Free Coding to backtest your strategy

Testing using Stochs only. Here it takes a buy trade when the stochs cross the 50 line in the sell direction? Or possibly (and more likely) a coincidental is it a late one from the stochs changing to over 50 a few bars before?

RSI only looks good. I'm going to stop for the night now and test the stochs only again tomorrow. I'm not 100% sure about them so I want to watch them again just in case. I'll also check with short trades as right now I'm only checking long.

Ah the joys of moving average crossover systems. Is too late, too early? Did it cross enough and was this one significant? And the periods, should it be this or that, or maybe on a different TF? So many parameters to adjust and so little time.

Well at least it will keep you busy and stop the mind from wandering. 😄

Im kind of new to coding and forex in general. I've kind of been into EAs though. would I just have to tell you a strategy and you would code it for me or would I need an custom indicator or something. I want a ea that basically trades a combination all time high and low points in a pair, and that also trades formations like double top, double bottom and bullish and bearish flag formations.

Hi, I'm sending you an EA build indicator. Based on the indicator signal to enter operation. Have a stop and take a profit, which if equal to zero, will work from signal to signal.

haha. To be honest the strategy of this one is pretty nailed down already. There are always refinements to any system but a lot of people have been using this live for several years with excellent results. I think the exits are actually a more complex endeavour. Though I 'may' have a pretty straightforward solution to those. That's for later though.

I haven't checked shorts yet. But I have checked the indis separately in long positions and it all looks great. All unsuitable trades seem to be caused by the lack of code to tell the EA how many bars from the confluence of the signals it's allowed to wait before taking a trade. Which I know hasn't been added yet, so isn't an error.

I'll do the same with shorts now.

Looking very good indeed I think

ADDED: I'm not getting any short trades at all. Are these enabled? I've tried with the original settings file and still none.

Hello Arteks,

Thanks for posting. Does this repaint??

Yes,Maksuel,
like all indicators, and for that I want the encoder to put a stop, which if there is any repainting or a sharp movement to hit the stop and to come in again with a new signal. The picture has a setting arrow signal to appear on a +1 candle.

In that case my skepticism is misplaced, my apologies. I just think of the function of a MA, which is to find the discriminant of a linear function, to draw a 'line in the sand' which is quite useful, but necessarily lagging behind price. And also to act as a low-pass filter, blocking higher frequency but irrelevant noise from the 'true signal' of price movement. When a system depends on two such filters (a filter of a filter) it is in most cases doubling the lag behind a signal and then a variety of barbarisms and absurdities must be introduced to compensate, torturing the MA to perform tasks for which it is unsuited.

I completely understand, and it's never going to replace an experienced human behind the monitor. I also think skepticism is the way forward! But that's not to say it couldn't be a profitable EA. For example. I've been testing the functions a lot today and have attached a pretty standard chart for major pairs with it below. This is just a visual run to check the EA was performing as expected.

Considering the following:

The EA doesn't currently have a proper strategy for exits, which I think in this case should be based on stochs retracement.

It's not taking short trades yet.

It's not taking continuation trades yet. These are actually the more reliable trades using these indicators and much less likely to suddenly reverse than the crossovers.

Finally (but crucially), about 20% of the trades it's taking are currently incorrect due to it entering too late on those occasions. The code to stop that happening has yet to be added. Because of this it's often exiting trades in profit but then taking the same trade again on the next bar simply because the trade indicators from before are still valid. Of course this means it's now sometimes entering at the end of the move and dumping the profit it just made. All this actually needs is the addition of code to say that it's only being allowed to enter a trade on the start of the next candle after the indis line up.

With the above done you would obviously want to add good money management. And then perhaps a filter to check the trade lines up on other charts (5M and 1D), and code to stop it trading around major news events. If you really wanted to make it 'all it can be'.

Somehow I overlooked the first EA (Tandem) that KA produced. I am now testing on Tickmill with amazing 82% accurate data. ;?

This is a randomly selected period, H1 using the supplied setfile. Promising. I will do more tests.


Strategy Tester Report
TANDEM EA V1.01
Tickmill-DemoUK (Build 1170)

SymbolXAUUSD (Gold vs US Dollar )
Period1 Hour (H1) 2016.07.05 04:00 - 2019.08.23 23:00
ModelEvery tick (the most precise method based on all available least timeframes)
ParametersInpMagicNumber=190812; InpMode=0; _Entry="=========="; InpPrimaryTF=43200; InpSecondaryTF=240; InpATRPeriod=21; InpATRPercent=50; InpATRBuffer=0.1; InpEntryCandle=false; InpFilterPTFCC=false; _Exit="=========="; InpCloseOpposit=false; InpSLEntryCandle=true; InpTPRR=20; InpStopLoss=200; InpTakeProfit=2000; InpSLATRMultiplier=1; InpTPATRMultiplier=20; InpTPSLMINMAX=1; _TrailStop="=========="; InpUseTrailingRR=true; InpTrRRTrigger=5; InpTrRRTrail=3; InpUseTrailingStop=true; InpTrMinProfit=200; InpTrailingStop=200; InpUseTrailingATR=true; InpTRbyMultiplier=2; InpTRstMultiplier=1; InpTRLMINMAX=1; _BreakEven="=========="; InpUseBreakEvenRR=true; InpBreakEvenRRTrigger=3; InpBreakEvenRRLock=0; InpUseBreakEvenStop=true; InpBeMinProfit=200; InpBeLockProfit=0; InpUseBreakEvenStopATR=true; InpBeMinProfitATR=3; InpBeLockProfitATR=0; InpBEMINMAX=1; _mm="=========="; InpUseMoneyManagement=true; InpRiskPercent=2; InpFixedLotSize=1; InpMaxTrades=3; InpMinATRbtwTrades=3; InpMaxATRbtwTrades=10; _oth="=========="; InpTradeComment="TandemBT"; InpShowReport=false; csvFile="TandemStats";
Bars in test14144Ticks modelled34789434Modelling quality82.09%
Mismatched charts errors20
Initial deposit3000.00Spread10
Total net profit4853.07Gross profit12670.52Gross loss-7817.45
Profit factor1.62Expected payoff9.59
Absolute drawdown989.48Maximal drawdown1269.36 (38.70%)Relative drawdown38.70% (1269.36)
Total trades506Short positions (won %)268 (73.51%)Long positions (won %)238 (50.00%)
Profit trades (% of total)316 (62.45%)Loss trades (% of total)190 (37.55%)
Largestprofit trade1320.37loss trade-133.10
Averageprofit trade40.10loss trade-41.14
Maximumconsecutive wins (profit in money)15 (257.79)consecutive losses (loss in money)7 (-189.80)
Maximalconsecutive profit (count of wins)2013.02 (2)consecutive loss (count of losses)-303.42 (6)
Averageconsecutive wins3consecutive losses2

Results using EU with the same settings.

EU H4 starting 2010.01.25

100 trades on the FXBlue simulator using the 'up & down' indicator traded 'signal-to-signal' where one signal closes the previous one, and multiple signals of the same type opens multiple trades. This seems to be a renamed and modified John Ehlers indicator, similar if not the same to the Fisher-Transform. One thing I noticed is that it places signal arrows in the past to make it seem more accurate than it really is. It also tends to produce more signals during ranges than during trends, probably using some kind of crossed lines method like the original F-T. So a signal-to-signal method where one trade is closed by the previous will create horrendous losses during ranging periods, but in trends it does quite well. Note that the big loss near the end of the graph was caused by a 'missing' buy signal which I was expecting but did not materialize until far too late for some reason.

I would not scorn an EA based on the original Ehler's F-T but it would need to be tuned to the cycle period being traded and it would need more rules than has been provided.

Sorry for my bad english can you make ea and test this for me

Забележки:

1. Ако персонализираният индикатор, използван в стратегията, има mq4 изходен файл, който може лесно да бъде вграден в кода на EA, той се счита за стратегия без индикатор (под приоритет 1).

Hi Abokwaik, post 145 is for you and if possible do EA. I explicitly emphasize it because my colleagues thank you for sharing. This is not a sharing but a task and if there is something unclear I am ready for an answer. Nice day

TMA Reversal attempts
All H1, all starting 2017.06.17

Default settings


Reversed strategy


Reversed strategy with TMA cross turned off


Reversed strategy with switch modes on TMA Slope (default slope)


Widened SL, period and StdDev

Hi Abokwaik & Clemmo,
TMA Reversal 1.2
EU, original method on H1, for 1 year on tick-data from TickStory, 99.90% accuracy.
This has also been run on other pairs with similar results (after adjusting half length and deviation %).
Abokwaik,
Thank you for starting this thread and for doing the work to make other people's ideas a reality.
Very generous and very informative effort.
I'd like to put your TMA Reversal 1.2 indicator into an EA to try forward testing on demo account.
Would you do that or should I ask another programmer?
Thanks again.
All the best,
John

Ignore this! For some reason the EA was defaulting to long only in the main settings, not sure why since others don't so I just overlooked the obvious. lol

This is a demo version that runs till Sept.30.

This is the message I got in the journal tab after attempting to optimize some settings (running overnight) using the genetic algo of the TANDEM EA.

2019.08.25 01:38:11.739 There were 293 passes done during optimization, 295 results have been discarded as insignificant

How can you discard more passes than were done? Also -reaalllly MT4? 293 optimizations had no significance?? Another argument in favor of live testing.