Can I get a few thoughts on these backtest results?
That long initial stretch of almost flat-line, and that last stretch of down-hill slide says the EA has a high level of "inconsistency". No, that "inconsistency" measure is not part of the report, but it may be quite useful if it is.
The front strech I can live with, its a slow build from 10k to 50K its the last 2 weeks of May that killed me. I need a better way to try and filter my trades, There is a 60 pip stop on all trades and all trades are closed on Fridays to prevent a Gap loss. I am also thinking of moving risk % down as balance grows instead of leaving it steady that just might make the chart better but not improve the EA. I am going to add in June this weekend, maybe extend the test period back another 12 months as well.
I am going to drop an indicator on my test template to see if I can find one that could knock out a few more losers then winners I could inprove my results dramatically.
Thanks for the comments.
Yes, finding the right filters would definitely help, as long as it doesn't also filter out the profitable trades.
I've never been a fan of fixed-pip stops. Neither have I been a fan of fixed risk %. They should be dynamic depending on the trading conditions, and risk/reward ratio - which becomes a matter of how you actually calculate those values in your EA.
Yes, finding the right filters would definitely help, as long as it doesn't also filter out the profitable trades.
I've never been a fan of fixed-pip stops. Neither have I been a fan of fixed risk %. They should be dynamic depending on the trading conditions, and risk/reward ratio - which becomes a matter of how you actually calculate those values in your EA.
| Symbol | EURJPY (Euro vs Japanese Yen) | ||||
| Period | 5 Minutes (M5) 2010.07.08 09:05 - 2011.06.23 23:55 (2010.07.01 - 2011.06.24) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Bars in test | 66161 | Ticks modelled | 16311952 | Modelling quality | n/a |
| Mismatched charts errors | 99827 | ||||
| Initial deposit | 10000.00 | ||||
| Total net profit | 165559.36 | Gross profit | 297982.38 | Gross loss | -132423.02 |
| Profit factor | 2.25 | Expected payoff | 325.26 | ||
| Absolute drawdown | 1917.06 | Maximal drawdown | 88562.83 (34.38%) | Relative drawdown | 35.57% (18406.04) |
| Total trades | 509 | Short positions (won %) | 248 (39.52%) | Long positions (won %) | 261 (46.74%) |
| Profit trades (% of total) | 220 (43.22%) | Loss trades (% of total) | 289 (56.78%) | ||
| Largest | profit trade | 24882.43 | loss trade | -9280.28 | |
| Average | profit trade | 1354.47 | loss trade | -458.21 | |
| Maximum | consecutive wins (profit in money) | 15 (58185.44) | consecutive losses (loss in money) | 10 (-577.67) | |
| Maximal | consecutive profit (count of wins) | 112674.15 (8) | consecutive loss (count of losses) | -13906.95 (2) | |
| Average | consecutive wins | 3 | consecutive losses | 4 | |
I added a dynamic risk feature. This improved the results by a decent amount. This test is a one year as opposed to the original which was only 9 months. It caught most of the winners at higher risk and most of the losers at the lowest risk. The risk variation is 3% -> .5 %. Thanks for your ideas.
Your mismatched chart errors have got something to do with MT4 getting data incorrectly from your broker. You can get rid of it, by downloading data from the MT4 servers, but that data does not match the data from your broker, or anyone else's broker, but it will allow you to model an EA and see how it reacts to some sort of data, that is close to what you are expecting.
There is some good help here;
http://www.forexnirvana.com/f41/getting-rid-mismatched-chart-errors-backtests-57/
It helped me get mine down. Although it never really seemed to make much of a difference to the results.....however I am somewhat of a noob to EA programming.
I hope the link helps.
Colin
Thought I might add this about you chart, judging by how clear cut your loosing streaks are, it would be statistically easy to see what was common on those loosing streaks by graphing the results and see the common denominator. Then you just have to not trade when those circumstances arise. Hope that does not sound too vague like someone reading your palm.......
Colin
Thought I might add this about you chart, judging by how clear cut your loosing streaks are, it would be statistically easy to see what was common on those loosing streaks by graphing the results and see the common denominator. Then you just have to not trade when those circumstances arise. Hope that does not sound too vague like someone reading your palm.......
Colin
Thanks for your input. I was trying to work on filtering out the losing trades. The second set of results that I posted, in reality I probably ran close to 100 sets of one year backtests to get that improvement, was my latest, I added a volatility indicator to try and cut some losses. I found some issues with it while backtesting other pairs. I am in the process of tweaking it now I will post the more results when I have them.
Dan
That last stretch of consistent downhill slope is a concern.
You appear to have an over-optimized EA that can make big jumps in profits, when the trends are favorable, for the period you are testing.
As someone has already mentioned, run optimization for one date-range, and then back-test it for a completely different date-range, and see what sort of curves you get. If it does well for the optimized period, and flops on anything else, I would not put real money on the EA, yet.
| Symbol | EURJPY (Euro vs Japanese Yen) | ||||
| Period | 5 Minutes (M5) 2010.07.08 09:05 - 2011.06.29 23:55 (2010.07.01 - 2011.06.30) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | TrendFastTimeFrame=60; TrendSlowTimeFrame=60; TrendFastPeriod=1; TrendFastMethod=0; TrendFastPrice=0; TrendSlowPeriod=24; TrendSlowMethod=0; TrendSlowPrice=0; VeryFastTimeFrame=5; FastTimeFrame=5; MedTimeFrame=5; SlowTimeFrame=5; VeryFastPeriod=5; VeryFastMethod=1; VeryFastPrice=0; FastPeriod=15; FastMethod=1; FastPrice=0; MedPeriod=30; MedMethod=0; MedPrice=0; SlowPeriod=45; SlowMethod=0; SlowPrice=0; iShift=5; ATRLevelNOWAY=0.32; ATRLevelFULL=0.1; ATRLevelLITE=0.06; StackSizeFull=7; StackSizeLite=1; DistanceApartFull=100; DistanceApartLite=100; StopLossFull=600; StopLossLite=600; TrailingStopLossStepFull=200; TrailingStopLossStepLite=200; TakeProfitFull=0; TakeProfitLite=0; RiskFull=0.03; RiskLite=0.005; LotSize=0.1; Lots=0.1; Slippage=0; MAttempts=1; MNumber=0; maGap=250; | ||||
| Bars in test | 67298 | Ticks modelled | 16961277 | Modelling quality | n/a |
| Mismatched charts errors | 99916 | ||||
| Initial deposit | 10000.00 | ||||
| Total net profit | 610317.14 | Gross profit | 871000.36 | Gross loss | -260683.23 |
| Profit factor | 3.34 | Expected payoff | 1925.29 | ||
| Absolute drawdown | 989.49 | Maximal drawdown | 227851.88 (28.80%) | Relative drawdown | 48.59% (227625.88) |
| Total trades | 317 | Short positions (won %) | 152 (46.05%) | Long positions (won %) | 165 (54.55%) |
| Profit trades (% of total) | 160 (50.47%) | Loss trades (% of total) | 157 (49.53%) | ||
| Largest | profit trade | 54394.65 | loss trade | -24521.58 | |
| Average | profit trade | 5443.75 | loss trade | -1660.40 | |
| Maximum | consecutive wins (profit in money) | 16 (162575.59) | consecutive losses (loss in money) | 11 (-3882.38) | |
| Maximal | consecutive profit (count of wins) | 368228.29 (13) | consecutive loss (count of losses) | -65899.92 (4) | |
| Average | consecutive wins | 3 | consecutive losses | 3 | |
Top graph is 15m timeframe. I have 30+ parameters in the EA. I only changed the trading ma's, 4 + 4 more that feed off that time frame that don't have there own setting in the properties. I left the ATR level settings but did change the timeframe. I left the trend ma's the same 1 hr. I did not expect to get the same results as the 5min but I was happy that it did almost double over the year without really having to optimize the reat of the EA( stacksize stop, trailing, risk maGap etc). Had I done this and adjusted my ATR setting, which would be very different on the 15, I could probably get a much better return on the 15 minute. I am more interested in droping the EA on different pairs on a 5m. Just from testing, I already know that the ATR setting and the levels that I use to determine trade size need to be adjusted for each pair and or timeframe. This prevents this EA from just being dropped on any pair any timeframe without changng a few settings.
The bottem set and graph 5m, has improved by leaps. I ran the optimizer over the last few days, 1400 runs. It helped me tweek a few settings stacksize and stack distance and trailing distance for both FULL and LITE trades.
This one is going back in my demo account to run for a few weeks.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Symbol
This is my first post and my first attemp at an EA. I have moved this to a demo account to watch it in real time and make sure it is opening closing trades correctly etc.
What are Mismatched chart errors? Is that alot 12081?