Settings Files - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
More optimization
I should have thought of leaving out last week. But I can redo the runs without last week just for good measure, if it is essential. Should we suppose that next week is going to be anomalous, given the holiday period?
Perhaps it would be better to wait until January to do forward comparisons.
Mode 1 optimized
For what it's worth, here is the report on the latest round of optimization. Settings seem to yield a lot of loot, but I don't know if it is mere curve-fitting or genuine profitability. For comparison, I include the report for the week prior to the 3-month test period. Nothing to get excited about.
Thank you Daraknor for a quick reply do you sleep some times
I really like how you handle these threads and thank you for your quick reply.
For my stealth mode question, i am no programmer but i thought that all you had to do was ad a command, when the order is triggered, for a sell stop loss saying something like ( don't send unless bid or ask is at or below stop loss ) But like i said i am no programmer. But by your answer i see that it is a bit more complicated.
My second question would be to Cockeyedcowboy would you care to elaborate on how you would proceed in testing with random number generator, i find this very intriguing. I a have looked it up thru Google, read thru a couple of articles and i am still trying to understand how it would be done in Metatrader and some of it, but if you would kindly show me how to implement it, i am willing to try to work with it.
i am one of those people that need to see how it would be set and work to understand at least the principles.
Thank you all for your time and Happy Holidays
USDJPY, Mode3, changed Signal2
I was trying to learn (and I failed lol), how to optimize, so i was playing with different settings and I got a little better backtesting results by changing Signal2:
- SMAPeriod = 10
- SMABars = 30
Pair USDJPY
Mode 3
After this change, Phoenix demo placed buy order and all three trades become profitable. Probably just luck lol
Anyway, try it out if you wish.
Mario
EURGBP settings anyone?
I have been casting about for settings in order to do optimizing for EURGBP, but can't seem to find anyone who has ventured onto this terrain. Is this pair uninteresting from the point of view of trading, or is there another reason for avoiding it?
If anyone can suggest settings for a starting point I would appreciate the gesture.
Thanks,
Autumn
Here is a file that I used to come up with some new settings. It is taking about 8 hours to complete the optimization. I have Genetic Algorithm turned on, which cuts the tested options from 9.5 billion to 10,500.
I also chop off the testing if the results fall under 80% of my original balance. Hopefully a series of tests will allow me to trim some of the permutations down, and decrease testing time or increase our range of potentially profitable condition. Hopefully I can leave the system running and come up new settings for most currencies.
I noticed that there are 9 threads running concurrently on this machine, which means a dual processor machine may cut the time down significantly. If anyone has dual core, I'd be interested to see if setting one processor only increases the total estimate time.
Long story short, using this may yield new interesting results. Optimization guide is coming.
I have some new 'best settings' for USDCHF. Data stream is FXDD.
The results were gathered using Mode 1 with 10 max trades, signal count=20
U_Mode12_MaxTrades=10 U_Mode12_ConsecutiveSignals=20 (in the new lingo)
This is done because I optimized the entry signal, not the mode. By optimizing entry signal, all modes should have more reliable trading. It is hard to accurately optimize 5 trades. The settings above executed 80-90 trades compared to the normal 20 trades for Mode1 over the same time period. (Nov 20 to Jan 13 total.) Having 4 times as many trades means greater accuracy in optimizing the signal.
USDCHF is backtested Nov 20 to Jan 13. Summary results on $10,000 demo.
Mode 1 standard settings and 1 max trade (20 total) profit=$100.51
Mode 3 standard settings 39/3 trades total, profit = $3903.17
Mode 1 standard settings and 10 max trades (41 total) profit=$2128.77
Mode 1 best settings and 1 max trade (22 total) profit = $2092.16
Mode 3 best settings 72/3 trades total, profit = $179.03
Mode 1 best settings and 10 max trades (153 total) profit=$15,751.
(Standard settings had a TP of 110, best settings have a TP of 48. Mode 3 with standard settings took advantage of the higher TP. I would recommend optimizing TP for Mode3 separately now that the signal is optimized. A very high TP may make more profit with Mode3, but it appears the big jumps Mode3 benefited from were news events.)
Methodology:
First key issue: time frame. I did my main optimization over 5 weeks of trading data. I filtered the results over 1 week after to cull the first set of results. I then filtered again over the 3 weeks previous to the first month. This is a total of 9 weeks of test data. Ideally, the pattern would be more like 6 weeks main optimization, 2 weeks current data, 4 weeks prior. This would give a schedule like the following in chronological order:
4 weeks Round 3 testing/filter
6 weeks Round 1 optimization
2 weeks Round 2 testing/filter
Actual performance:
3 weeks Round 3 (nov)
5 weeks Round 1 (dec/jan)
1 week Round 2 (jan)
I used the settings file with optimization turned on that I posted above. There were over 10,000 results. I turned on a filter that stopped testing if the account went to $8000 from $10000, because this wasn't likely to be a good result long term. This took 8 hours.
When I had results, I clicked the "Optimization Results" tab, right clicked, and saved them all to a file. This file is attached as "PhoenixOptimizations2.txt" These were the best results over the optimized time. Unfortunately, it could just be lucky curve fitting and not an actually good result to use in the future.
I took these results and ran the data over the next 1 week in Jan. (recommended 2 weeks) Right click an "Optimization Result", choose "Set Input Parameters". All of the top winners lost over the next week. I noticed that the results came in clumps. Later on I will try to pick a better optimization set to reduce clumping, but I didn't want to miss more "potential best" settings. Clumps that performed the same during optimization tended to perform the same over the 1 week follow up.
After filtering out the first few dozen results, I started to see winners over the new week of data. This was very interesting and all potential candidates for the new best. I copy and pasted each of these into a new file, "PhoenixOptimizations3.txt" and wrote down the amount they won. After finishing this for the top 1000 hundred results, I had a much smaller set to work with.
I tested the top results of each clump on 3 weeks of previous data (Nov). This gave me some relative ranking on how these settings will perform long term, but my main goal was getting the best setting on the previous week, with a decent setting on the 3 weeks of previous data.
I picked my favorites based on this weighted system and saved the "Optimization Results" as settings files on the main page. ("Set Input Parameters" "Expert Properties" "Save")
Here are my favorite 6 settings files, and the two Phoenix Optimization files I created for this as my working data. When looking in notepad, please note there are variables chopped off due to length. We'll fix this soon so the notepad files will contain all relevant data.
I recommend running this in Mode 1, Signals=20 MaxTrades=10 but any mode will work. (Mode3 works ok on these settings with a profit of 179.03, but I prefer $15k in Mode 1. As stated before, reoptimize TP and Mode3 specific settings if you want to use these signal settings with Mode3.)
Unbeleavable work Daraknor !
I recommend running this in Mode 1, Signals=20 MaxTrades=10 but any mode will work. (Mode3 works ok on these settings with a profit of 179.03, but I prefer $15k in Mode 1. As stated before, reoptimize TP and Mode3 specific settings if you want to use these signal settings with Mode3.)
Unbeleavable work Daraknor
! I will study your recommendations and test it as soon as possible.
Thanks a million for sharing this with the community !
Computer for testing
I noticed that there are 9 threads running concurrently on this machine, which means a dual processor machine may cut the time down significantly. If anyone has dual core, I'd be interested to see if setting one processor only increases the total estimate time.
I could buy and make available, 100 % of the time, a high performance dual core computer for testing. I presume that you and a few other collaborators could access it via remote desktop or anything else. We could run tests on very long periods of time and optimize settings and entry signal on all major currencies.
Let me know if interested and if positive, what to buy.
Great work! =)
Great work Daraknor!
Really thankful to you for sharing.
By the way, just confuse bout the TP and SL for best setting.
Is it 42 or 51 for TP(mode 1) . SL equal to 90..is it right?
I will start to forward test this setting and will post the results later.