Great EA in backtest! - page 123
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
follow the last version that i maked some changes *only logging system*
Cyberia Statistical File Final 2005b.zip:
Upload of file failed.
Sorry I can't upload my draft worksheet with the analysis of results variables.
My statistical lessons at college are finally making some sense now! Also next week I'm going to show it to my discrete mathematical teacher, that's starting to get really interesting! in a short time we should get a serious analysis over it, and some others news ;]
**If someone want to see that worksheet just PM me and I send it by e-mail.
Today I contacted the author of CyberiaTrader, OpenStorm at Automated Trading Championship 2006.
Bellow is what i write him in english.
Could you please answer my questions?
1. At that championship why don`t you increase your risk to reach the top? increasing lot size i know that you could reach them!
2. Now, about your CyberiaTrader that you built, it's just amazing! i really appreciate your actions to release it as open-source in the first stages, my ask is for are you paying attention over the discussion about CyberiaTrader open-source on forums? (https://www.mql5.com/en/forum/174700)
I'm the user templar, which with help of some friends to bring the ideas, already started to make a good analysis over the CyberiaTrader's variables, such as, analyzing you deviations, distributions, mode, general patterns, and with quartiles intervals getting the quantity of "win / loss" with that thinking about some low impact filters (that block minimal quantities of trades possible, only chop down the losses).
I think with you [as creator, and your nice formation] we all could get much over it, some of that analysis process could be improved and done too to your professional-version, you help us to help you!
We here wish you good trades!Week report...
Two weeks of demo with CT v1.93b:
Doing like the backtest, but in a lower trading rate...
lets... wait...
Two weeks of demo with CT v1.93b:
Doing like the backtest, but in a lower trading rate...
lets... wait...good job
75% same as mine.
I noticed that it won 100% of its long trades. You can create another user profile on your PC and create another demo. Just leave each user logged on, and you can run multiple demos on one PC.
Mabe try running one with the settings to place long orders only.
good job
75% same as mine.
I noticed that it won 100% of its long trades. You can create another user profile on your PC and create another demo. Just leave each user logged on, and you can run multiple demos on one PC.
Mabe try running one with the settings to place long orders only.That is good, indeed, but over 2 weeks 100% isn't much significant; I think that the buy success was a result of the up trend it got in these 2 weeks.
Perhaps in a month, if it kept the high win %, I would start another demo with longs only (that's nice, I didn't know that
);
anyways, me and Templar are working (we have to finish the college exams to continue it properly) on a statistical analysis over 1.93b;
We are separating all the variables used by CyberiaLogic on the 3 main possibilities (Buy; Sell; Uncertainty) and their values by percentiles as well as crossing the results (win/loss) with them, to take conclusions and develop filters over CyberiaLogic (which could be enabled/disabled at any time);
The main idea is to work on the losses, just to reduce them, without compromising the wins, what seems possible so far, increasing the win % significantly (what is really hard);
For that, we ask help from people who are interested on develop that work with us;
That is good, indeed, but over 2 weeks 100% isn't much significant; I think that the buy success was a result of the up trend it got in these 2 weeks.
Perhaps in a month, if it kept the high win %, I would start another demo with longs only (that's nice, I didn't know that
);
anyways, me and Templar are working (we have to finish the college exams to continue it properly) on a statistical analysis over 1.93b;
We are separating all the variables used by CyberiaLogic on the 3 main possibilities (Buy; Sell; Uncertainty) and their values by percentiles as well as crossing the results (win/loss) with them, to take conclusions and develop filters over CyberiaLogic (which could be enabled/disabled at any time);
The main idea is to work on the losses, just to reduce them, without compromising the wins, what seems possible so far, increasing the win % significantly (what is really hard);
For that, we ask help from people who are interested on develop that work with us;that was what I tried to do and the best I got was about 71% in live trading. I hope you guys can do better. In live trading 71% is still unprofitable for me.
that was what I tried to do and the best I got was about 71% in live trading. I hope you guys can do better. In live trading 71% is still unprofitable for me.
It could be just a drawdown...
Besides, ppf version decrease the drawdown, sacrificing a lot of profit... it just made the graph flatter...
I am thinking on creating my live soon with CT1.98b; and work on the statistical analysis over CT with Templar...
Crown Forex
"Also now Crown forex shuts down my computer - it happened twice already"
Mmm i think you need to make sure microsoft updates are not doing its stuff at 3 am and rebooting. Its happened to me before i realised all my progs were closed one morning. Check you system log files and make sure autoupdate is OFF.
One other thing MT just upgraded to build 198 gives complelty different results to 197. Im not complaining its a lot better but having build standards change backtesting makes life doubly difficult. I now get better results on 15 mins with no indicators to filter just the pure logic. However seems the most important setting is the values period settings. This checks back how many bars it bases the market information on to establish the risk on opening future orders. Interesting results to be found using fibo sequnce here 8, 13, 21, 34, 55, 89, 144, 233 although the later numbers take HOURS to trawl through just a few weeks of data as its so very PC intensive. I believe you should stick to fib values here. We are dealing on luck, odds, and ratios in the noise and need all the help we can get:)
I also agree that risk should be 0.5 which opens about 2 lots per 10k. Any more then that is just asking for trouble.
Have a good weekend.crown forex started at saudi arabia at 2004, the owner from jordan with saudi partner ,his first name is ibrahim, all offices of crown forex is closed now at saudi arabia after they stolen millions from saudis..... ibrahim now is requested by law at saudi arabia... it was registered with NFA, but NFA cancelled registration later.......
i can provide more detailes about the cases of crown forex at saudi arabia....
the saudi partner of crown forex is now at jail, with about 50 million us dollar stolen from people here....it is small refco...
this is brief background about company and owner...
Maybe an alternative to improve backtest results. I don't know if this is valid, but it caught my attention on mql forum. http://forum.mql4.com/4965
I think the PeriodGen is a script. I put it in the script folder and attached it to M1 offline chart and not sure anything happened. I believe using the the Period_Converter script will also synchronize all timeframes OK.
I will try to hardcode timeframe in a few EAs and then backtest on M1 Strategy Tester.
irusoh1 2006.11.25 02:31
For most possible accurate results in backtester:
1. use alpari data only (MT history gives weird results with sensitive EAs)
2. synchronize all timeframes (I wrote a little script to do that) and put them into history folder
3. hardcode your timeframe in tester (i.e. use iTime, iHigh, etc. use explicit PERIOD_?? instead of 0 or Period() function)
and run your EA on 1 minute frame only)
4. Run your EA on demo for a week or two then compare results with tester for the same period.
Then maybe you will get closer to real life. That is my bitter experience with backtesting.
Attached files:
periodgen.mq4 (7.83 KB)
Maybe something, maybe not.
Wackena
Maybe an alternative to improve backtest results. I don't know if this is valid, but it caught my attention on mql forum. http://forum.mql4.com/4965
I think the PeriodGen is a script. I put it in the script folder and attached it to M1 offline chart and not sure anything happened. I believe using the the Period_Converter script will also synchronize all timeframes OK.
I will try to hardcode timeframe in a few EAs and then backtest on M1 Strategy Tester.
irusoh1 2006.11.25 02:31
For most possible accurate results in backtester:
1. use alpari data only (MT history gives weird results with sensitive EAs)
2. synchronize all timeframes (I wrote a little script to do that) and put them into history folder
3. hardcode your timeframe in tester (i.e. use iTime, iHigh, etc. use explicit PERIOD_?? instead of 0 or Period() function)
and run your EA on 1 minute frame only)
4. Run your EA on demo for a week or two then compare results with tester for the same period.
Then maybe you will get closer to real life. That is my bitter experience with backtesting.
Attached files:
periodgen.mq4 (7.83 KB)
Maybe something, maybe not.
WackenaThanks Wackena, but I think that there is a simpler way do have good backtest:
1. Download MT4 build 200 from any broker and at History Center, press the "Download" button;