Great work! I have to ask where the nickname snailbeard comes from?
| stevehopwoodforex.com https://www.stevehopwoodforex.com/phpBB3/ Print view |
|
| MNB 10.4 Streamlined by Snailbeard https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=2443 |
Page 1 of 9 |
| Author: | snailbeard [ Wed Jun 05, 2013 12:40 pm ] |
| Post subject: | MNB 10.4 Streamlined by Snailbeard |
SERIOUS WARNING Most Forex traders lose all their money. Using the robot posted here in trading Forex does not guarantee success. Trading this robot could lead to serious financial loss. Trading this robot without understanding its underlying trading strategies guarantees traders will lose their money. To trade this robot, you have to understand all concept MNB 10.4 Streamlined This is a derivative from an earlier version of MrLong's Multi-NB 10.4. Our EA's diverged significantly as my goals were different. Both MrLong and I have moved on from the original EA which we don't consider meets our requirements: The draw-down can be painful. MrLong is developing a new EA which will no doubt be even better than the last - I look forward to playing with it... However, I am not entirely unhappy with the original EA as it contains a framework on which to build on. My derivative is not intended to run on a live account, nor is it intended to be a dashboard (as that already exists). Like many people here I don't plan to write much more MQL because better platforms and programming environments are available, so I can't promise any updates to this. I have used MrLong's EA as a template for NB 10.4 trading logic but changed things around according to back-testing & forward-testing testability and repeatability. This is more of reworking so that: a) It behaves consistently and predictably (bug fixes) b) Code that only needs to run once an hour/H4/D1 is run as needed and results are cached - (only partly achieved so far) c) Decisions about trading or not can be tracked and logged - otherwise we are just flying blind. d) Lots of pending orders are generated on a demo account and a user selected sub-set can be manually traded on a live account. e) Any decision by the EA to not trade when MA & RSI are 'good to go' has to be justified - we can see why a potential trade was abandoned by looking in the log file. f) My EA's can always be back-tested - back-testing doesn't promise anything but if the EA can't survive a back-test then who know's what will happen on live data... During the reworking of the EA, I introduced lots of new bugs, but with the help of back-testing and reporting all/most of these new and old bugs have been fixed as well. Generally, I have tried not to change functions that do fundamental calculations, but for clarity and performance reasons I have changed so much (in places) that you might not think the code is a derivative. If you are not back-testing then it is unlikely that you would notice any performance improvement. Nevertheless, if you want to run multiple copies with different parameters and magic numbers then this version will use less CPU but more RAM. There is still at least one issue/bug coming out of a 6 month back test of GPBUSD. The attached graph shows that the EA has both good and bad periods - so best results would be with manual intervention for each trade/potential trade. Forward testing on demo without intervention had a positive result this week so far, but next week could be the opposite! It's here in case anyone finds it useful. A look at the results EA's can be unpredictable because an unusual combination of data or corrupt data invalidates the logic. I thought I was looking at a bug earlier today - the EA logged the bad event without crashing so that is a positive sign that some of the logic is robust. (Those of you who follow my posts will remember that I rewrote the correlation logic to prevent the EA going into an infinite loop when the data was incomplete). Here NZDJPY shows another example of screwy data: |
|
| Author: | snailbeard [ Wed Jun 05, 2013 12:45 pm ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
6 months of back-testing on GBPUSD: Don't be fooled by the final balance, what the graph shows is if you start running at the wrong time the account will lose money. There was one really good hit and then so-so. |
|
| Author: | snailbeard [ Wed Jun 05, 2013 12:50 pm ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
Recent forward-test demo account trade history: OK so it made lots of money in a demo account in the last couple of days but next week could be the opposite! |
|
| Author: | snailbeard [ Wed Jun 05, 2013 1:06 pm ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
The source and executable: This update addresses a couple of issues with next support/resistance and oversized lotsize protection. I am not planning any more updates to this EA, its just here for reference and learning. The only benefit of using this EA is that the log file records information about why it made particular decisions. It is also a good starting framework for making a multi-pair EA which is back-testable (testable on one currency at a time) |
|
| Author: | Jemook [ Wed Jun 05, 2013 1:12 pm ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
Great work! I have to ask where the nickname snailbeard comes from? |
|
| Author: | snailbeard [ Thu Jun 06, 2013 5:37 am ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
After more investigation of profits and losses, it appears that the big swing up in the profit is due to more dodgy data... However, it is good to see the EA coping with the rubbish that comes down the wire and actually logging useful stuff. Obviously, there is a need for any EA to check on all time frames for consistency and take action or no action accordingly. It is on my wish list to have a utility which ploughs through data and does its best to clean it up, although it is always good to test EA's on bad data first. |
|
| Author: | snailbeard [ Thu Jun 06, 2013 5:49 am ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
My wife thinks I am slow and I used to have a black beard. Of course she is a genius - I can never make sense of anything she tells me  |
|
| Author: | Jemook [ Fri Jun 07, 2013 2:45 am ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
Hi SB, Can you please let me know of any other corrupted data you find on GP demo? Appreciate your help. Jeremy |
|
| Author: | snailbeard [ Fri Jun 07, 2013 7:08 am ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
Will do I might invest in some utilities that do a clean up for me, or perhaps follow the recommendations about using Dukascopy as having the best data for testing. However, extreme data is good for picking up dangerous bugs in EAs. I have manually cleaned up the M5 data and rerun the BT with hopefully more realistic results: |
|
| Author: | MrLong [ Mon Jun 10, 2013 7:46 pm ] |
| Post subject: | Re: MNB 10.4 Streamlined by Snailbeard |
Hello Brian, Nice work mate!! I look forward to your improvements. Andy |
|
| All times are UTC | Page 1 of 9 |
|
Powered by phpBB® Forum Software © phpBB Limited |
|