Old Dog with New Tricks | Page 40 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Old Dog's Taming of the Beast? 34 replies

Old Dog Fib Trading - EA Development 18 replies

Old Dog Fib Trading - EURUSD 399 replies

5 Tricks of Trend Trading, by Rob Booker 25 replies

How To Work Around A Market Maker's Tricks 0 replies

Attachments: Old Dog with New Tricks
Exit Attachments

Old Dog with New Tricks

  • #781
  • Aug 13, 2010 7:39am Aug 13, 2010 7:39am
Quoting stevegee58
Frankly I've been running around lately and haven't had a chance for a sit-down with the csv files that the EA generates. I'm hoping to have a look this evening after trading stops. It could very well be that I've tightened things to the point that there will never be any trades.
I have managed to get one or two hours of trading on EU with win rate 50 and success score 0.8. Anytime I try a higher win rate it blocks out all trading. At least the success score ensures my risk reward is well above 1.

Paul
  • #782
  • Aug 13, 2010 7:51am Aug 13, 2010 7:51am
Thanks Paul for sharing this.
But if you have only one or two hours where trades can be trigerred, the EA can spend a month before finding a trade opportunity.
But as you said your success score statistacally ensures that you will end up by winning.
So maybe the best thing to do is to run the EA on as much pairs as possible with the highest statistical sucess score and increase the lot size.

Quoting paulbt38
I have managed to get one or two hours of trading on EU with win rate 50 and success score 0.8. Anytime I try a higher win rate it blocks out all trading. At least the success score ensures my risk reward is well above 1.

Paul
  • #783
  • Aug 13, 2010 8:46am Aug 13, 2010 8:46am
Quoting Mindsphere
Ronald: Does the optimizer take into account spread at all when making its assessment? If so is it possible to add an option which lets us set the spread as it may vary at different times of the day. If it's not included, any chance you could add this feature, as the spread could heavily change profit/loss calculations.
... yes, that would be great ... I did mentioned about spread filter somewhere earlier in the thread, ... Ron what do you think about it ???
  • #784
  • Aug 13, 2010 9:48am Aug 13, 2010 9:48am
Quoting bluemele
i have had nothing but losses all week.
If your bar count was relatively low, then that would be expected. Why? This week was the first down week in awhile.

As for those asking about spread, yes, I'm working on putting in a more realistic spread model to put in. Sunday spreads tend to be wider than normal spreads.
  • #785
  • Aug 13, 2010 9:56am Aug 13, 2010 9:56am
To give an update on what I ran for the week.
4 pairs (U/J seems to still be the best)
Had MM set to True, risk of 1
Turned off the scrolling RR during opto to speed things up a bit.
50K bars to opto
70 min success score up from the default of 50.

Start off the week well...then had 19 losses in a row to blow out all the gain...and take back some of the gains from last week.

Last night around 6pm est. I lowered the max trades at once to 5. Didn't get any trades between last night and this morning. Turned it off for the weekend around 8am est to give my pc a break.

I'm tempted to just try it on U/J with more optimization the beginning of next week
  • #786
  • Aug 13, 2010 9:56am Aug 13, 2010 9:56am
Quoting Ronald Raygun
If your bar count was relatively low, then that would be expected. Why? This week was the first down week in awhile.

As for those asking about spread, yes, I'm working on putting in a more realistic spread model to put in. Sunday spreads tend to be wider than normal spreads.
Yes, my CPU's (pretty much have a super computer) all the instances of MT4 live, with all the currency pairs, while trading futures using tradestation and ATC brokers interface. Computer gets a little bogged if I have anything much over 50K bars..

It was working beautifully last week when I had it just on one instance and 300K-400K bars... I guess that is the trick...
We are our own best indicator.
  • #787
  • Aug 13, 2010 4:47pm Aug 13, 2010 4:47pm
Quoting networker75
When I try to be more selective on EU (Min Success Rate=55), the EA is not able to find any winning combination. By looking at the optimized settings CSV file I can see all the values to -1.
Maybe my historical data is not acurate, or that just the HiLo strategy didn't perform historically well on this EU pair.

were you able to find winning combinations with the below settings on EU?
You might have missed this earlier but min successrate needs to be set with a decimal point - i.e. 0.55 = 55%

Ronald said it shouldn't be this way, but according to my testing it works.
  • #788
  • Aug 13, 2010 4:50pm Aug 13, 2010 4:50pm
Quoting Ronald Raygun
If your bar count was relatively low, then that would be expected. Why? This week was the first down week in awhile.

As for those asking about spread, yes, I'm working on putting in a more realistic spread model to put in. Sunday spreads tend to be wider than normal spreads.
Thanks Ron, good to hear

One other question that may have been asked earlier in the thread - does the optimized back test take into account the max number of trades setting? e.g. Will it take all 10 trades during a breakout hour if the breakout continues on each consecutive m5 bar, or does it wait until the first trade is closed? If so, might we be skewing our own results by using a max trades of 10 when the backtest uses the equivalent of 1 max trade?
  • #789
  • Aug 13, 2010 5:49pm Aug 13, 2010 5:49pm
Thanks Mindsphere.
I Will try a SuccessRate with a decimal value.

I still have 2 questions, maybe someone can help :
SucessRate is a combined metric of Win Rate and Risk Reward, so I think it is better to set only the value of the minimum SuccessRate (and not the win rate) and optimize the results according to SuccessRate also.
Is it redundant to proceed this way?

What is the difference in terms of performance between the two following settings:
A- 0.1 lot with a maximum of 10 trades
B- 1 lot with a maximum of 1 trade

I tend to say that scenario A is safer, but I am not sure...

Quoting Mindsphere
You might have missed this earlier but min successrate needs to be set with a decimal point - i.e. 0.55 = 55%

Ronald said it shouldn't be this way, but according to my testing it works.
  • #790
  • Aug 13, 2010 7:19pm Aug 13, 2010 7:19pm
Quoting Mindsphere
Thanks Ron, good to hear

One other question that may have been asked earlier in the thread - does the optimized back test take into account the max number of trades setting? e.g. Will it take all 10 trades during a breakout hour if the breakout continues on each consecutive m5 bar, or does it wait until the first trade is closed? If so, might we be skewing our own results by using a max trades of 10 when the backtest uses the equivalent of 1 max trade?
The backtest takes all trade opportunities regardless of whether or not there are other simultaneous trades open. The reason why is in the nature of how the TP/SL optimization works as I explained in the webinar.
  • #791
  • Aug 13, 2010 7:26pm Aug 13, 2010 7:26pm
Well I finally got around to looking at the files that the EA generates. The files named "HiLo Self Optimizing XXXXXXm Optimized Settings.csv" contained all -1 values.

I guess that means no trades were found in the 300,000 bars over which I had it optimize with my criteria? (65% win rate, 1.5 reward:risk)
You are in a maze of twisty little passages, all alike.
  • #792
  • Aug 13, 2010 7:46pm Aug 13, 2010 7:46pm
The easiest way I found to get a success score over 100 is to lower the win rate and increase risk reward. Next week I'm going to test a win rate of 30% and R/R of 4. This should give me a success score of 120. 12 pairs and 250,000 bars. Lets see if my results fall in line with statistics...
  • #793
  • Aug 13, 2010 7:59pm Aug 13, 2010 7:59pm
Quoting stevegee58
Well I finally got around to looking at the files that the EA generates. The files named "HiLo Self Optimizing XXXXXXm Optimized Settings.csv" contained all -1 values.

I guess that means no trades were found in the 300,000 bars over which I had it optimize with my criteria? (65% win rate, 1.5 reward:risk)
Doesn't look like it.

What does All Permutations say?
  • #794
  • Aug 13, 2010 8:01pm Aug 13, 2010 8:01pm
Quoting Ronald Raygun
The backtest takes all trade opportunities regardless of whether or not there are other simultaneous trades open. The reason why is in the nature of how the TP/SL optimization works as I explained in the webinar.
Interesting... If that is the case, is it possible that extremely long trends, over consecutive m5 candles for example, could skew the backtest results?

i.e. if it takes 1 trade per bar, and there are situations where the trend continues slowly for a couple of hundred pips on a given day, triggering a breakout on m5 every 15 minutes or so, it's possible that the backtest system will over-weight that day, thus rendering the backtest less reliable as a predictor of future moves.

On that point, is there a way to get the EA to draw the trades that its calculated? Might be a little much for a 1million strong bar chart, however
  • #795
  • Aug 13, 2010 8:41pm Aug 13, 2010 8:41pm
Quoting Ronald Raygun
Doesn't look like it.

What does All Permutations say?
It's full of numbers that aren't -1. What's the interpretation for those records?
You are in a maze of twisty little passages, all alike.
  • #796
  • Aug 13, 2010 10:06pm Aug 13, 2010 10:06pm
Quoting stevegee58
It's full of numbers that aren't -1. What's the interpretation for those records?
The column headings are the same as the optimized settings file. Difference is this is every possible permutation. If you see something there that you like, replace the corresponding hour with the new line.
  • #797
  • Aug 13, 2010 10:08pm Aug 13, 2010 10:08pm
Quoting Mindsphere
Interesting... If that is the case, is it possible that extremely long trends, over consecutive m5 candles for example, could skew the backtest results?

i.e. if it takes 1 trade per bar, and there are situations where the trend continues slowly for a couple of hundred pips on a given day, triggering a breakout on m5 every 15 minutes or so, it's possible that the backtest system will over-weight that day, thus rendering the backtest less reliable as a predictor of future moves.

On that point, is there a way to get the EA to draw the trades that...
Yes, it's possible. Take a look in the master copy file and see how frequently trades happen.

That's largely why I backtest against longer periods. I'm more likely to catch both a long up trend and a long downtrend.
  • #798
  • Aug 13, 2010 10:44pm Aug 13, 2010 10:44pm
Quoting Ronald Raygun
Yes, it's possible. Take a look in the master copy file and see how frequently trades happen.

That's largely why I backtest against longer periods. I'm more likely to catch both a long up trend and a long downtrend.
I see... So the idea is that by taking an extremely large sample we correct for such possible discrepencies? If that's the case can you possibly recommend a good source for accurate historical data that goes back further than 2001? The free source quoted throughout this thread gives me only about 700,000 bars on m5.

Thanks again!
  • #799
  • Aug 13, 2010 11:05pm Aug 13, 2010 11:05pm
Quoting Mindsphere
I see... So the idea is that by taking an extremely large sample we correct for such possible discrepencies? If that's the case can you possibly recommend a good source for accurate historical data that goes back further than 2001? The free source quoted throughout this thread gives me only about 700,000 bars on m5.

Thanks again!
Are we to go off the M1 or M5, as I have not been keeping up w/ this thread, but when I do M1 data, I can import about 380K records. Then, when I reboot my MT4, it goes back to 83K bars???

Anyone know why this might happen?

I have max bars set in history... I used to have 300K bars, but now it just keeps going back...
We are our own best indicator.
  • #800
  • Aug 13, 2010 11:13pm Aug 13, 2010 11:13pm
Quoting Mindsphere
I see... So the idea is that by taking an extremely large sample we correct for such possible discrepencies? If that's the case can you possibly recommend a good source for accurate historical data that goes back further than 2001? The free source quoted throughout this thread gives me only about 700,000 bars on m5.

Thanks again!
Why are you using M5?
0 traders viewing now