,

I am aware that the typical MA crossover strategies usually get slaughtered in the chop of the shorter time frames but are somewhat more benevolent in daily charts.

I am doing my first attempts at programming my first Experts for TradeStation and MetaTrader and I'd like to hear about relatively simple but yet feasible Moving Average-based systems for the longer time frames so I can tweak them and get acustomed to programing my own stuff.

Thanks!

Have you seen my post at http://www.forexfactory.com/forexfor...ead.php?t=3291?

I don't know coding at all, but it seems like having a MA change direction 5 pips between 1 am EST and noon EST on the 10 minute GBP shouldn't be too hard. Also, I think 25 period MA works better the the 30 period I was initially watching.

Daily average has been consistently 17 pips even with a worse case largest drawdown of 300 pips in the month.

Regards,

Agent 2005

I came over to the Forex world after trading equities based on daily charts. One of the methods I used was simple MA to determine when to get in & out of a position. So, when I switched over to Forex, I was interested to see whether a similar approach would work. After doing some testing, I found that it does indeed.

I played around with various optimizations and the following works quite nicely: 3 EMAs (short, medium, long) with a buy/sell when the medium crosses the long and get out when the short crosses the medium the other way. Now, each currency pair exhibits its own patterns and behaviours, so the MA values will be slightly differently in each case. I simply got the computer to run through a loop with different values for each MA and then kept the values with maximum profit. Here are the optimal EMA settings for the pairs I tend to trade (using 2001-2005 EOD data):

EURUSD: 9, 28, 40
GBPUSD: 7, 26, 45
USDJPY: 3, 22, 55

Caveat:
1) I did not do any forward testing
2) you might find that calculating how quickly the short MA goes in the other direction (its tangent) can be a better indicator as when to get out of the trade
3) a SMA (vs. EMA) might work better for the short MA
4) initially I wanted to calculate a profit target as well, but found that leaving it out altogether gave better results

Julian

bluemeasure,

Do you apply your EMA, to high, low, open, close, or median(hl/2)? Or does it really matter as long as all three are the same?

Thanks in advance,
Jimmy

I used the close in each case; that's what I had used in equities, so I just transferred the same logic on over to Forex. Because of the timeframe (daily), I suspect the results would be much the same as with e.g. (h+l)/2. It seems to me that trading approaches based on MAs become more sensitive in their values and parameters as the timeframe decreases.

Bluemeasure, how did you deal with the choppy markets? From watching at the daily charts, it seems to me that the pre-2000 charts show trendier markets altoghether, but post-2000 charts are more range-bound, even on the daily. And MA-based strategies are all based on riding the trend.


Thanks!

As I mentioned above, I used 2001-2005 EOD data (up to the end of Jun for this year). The values that I posted were the ones that generated the most profit over that period of time; that does not mean that it didn't get caught in some bad trades during choppy trends, but the returns overall more than made up for this drawback.

Hi Just a cautionary note as I've fallen many a time into a trap of optimizing for the whole data period. There's been a few great systems I've tested in the past that worked great when optimized, but became very mediocre when it was tested on new data.

I would want compare the best values for 2001-2004, and then see if that would work on 2005 data. And to check validity, I would also optimize for 2001-2003 and use those values to check for 2004 trades. That would then be more realistic of what you can achieve. If you get good results great. If the results wobble then one would have learned something.

There's a post around here somewhere on the forum on system testing, you may want to do a search for it.

Best regards
Alan

Absolutely right, Alan, thanks for pointing that out. You'll note that to drive this point home I cautioned in my initial message:

"1) I did not do any forward testing "

The info should therefore be taken as fyi and not as something you can just chart and start trading with.

The other thing to keep in mind with most any MA system is that when used live, you'll find one MA crossing another signaling an entry only to have the price retrace enough to negate the signal. Now what? get out of the trade? stick it out? That's why the initial question about long-term type trading using MA crossovers makes more sense in my mind because you can take a cut-off point (e.g. EOD) and either the MAs crossed at that point or they didn't and also the larger timeframe will be more indicative of a trend.

In any event, I merely wanted to present MA numbers that worked for those currency pairs for that timeframe that show that each exhibits different characteristics.

Good trading,
Julian

Bluemeasure:

What timeframe charts are you using for these EMA settings? Thanks.

These are the results for Daily data.

bluemeasure,

I just tested this for the last 10 years of data trading 1 contract and it loses 33K during this time. This was the GBP parameters that you describe above. I didn't try the others as I suspect the results are similar. The data is Olsen data from 1/1/94 to 22/3/05. I am buying on open the day after the crossover and selling at open the day after the crossover. Open/close of the bar is midnight GMT (although this would make little difference).

Simon

Yes, that shows that markets change over time. I would have been surprised if these particular numbers would have worked for time periods outside of the one I used. My aim was to answer the initial question of whether a MA trading methodology can work. I believe it can and then keyed in the paramaters that worked for those pairs for that timeframe to make the point (off course with the help of hindsight) If you are looking for numbers that will work for longer timeframes, you will have too loosen those numbers (and therefore profits) considerably.

It could be argued that results based on historical data invalidate the theory altogether. I don't believe so as there are such trading methods out there with more "vanilla" number that work over longer time periods and over most if not all pairs. Posting these MA numbers was to make a second point: each pair behave differently and that a little tweaking of these MAs can help the bottom line.