MA Cross Optimization EA (very cool) | Page 7 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

New very cool CandleTime and session time Clock indicators 189 replies

Rsi cross indicator? (NOT cross 50, etc) 44 replies

very cool Trading Panel EA needs some adds 19 replies

Royal Cross/Fresh Cross 7 replies

World Market Hours - Very Cool 8 replies

Attachments: MA Cross Optimization EA (very cool)
Exit Attachments

MA Cross Optimization EA (very cool)

  • #121
  • Aug 16, 2008 4:10am Aug 16, 2008 4:10am
Would someone answer my questions in post 117?
  • #122
  • Aug 16, 2008 6:12am Aug 16, 2008 6:12am
Quoting eagle4x
For MA_Optimizer, I obtained similar results as Tdion using daily charts. However, if I try any time frame lower than daily, the optimization results are not good. When I first tried optimization, there were no results, and I looked in the journal and there was a message, "81 passes were made and 81 were disregarded as insignificant". I then unchecked the box to skip useless results. I was then able to get results - for 1m gbpusd, the optimized profit ranged from 0 to -10K from 7/08 to 8/08. Has anyone been able to get good optimization results on charts less than daily? If so, would you post in this thread? Any thoughts on what I am doing wrong? I have downloaded history records within my platform (FxPro): # of records are: 1M = 3000K, 5M = 703K, 15M = 237K, 30M = 119K, 1HR = 60K, 4HR = 15K, Daily = 3K.
Try using the aplari history database.
  • #123
  • Aug 16, 2008 1:18pm Aug 16, 2008 1:18pm
That's right rabid.... your assessment is correct, and I only take longs when the RSI 96 is below 50 and shorts when 96 is above 50 (on the hourly) ... the results speak for themselves.

As for the exits.... I optimized those after I found the optimal RSI period and thresholds...

Quoting Rabid
So long w/ an RSI above X? You're optimizing for X and the RSI period?

What exit?
I am a bit disappointed with the backtest optimizer. Although it won't lie to you (the results it finds are true based on the parameters it finds) I have found cases where it ignores some permutations which were actually profitable.... in some cases even the most profitable. So although there is value in this approach, it isn't perfect.



MACDoug - support and resistance aren't a part of the equation.... and I don't think there is ever a guarantee that a profitable algorithm will continue, based on our speculative nature (and no knowledge of what underlying factors are moving the price) but if I am going to gamble, I'd rather do it like this.
  • #124
  • Sep 5, 2008 2:00pm Sep 5, 2008 2:00pm
The result of combining the 18/16 yen cross (w/ sl & tp) w/ the EURUSD monthly strategy (w/ trailing stop)....

30,000 pips in 20 years....

bottom graph show maximum drawdowns....

if i can find a few more high timeframe strategies that are nice and smooth over the last 20 years (say, getting the total up to 100,000 pips) then i think I may have something worth trading.
Attached Image
  • #125
  • Sep 5, 2008 4:17pm Sep 5, 2008 4:17pm
Are you familiar with the math for "margin of error" ? Since you're dealing with return series data it should be close to a normal distribution.

Linky:
http://en.wikipedia.org/wiki/Margin_of_error

So lets say you want a 99% confidence you'd have:
margin of error = 1.29/sqrt(n)

So let's say we want a margin of error no greater than 5%, just plug it in and solve for n.

Ends up slightly under 700 trades. Naturally less if you want less confidence or a higher margin of error. Still, tells you how accurate your returns should be.
  • #126
  • Sep 5, 2008 4:21pm Sep 5, 2008 4:21pm
Hi Rabid I feel like we might be cheating the system a little bit with that type of statistical analysis.... (but maybe not)

Let me explain. By taking several pairs and trading different algorithms on them, we will have several small population sizes instead of one large one.... so the analysis can't be applied to all the trades together.

Therefore, even if I have 2000 trades in 20 years, I actually have 4 to 500 trades in isolatation that must be judged seperately.

Quoting Rabid
Are you familiar with the math for "margin of error" ? Since you're dealing with return series data it should be close to a normal distribution.

Linky:
http://en.wikipedia.org/wiki/Margin_of_error

So lets say you want a 99% confidence you'd have:
margin of error = 1.29/sqrt(n)

So let's say we want a margin of error no greater than 5%, just plug it in and solve for n.

Ends up slightly under 700 trades. Naturally less if you want less confidence or a higher margin of error. Still, tells you how accurate your returns should be.
  • #127
  • Sep 5, 2008 4:27pm Sep 5, 2008 4:27pm
Quote
Let me explain. By taking several pairs and trading different algorithms on them, we will have several small population sizes instead of one large one.... so the analysis can't be applied to all the trades together.

Hrm, that's a good question. If you're using the same settings for all pairs then it probably wouldn't matter, but if you're optimizing for each pair specifically...

Still, 500 trades in isolation can give you a confidence level and a margin of error. As such, you could probably determine when a series of trades goes outside of your bounds and alert yourself accordingly that things may have changed. With that you could probably determine a maximum allowable drawdown and determine if the system is profitable even under changing conditions.
  • #128
  • Sep 5, 2008 4:30pm Sep 5, 2008 4:30pm
And how does a statistician (such as yourself) deal with market conditions changing, thus adversely effecting a characteristic that was present for 20+ years?

Quoting Rabid
Still, 500 trades in isolation can give you a confidence level and a margin of error. As such, you could probably determine when a series of trades goes outside of your bounds and alert yourself accordingly that things may have changed.
  • #129
  • Sep 5, 2008 5:14pm Sep 5, 2008 5:14pm
Well, I'm just saying that if things go outside of your set bounds you could alert yourself to perform other tests.
  • #130
  • Sep 6, 2008 5:51am Sep 6, 2008 5:51am
Quoting tdion
And how does a statistician (such as yourself) deal with market conditions changing, thus adversely effecting a characteristic that was present for 20+ years?
Tdion,

As usual, don't consider my post as being critical - I am just providing "food for thoughts", if you are not hungry or don't like that food, simply throw it away.

Backtesting is good useful exercise, but... As they say "Past performance is never a g'tee of the future one". You made good discoveries with your backtesting, now it's time to get into a gear.

I believe you've been to Bagovino thread. Have a look at Strada - he spent a year (!!) FORWARD-testing the systems (similar MA crosses), found his sweet spot in terms of the rules and (this is the MOST important part) put those rules into practice. He now routinely makes 100+ pips a day.

I am not not (at all! being provocative, but how many pips a day do you make on a consistent basis?.. Bagovino rules which Strada is using are VERY simple. The trick is in... following them 100%. If you start bend the rules even a bit, no amount of backtesting will save you.

Have a nice weekend!
One Chart is Worth Ten Thousand Words.
  • #131
  • Sep 6, 2008 12:18pm Sep 6, 2008 12:18pm
Hmmmm....

What's wrong with learning and experimenting? Did it occur to you that I might ENJOY this little project? I know your intentions were good, but....

Okay, so 100 pips a day.... those are his claims? Does he keep a live journal? Are you trading this method everyday yourself?

PM me because I don't want the topic of the thread changed.

Thanks.

Quoting Forex007
He now routinely makes 100+ pips a day.
  • #132
  • Sep 6, 2008 3:28pm Sep 6, 2008 3:28pm
Show me the pips......
Same Whore .... Different Dress
  • #133
  • Sep 6, 2008 5:26pm Sep 6, 2008 5:26pm
Tdion i aways though that frequency was one of the most important things in trading. That was confirmed by some big traders (thoose that you consider too i know).

in your findings there was not results good enough to use in intraday? Maybe not 1 min neither 5 but maybe 30 min.

thanks in advance.
  • #134
  • Sep 6, 2008 5:28pm Sep 6, 2008 5:28pm
Just another home brewed creation out of my garage operation....

Based on a EURUSD daily LWMA SMA cross given to me by my good friend johnedoe, I have put together this little baby. (john.... the strategy is VERY different than yours, but they still use the same time period cross.)

The graph is pip performance over the last 20 years... 11,000 or so....

This is not to tantalize, but to show that MA crosses have nice potential on certain pairs... and show consistency over time.
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 17 KB
  • #135
  • Sep 6, 2008 5:42pm Sep 6, 2008 5:42pm
giraia time will tell.... not enough research has been done yet for me to have anything worth saying....

Quoting giraia_br
Tdion i aways though that frequency was one of the most important things in trading.
  • #136
  • Sep 6, 2008 8:48pm Sep 6, 2008 8:48pm
Very good Tom, what are the changes you made?
Also if you look at my spreadsheet you have seen that using the strategy I use with a 100 pip SL we're over 3700 pip YTD ... +/- a small margin of error.
Same Whore .... Different Dress
  • #137
  • Sep 13, 2008 4:39pm Sep 13, 2008 4:39pm
I have read that in order to properly backtest an EA trading on a daily timeframe that one needs to use 3 years of data, but for an EA that trades on a 1 hour timeframe how many years of data should I use? I have searched all over the internet and I haven't found anything that would answer my question!!!
  • #138
  • Sep 13, 2008 8:57pm Sep 13, 2008 8:57pm
Use as much data as possible, you really can't have too many trades in a sample.
  • #139
  • Oct 1, 2008 8:22pm Oct 1, 2008 8:22pm
Here is a 30 year backtest for the GBPUSD w/ the RSI Optimization MA (I never posted it)

17,000 pips after spreads/rollover.

So that is a 3rd COMPLETELY DIFFERENT strategy for trading daily's to add to the EURUSD monthly strategy and the USDJPY cross.

If I get up to 100,000 pips with a few more strategies as consistent as these, I will trade them live.
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 20 KB
  • #140
  • Feb 2, 2009 7:21am Feb 2, 2009 7:21am
Quoting tdion
Here is a 30 year backtest for the GBPUSD w/ the RSI Optimization MA (I never posted it)

17,000 pips after spreads/rollover.

So that is a 3rd COMPLETELY DIFFERENT strategy for trading daily's to add to the EURUSD monthly strategy and the USDJPY cross.

If I get up to 100,000 pips with a few more strategies as consistent as these, I will trade them live.
Was reading your thread and since you're interested in RSI, how about a RSI corssover? Put it into your E/U D1 and see if you can use this. It also works for other TF. You'll need 2 of these in different TF. Lower for entry. Higher for direction.
Attached File(s)
File Type: tpl tdion.tpl   3 KB | 350 downloads
0 traders viewing now