Old Dog with New Tricks | Page 35 | 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

  • #681
  • Aug 5, 2010 10:33pm Aug 5, 2010 10:33pm
Ronald,

Good work on the Webinar. Very informative.

We are our own best indicator.
  • #682
  • Aug 6, 2010 12:44am Aug 6, 2010 12:44am
I am going to update My version soon. Those who use it, please let my know whether you would prefer RR ENC, or Steve"s ENC. I want to know what you think.
  • #683
  • Edited 5:01am Aug 6, 2010 4:48am | Edited 5:01am
Thanks for the further clarification Ronald. I am, however, a little perplexed as to why simple trading strategies should increase the optimisation time greatly. Currently, the EA just looks at trades when the high/low is breached. A simple MA cross would just look at trades when one MA crosses the other - surely a very quick calculation. I'm just saying that it might be worthwhile exploring the same backtest approach on different simplistic trading logic to see if it can provide an edge.

I also think that you should not feel limited by the lowest common denominator for processing power. Even if it takes 10-20 minutes to run the calculation for one pair on slower machines, it's only run once a week. That's a very small percentage of the actual trading time to be sure. On this point, perhaps with a more long-term outlook I'm wondering if you have any plans to implement some sort of cloud system whereby optimization data could be shared across willing clusters of users?

Finally, I've been having weird issues with utilising the EA. On some occasions it seems to work just fine, while on others the EA shows the following. Default settings except for ECN true, minimum success rate = 55. Or is this expected behaviour? If so, what does this actually mean?
Attached Image
  • #684
  • Aug 6, 2010 5:00am Aug 6, 2010 5:00am
Quoting Mindsphere
Thanks for the further clarification Ronald. I am, however, a little perplexed as to why simple trading strategies should increase the optimisation time greatly. Currently, the EA just looks at trades when the high/low is breached. A simple MA cross would just look at trades when one MA crosses the other - surely a very quick calculation. I'm just saying that it might be worthwhile exploring the same backtest approach on different simplistic trading logic to see if it can provide an edge.

I also think that you should not feel limited by the lowest...
I've had the same before due to a small gap in my historical data. It may be worth checking if your data history doesn't have any gaps.
  • #685
  • Aug 6, 2010 5:13am Aug 6, 2010 5:13am
Actually now that I look at the screenshot I took earlier during the webinar of your data, I see that you the first few entries also have -1, I guess this means insufficient data?

Looking at that screenshot has further puzzled me, however. If you are working off 1.5 million bars of prior data (on M5 I'm assuming), how can there only be 98 trades at 5am? Ahh.. wait a minute, are the "number of trades in sample" affected by whatever you set the minimum success rate to? I note in my data I have something like 1529 trades on 100,000 bars at 5am for EURUSD.

So, if I understand correctly, it will display -1 for all values depending on how stringent you set your minimum successrate or winrate. Furthermore, in the recorded CSV, the "number of trades in sample" will be reduced accordingly.

Sorry for all the thinking out loud, I'm just trying to work it all out. Hope I didn't confuse anyone
Attached Image (click to enlarge)
Click to Enlarge Name: why2.png Size: 38 KB
  • #686
  • Aug 6, 2010 5:46am Aug 6, 2010 5:46am
Quoting Ronald Raygun
The relation between win rate and risk reward is inverse within the same success score. Success score of 50 could mean a win rate of 50 and a risk reward of 1. It could also mean a win rate of 90 and a risk reward of 1.8.
I may have picked this up wrong but on any tests I have run so far I had thought the success score was simply calculated by multiplying win rate by risk/reward. In which case 0.9 * 1.8 would give a success score of 1.62 and not 0.5.

Paul
  • #687
  • Aug 6, 2010 7:00am Aug 6, 2010 7:00am
Quoting Mindsphere
Thanks for the further clarification Ronald. I am, however, a little perplexed as to why simple trading strategies should increase the optimisation time greatly. Currently, the EA just looks at trades when the high/low is breached. A simple MA cross would just look at trades when one MA crosses the other - surely a very quick calculation. I'm just saying that it might be worthwhile exploring the same backtest approach on different simplistic trading logic to see if it can provide an edge.
When writing the trading logic within the EA's backtester, I need to write it in such a way that I can calculate all the necessary variables as if the EA were just started that bar. It's not always easy to do that for more complex systems.

Quoting Mindsphere
I also think that you should not feel limited by the lowest common denominator for processing power. Even if it takes 10-20 minutes to run the calculation for one pair on slower machines, it's only run once a week. That's a very small percentage of the actual trading time to be sure. On this point, perhaps with a more long-term outlook I'm wondering if you have any plans to implement some sort of cloud system whereby optimization data could be shared across willing clusters of users?
I have a version of this EA which optimizes the Range-based breakout. It's basically the london breakout style system. That--by my estimates--takes something on the order of 122 days to optimize. Why? There are 4.14M permutations to handle. The core system of this EA (HiLo) is so very simple that the backtest runs incredibly fast.

Quoting Mindsphere
Finally, I've been having weird issues with utilising the EA. On some occasions it seems to work just fine, while on others the EA shows the following. Default settings except for ECN true, minimum success rate = 55. Or is this expected behaviour? If so, what does this actually mean?
When the system settings don't make sense (TP of -1 is a good example) the EA couldn't find any settings that met your criteria, so the optimizer set the EA using impossible settings so it doesn't trade.

Quoting Mindsphere
Actually now that I look at the screenshot I...
The EA that sample came from isn't the one being used on my live account. That terminal is where I do most of my debugging/testing.

Quoting paulbt38
I may have picked this up wrong but on any tests I have run so far I had thought the success score was simply calculated by multiplying win rate by risk/reward. In which case 0.9 * 1.8 would give a success score of 1.62 and not 0.5.

Paul
You're right. It should be 0.9 * 0.555555 (5/9)
  • #688
  • Aug 6, 2010 8:09am Aug 6, 2010 8:09am
Quoting Ronald Raygun
When the system settings don't make sense (TP of -1 is a good example) the EA couldn't find any settings that met your criteria, so the optimizer set the EA using impossible settings so it doesn't trade.
Say I enter say a MinimumWinRate of 60%, MinimumRiskReward of 1.5 and Optimize Based On HighestSuccessScore and the optimizer does its thing and returns a TP of -1. Should I keep trying different settings until it returns a result that makes sense?
You are in a maze of twisty little passages, all alike.
  • #689
  • Aug 6, 2010 8:31am Aug 6, 2010 8:31am
Quoting stevegee58
Say I enter say a MinimumWinRate of 60%, MinimumRiskReward of 1.5 and Optimize Based On HighestSuccessScore and the optimizer does its thing and returns a TP of -1. Should I keep trying different settings until it returns a result that makes sense?
Not at all. TP of -1 means the EA couldn't find any settings that match your criteria. It basically means "Stand Aside"
  • #690
  • Aug 6, 2010 9:00am Aug 6, 2010 9:00am
Quoting Ronald Raygun
Not at all. TP of -1 means the EA couldn't find any settings that match your criteria. It basically means "Stand Aside"
So as long as the settings aren't otherwise unreasonable, just set it and let the EA do its job?

I'm assuming my hypothetical settings are reasonable...
You are in a maze of twisty little passages, all alike.
  • #691
  • Aug 6, 2010 9:08am Aug 6, 2010 9:08am
Quoting stevegee58
So as long as the settings aren't otherwise unreasonable, just set it and let the EA do its job?

I'm assuming my hypothetical settings are reasonable...
Pretty much.

You can set what you want the EA to achieve and if the EA finds a scenario that matches, it will trade that scenario.
  • #692
  • Aug 6, 2010 9:14am Aug 6, 2010 9:14am
Quoting Ronald Raygun
Pretty much.

You can set what you want the EA to achieve and if the EA finds a scenario that matches, it will trade that scenario.
Ahh! This has helped clear some things up for me I think as well.
  • #693
  • Aug 6, 2010 9:18am Aug 6, 2010 9:18am
Replays are up.

http://ronaldraygunforex.com/WebinarReplay/

Webinar 2 should be done in 40 mins. So pls wait until that finishes.
  • #694
  • Aug 6, 2010 9:34am Aug 6, 2010 9:34am
Quoting Ronald Raygun
Replays are up.

http://ronaldraygunforex.com/WebinarReplay/

Webinar 2 should be done in 40 mins. So pls wait until that finishes.
Great! Thanks!
  • #695
  • Aug 6, 2010 10:25am Aug 6, 2010 10:25am
Quoting Ronald Raygun
Pretty much.

You can set what you want the EA to achieve and if the EA finds a scenario that matches, it will trade that scenario.
ok, that clears one of the issues for me too. If I set let's say winrate55 r/r 1 success 60 it shows me negative values (TP-1) SO that means it stands aside till it finds this minium conditions so it can start to trade.
  • #696
  • Aug 6, 2010 10:38am Aug 6, 2010 10:38am
Quoting iulli
ok, that clears one of the issues for me too. If I set let's say winrate55 r/r 1 success 60 it shows me negative values (TP-1) SO that means it stands aside till it finds this minium conditions so it can start to trade.
If it doesn't find the minimum condition to trade...how often is it looking for it again?
  • #697
  • Edited 11:14am Aug 6, 2010 11:03am | Edited 11:14am
Quoting unknown2u
If it doesn't find the minimum condition to trade...how often is it looking for it again?
and is it possibile to make ea show some values (best real values instead of negative values) even if it does not meet the condition to trade? That way we can visually what works most of the time and/or highest values possibile and edit our settings acordingly.
Attached Image (click to enlarge)
Click to Enlarge Name: for today.gif Size: 29 KB
  • #698
  • Aug 6, 2010 11:46am Aug 6, 2010 11:46am
Quoting unknown2u
If it doesn't find the minimum condition to trade...how often is it looking for it again?
Once a week

Quoting iulli
and is it possibile to make ea show some values (best real values instead of negative values) even if it does not meet the condition to trade? That way we can visually what works most of the time and/or highest values possibile and edit our settings acordingly.
That seems counter-intuitive to how the EA works. EA basically looks for what you define as your holy grail.
  • #699
  • Aug 6, 2010 1:10pm Aug 6, 2010 1:10pm
Quoting Ronald Raygun
Once a week
I guess then it would make sense not to leave settings in the "take a pass" mode as it won't look again for that possible situation for a full week....thus if I'm understanding correctly there would not be any trades for a whole week??
  • #700
  • Aug 6, 2010 1:18pm Aug 6, 2010 1:18pm
Quoting unknown2u
I guess then it would make sense not to leave settings in the "take a pass" mode as it won't look again for that possible situation for a full week....thus if I'm understanding correctly there would not be any trades for a whole week??
During that specific hour of the day.
1 trader viewing now