Trend Champ | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Are you a trend trader or a counter-trend trader? 29 replies

Ichi Trend & RSI Trend Catch 4 replies

trend up trend down 1 reply

Automatic Trend line, trend channel, support, resistance detection 3 replies

"Trend" , how do you actually trace the current trend ? 23 replies

Attachments: Trend Champ
Exit Attachments

Trend Champ

  • #1
  • First Post: Sep 17, 2008 3:55am Sep 17, 2008 3:55am
Hi Guys!

I have finished writing the first version of an EA based on the Turtle trading system, with a MACD + Stoch filter. I will not explain the Turtle trading system, since its already well known. My filters on top of the breakout entry signals are (1) H1 MACD(5,13,1): check if th last 6 bars are positive for a long entry, or negative for a short entry, AND (2) M15 Stochastic(5,3,3): for long signals, entry when stoch crosses 30 up; for short signals, entry when stoch crosses 70 down. Thats all!

If you have read about the Turtle trading, the emphasis is on discipline and following rules exactly. So I thought what better way to implement this strategy than automated trading. I basically wanted an EA that I can put on a small account, say $500, and let it trade without any manual intervention. I have finished the first version of my system, and the backtest results look great! I'm attaching the report from strategy tester. The test was ran from Jan 1 2008 to Sept 16 2008. It was started on initial deposit of $500, and was able to produce net profit of $19k. It produced 83.33% winning trades.

Thanks, and good luck!
Attached Image (click to enlarge)
Click to Enlarge Name: TrendChamp.gif Size: 8 KB
Attached File(s)
File Type: zip TrendChamp.zip   31 KB | 1,167 downloads
  • #2
  • Sep 17, 2008 4:41am Sep 17, 2008 4:41am
Quoting BitShifter
Hi Guys!

I have finished writing the first version of an EA based on the Turtle trading system, with a MACD + Stoch filter. I will not explain the Turtle trading system, since its already well known. My filters on top of the breakout entry signals are (1) H1 MACD(5,13,1): check if th last 6 bars are positive for a long entry, or negative for a short entry, AND (2) M15 Stochastic(5,3,3): for long signals, entry when stoch crosses 30 up; for short signals, entry when stoch crosses 70 down. Thats all!

If you have read about the Turtle trading, the emphasis is on discipline and following rules exactly. So I thought what better way to implement this strategy than automated trading. I basically wanted an EA that I can put on a small account, say $500, and let it trade without any manual intervention. I have finished the first version of my system, and the backtest results look great! I'm attaching the report from strategy tester. The test was ran from Jan 1 2008 to Sept 16 2008. It was started on initial deposit of $500, and was able to produce net profit of $19k. It produced 83.33% winning trades.

Thanks, and good luck!
Those must be pretty good filters =D

The original turtle trading system I believe had a 30% win rate.
  • #3
  • Sep 17, 2008 5:00am Sep 17, 2008 5:00am
Good job! When can we begin forward-testing it?
  • #4
  • Sep 17, 2008 5:07am Sep 17, 2008 5:07am
Which Turtle Trading thread in here did you utilize for your system?
  • #5
  • Sep 17, 2008 6:38am Sep 17, 2008 6:38am
Hi BitShifter,

It is very good backtest result. Did you do some forward tests your system and EA too ? What were results of forward test ?

We would like to test your system (maybe EA, if you attach a demo version), can you give us a more detailed description about your trading system and EA ?

Best regards,

Viatoris
  • #6
  • Sep 17, 2008 7:24am Sep 17, 2008 7:24am
what happens if you test over the last five years with the same ea on the same settings?
Also why is your modelling quality so low?
Max
Gravity goes both ways in forex...
  • #7
  • Sep 17, 2008 12:55pm Sep 17, 2008 12:55pm
1) Its using the original Turtle system, I read this in a book. I am not aware if there is any forum thread on it. Perhaps there is, I haven't searched. But you can find a lot of info about Turtle trading on the web if you do a quick search.

2) As I mentioned earlier, I have "just" finished developing the EA. And by "just" I mean last night I have set it up on a demo account for forward testing now. Lets see what happens. One thing you will notice from the back testing results is that it doesn't take a crap load of trades. It waits for a trend to form and then try to get on it. That is the reason why its success rate is more than 80%.

3) I have back tested to as long as Jan 1 2007. Any time before that doesn't work with my broker account, I don't think they have the data. Even if I use dates before Jan 1 2008, the tick modeling % drops severely, so the results are not accurate. Basically, the results I have shown are from the longest time period I can take without compromising the test accuracy too much.

4) I also did incremental back testing (i.e. start with Jul 2008 to Sept 2008, then move back a couple of month: Apr 2008 to Sept 2008, etc), and the results are all great. As you can see, I am starting with an initial deposit of only $500. This is considered extremely low capital to trade with. But even with that, the account is never wiped out or gotten close to it. Yet the system still trades aggressively enough that it can grow a $500 account to $20k in less than a year! I attribute this to the very good trading volume selection strategy of the Turtle system. I should mention though, that one other thing I have changed from the Turtle system is the stop loss. I have fixed that at 150 pips.
  • #8
  • Sep 17, 2008 2:39pm Sep 17, 2008 2:39pm
your backtest is insufficient.

the lack of data is easily solved this way:

from historical center, select download data, then for every pair, u may usually download data until 2006 i think.

otherwise, ur backtest period is too short unless you are doing a 5 min/15 min, tens of trades/day system.

for example, the two loss at the top, how do you make sure you won't meet these kind of loss consecutively? if in a month, you have 3 of that kinds of trades, will that blow up the capital?

that's one of major reasons that I guess you need at least 1 more backtesting year.


Quoting BitShifter
3) I have back tested to as long as Jan 1 2007. Any time before that doesn't work with my broker account, I don't think they have the data. Even if I use dates before Jan 1 2008, the tick modeling % drops severely, so the results are not accurate. Basically, the results I have shown are from the longest time period I can take without compromising the test accuracy too much.
  • #9
  • Sep 17, 2008 2:51pm Sep 17, 2008 2:51pm
I have already tried updating the data from the History Center, but I keep getting No Updates Available. If you have longer, and more accurate history dataset, send it to me and I will run the EA on it.

As for your comment about blowing up the capital, there are two reasons:

1) Look at the consecutive losing trades. Over a year of testing it only has 1 consecutive loss at any time. Of course its possible that this figure COULD change if some major screw up happens in past years that have not been tested, but that argument can be applied to any system. If there are 'x' consecutive losses, the capital will be blown. If you consider this year alone for the EURO, I believe it has sampled a lot of the market conditions. Euro was on serious up trend until July, where it started falling. Last year, the trend was mostly up. So if you think about it, the extreme fluctuation has already been accounted for in the testing period. And even in that, it is showing a success rate of 80%. I seriously doubt that is a fluke.

2) Also keep in mind that the volume is not fixed. Its computed based on free margin. So in the case you mentioned where you suffer 3 consecutive losses, the volume would be reduced every loss, and hence account will not be gone within 3 hits.
  • #10
  • Sep 17, 2008 4:27pm Sep 17, 2008 4:27pm
thanks for explaination. please understand that i didn't mean that your account WILL blow up for 3 loss. I just say a random situation, potentially.

as to the data, it is too large to send, i have 2005 till now.

i may test your EA, if possible.

Quoting BitShifter
I have already tried updating the data from the History Center, but I keep getting No Updates Available. If you have longer, and more accurate history dataset, send it to me and I will run the EA on it.
  • #11
  • Sep 17, 2008 4:58pm Sep 17, 2008 4:58pm
Can you run a backtest on any EA with your historical data and see what the tick modeling accuracy is? I can also go as back as Jan 1 2005 on my account, but the accuracy drops to 30%.

Also, what broker are you using? Perhaps I can just download a terminal from the same broker and get the history data from there.

I am currently pondering on what course of action I will take with regards to releasing this EA. However, the system is exposed and in front of you if you want to do manual back testing perhaps, or write your own implementation. But I would be more than glad to do the tests on my EA as long as I am provided the input data.
  • #12
  • Sep 17, 2008 5:13pm Sep 17, 2008 5:13pm
90% is my quality from 2006, didn't bother to check the others.

mine is FxPro.

I understand that every developer value his own product a lot.

however, imHUMBLEo, your system has a long way ahead.

I have tons of EAs produce 2~4 times return in 0.5 yr or 1 yr, but dare you to use it, if i tell u for 2~3 years they will burn out?

good luck.
  • #13
  • Sep 17, 2008 5:44pm Sep 17, 2008 5:44pm
Depends on how he optimized it. That would determine how robust the EA is and consequently how well it will perform in the market.
  • #14
  • Sep 17, 2008 5:59pm Sep 17, 2008 5:59pm
Well, I've developed many EAs, so I really don't have any special "sentimental" attachment to this one, if that is what you are suggesting.

But yes, there are EAs that can produce much higher profits than this one. I can write an EA that takes random trades, and plug in some crazy martingale money management and trade volume adjustment scheme, and it could possibly turn $500 into $100k. But thats not the point. The point is to show consistent wins. If you can show me EAs that have winning rates 2 ~ 4 times better, I will be very interested in seeing that.

Secondly, if there was a static EA out there who consistently made $40k to $80k on a $500 capital in less than a year, as you are suggesting, then I don't think anyone would be doing anything else... theyw ould all just use this EA and live in peace. The point here is every EA has its days. The best you can do is try to make it adapt with current and near past market conditions. History is all you have to make your decision. I am not sure how you are making a prediction by saying "would you use it if I told you they will burn out in 2 to 3 years". Well, here's my response to that: "would you use it if I told you it will earn you a million in 2 to 3 years". Of course, both claims are absolutely baseless. We don't know what will happen in the future. It could go either way. Maybe it will bomb out, or maybe it'll just grow even faster. Who knows. All we can do is make a decision based on performance on past data.

Lastly, I would like to say there is no "optimization" involved. I have laid out exact system rules. What exactly are you thinking can be optimized here? There arn't any variable parameters or thresholds. Everything is fixed. The s/l is fixed, the t/p is fixed, and the trading volume is computed based on free margin and market volatility as it is clearly defined in the Turtle system.
  • #15
  • Sep 17, 2008 6:01pm Sep 17, 2008 6:01pm
Quoting BitShifter
Lastly, I would like to say there is no "optimization" involved. I have laid out exact system rules. What exactly are you thinking can be optimized here? There arn't any variable parameters or thresholds. Everything is fixed. The s/l is fixed, the t/p is fixed, and the trading volume is computed based on free margin and market volatility as it is clearly defined in the Turtle system.
MACD and Stochastics, those aren't standard values as far as I know, and how did you come to choose the timeframes that each uses?
  • #16
  • Sep 17, 2008 6:11pm Sep 17, 2008 6:11pm
Bit, I believe that u r a better/more experienced developer.

what i suggested is just that you need more period to determine the quality, and i believe that 8 months is anyway NOT enough.

see ur own words, u mentioned "a year", you didn't say 2008, right? not "the year", but "a year", that's what we look for.

in fact, time doesn't matter here. Ultimately, number of trades matters, if this is 200 trades records, you should be congradulated.


Quoting BitShifter
Secondly, if there was a static EA out there who consistently made $40k to $80k on a $500 capital in less than a year, as you are suggesting, then I don't think anyone would be doing anything else... theyw ould all just use this EA and live in peace. The point here is every EA has its days.
  • #17
  • Sep 17, 2008 6:26pm Sep 17, 2008 6:26pm
the MACD and stoch numbers I determined on my own. Well not quite. 6 bars MACD is generally taken when looking for trade patterns in MACD based systems. Similar thing in CCI based systems, they look at last 6 bars (see Woodie's system). As for the stochs, again, I've seem them in various systems. Perhaps they can be optimized, but in the system as I presented, they are not variables. Perhaps optimizing them could give better results, I don't know. The stochs value is probably still a bit of handwaving, I cannot justify those numbers. But the 6 bars on MACD, that is common practice from what I've seen.

I am not trying say that I should only look at one fixed time period and declare the system a winner based on that. I'm saying I "want" to do more testing, thats why I was asking you for your history data. Unfortunately, on the terminal I have, the result from testing back to Jan 1 2007 looks almost identical to Jan 1 2008, which leads me to believe the EA can't find any signals in the older history because of the lack of data.
  • #18
  • Sep 17, 2008 8:12pm Sep 17, 2008 8:12pm
I finally got a chance to read the Turtle Trading manual tonight. Your entries and exits based on your statement does not look like the Turtle method based on 20 days high and low. I was wondering if your using a different Turtle system besides what I have reviewed. Are there different Turtle entry methods out there and you created your own Turtle entry and exit rules?

By the way, I think the best way to test-drive your EA is by forward-testing. More traders are aware of back-testing problems so live forward-testing is the way to go.

Good luck!
  • #19
  • Sep 17, 2008 10:52pm Sep 17, 2008 10:52pm
Where can I find the turtle trading method? I've been looking on google with no success
  • #20
  • Sep 30, 2008 8:06pm Sep 30, 2008 8:06pm
Well today is the last day of the month, so I figured I should post the forward tests results so far.
Attached Image (click to enlarge)
Click to Enlarge Name: TrendChamp.gif Size: 5 KB
Attached File(s)
File Type: zip TrendChamp.zip   2 KB | 410 downloads
0 traders viewing now