Please Review My EA Test Results: 287% Return Annually

New comment
 

Now I know that this is probably mostly luck because I'm not a great programmer, but I wanted you guys to pick apart where my EA might be weak. I've heard max draw down is important, but am not sure why. Can you tell me from this testing report where my EA might be weak, or (better yet), that I've got an EA that looks like it is a solid performer?


Here is the test results. NOTE: They haven't even been optimized yet....



Strategy Tester: Nemesis - v3.4
Strategy Tester Report
Nemesis - v3.4
IBFX-MT4 Mini-1 (Build 225)

SymbolEURUSDm (Euro vs US Dollar)
Period4 Hours (H4) 2009.01.01 20:00 - 2010.01.11 20:00 (2009.01.01 - 2010.01.12)
ModelEvery tick (the most precise method based on all available least timeframes)
ParametersStopLoss=25; TakeProfit=25; AgainstTrendLots=0.3; WithTrendLots=0.3; SetDeviationPrice=0.5; Dev=50; MaxDev=55; KellyCriterion=0.8; MaxLots=7.5; MultiFrame=false; TradeAgainstTrendFalling=true; TradeAgainstTrendRising=true; TradeWithTrend=true;
Bars in test2142Ticks modelled5163270Modelling quality55.32%
Mismatched charts errors1
Initial deposit10000.00
Total net profit27805.67Gross profit82761.47Gross loss-54955.80
Profit factor1.51Expected payoff121.42
Absolute drawdown1717.34Maximal drawdown5097.07 (12.70%)Relative drawdown20.72% (2164.89)
Total trades229Short positions (won %)123 (66.67%)Long positions (won %)106 (66.04%)
Profit trades (% of total)152 (66.38%)Loss trades (% of total)77 (33.62%)
Largestprofit trade1787.77loss trade-1937.28
Averageprofit trade544.48loss trade-713.71
Maximumconsecutive wins (profit in money)8 (8493.80)consecutive losses (loss in money)4 (-3849.62)
Maximalconsecutive profit (count of wins)8493.80 (8)consecutive loss (count of losses)-3849.62 (4)
Averageconsecutive wins3consecutive losses1

#TimeTypeOrderSizePriceS / LT / PProfitBalance
12009.01.01 23:15buy17.501.39651.37771.4103
22009.01.05 08:14s/l17.501.37771.37771.4103-1414.898585.11
32009.01.06 17:14buy27.501.34711.32341.3645
42009.01.07 11:31t/p27.501.36451.32341.36451302.559887.66
52009.01.07 13:42sell37.501.37081.39211.3552
62009.01.08 08:18t/p37.501.35521.39211.35521161.7211049.38
72009.01.08 12:47sell47.501.36191.38411.3456
82009.01.09 17:31t/p47.501.34561.38411.34561219.7412269.12
 

Also, I'd like to know what type of numbers I should be shooting for in these reports?


What type of:

  • Modeling quality percentage?
  • Profit Factor?
  • Maximal Draw Down?
  • Relative Draw Down?
  • Win/Loss Percentage?

I've also read about a ratio that is a good judge of strength, but can't remember the name. It takes the max draw down and divides it by total profits....what is a good figure for that ratio?


Thanks to all the elders for helpin out this newb :)

 

i test my EAs with an initial deposit around 200 Euro and a leverage of 1:500

(if you have a large deposit, you can even throw a coin and wait until your profit comes in - no problem with large drawdowns and bad indicators)

if they survive with that money more than 3 month, i call them average.

if they survive 1 year, and endbalance is at least 70% of startbalance, i call them good.

if they survive 1 year and make 10% profit per month, i am satisfied.

and if i have a result like this, i am wondering and search for bugs ;-)

these are my testing conditions.

 

das ist gud, ya? :)


Thanks for the advice mike.

 

2009.01.01 20:00 - 2010.01.11 20:00 (2009.01.01 - 2010.01.12)

Ten days, How about testing for a year.

 
WHRoeder:

Ten days, How about testing for a year.

2009!

 
LOL! I make that mistake at the start of every year too :D
 

In answer to your questions:

Modeling quality roughly interprets as granularity of invented data. Depending on how your EA works, I wouldn't spend any time on this. Once you've got past functional testing of your EA, I'd focus more on testing against a number of instruments and timeframes with data that you know the provenance of. Since the strategy tester has it's limitations, that obviously means forward testing too.

Acceptable levels for profit factor and drawdown will be driven in many cases by your investors.

And in terms of win/loss percentage, aim for as good or just better than a coin toss. Don't try too hard to get your signal right all the time as you'll run the risk of overfitting. Put the effort into leveraging the percentage by good money management.


Just my humble opinions.


CB

 
meikel:

2009!

Sorry, I mis-read.


The question is did you optimize for the full 2009 or some other period. See Updated EA, Backtesting, common mistakes and some working code

Don't try too hard to get your signal right all the time as you'll run the risk of overfitting.

I optimized my EA for 4/2007-5/2009. I got Starting Balance=$100.00, Ending Balance=$115,026,307! (sic) Then I tried 5/2009 to present. Got Starting Balance=$100.00, High of $111,000, and final balance of $8.00 So far I can't get any setting to work outside of 2008.


You must run on other periods before you can think you have anything. Then, forward test for several months. Finally try live.

New comment