Strategy SR Dow J Day
-
Daily strategy with indicator created by Roberto Gozzi.
This strategy has been optimized with tick by tick mode until December 2017.
To the indicator created by Roberto I added two trend indicators being based only on long purchases, to this I put a monthly stop to not increase the Drawdown in case of sudden and prolonged sales …. (as it is in 2020).// Strategy SR Dow J. Day indicator by roberto gozzi DEFPARAM CumulateOrders = false indicator1 = ADX[10] indicator2 = ADXR[14] c2 = (indicator1[1] > indicator2[1]) indicator11 = ADX[10] c1 = (indicator11[1] >= 15) ONCE PrezzoC = 0 ONCE PrezzoB = 0 ONCE PrezzoA = 0 Prezzo = CALL SRtest IF Prezzo > 0 THEN PrezzoA = PrezzoB PrezzoB = PrezzoC PrezzoC = Prezzo ENDIF If longonmarket and (barindex-tradeindex)>=43 then SELL AT MARKET ENDIF LossMonth =200 once StopMonth=1 IF Month <> Month[1] THEN ProfitMonth=strategyprofit StopMonth=1 endif If strategyprofit<=ProfitMonth-LossMonth then StopMonth=0 endif Cond = PrezzoB > PrezzoA AND PrezzoB > PrezzoC IF StopMonth=1 and close CROSSES UNDER PrezzoB AND C2 AND C1 AND Cond > 0 THEN BUY 1 Contract At Market SET TARGET PROFIT 5* averagetruerange[1](close) SET STOP %LOSS 8 ENDIF Trailinglong = 2* averagetruerange[5](close) Trailingsteplong = 5* averagetruerange[1](close) TGL = Trailinglong STPL = Trailingsteplong if not onmarket then MAXPRICE = 0 PREZZOUSCITA = 0 ENDIF if longonmarket then MAXPRICE = MAX(MAXPRICE,close) if MAXPRICE-tradeprice(1)>=TGL then PREZZOUSCITA = MAXPRICE-STPL ENDIF ENDIF if onmarket and PREZZOUSCITA>0 then EXITSHORT AT PREZZOUSCITA STOP SELL AT PREZZOUSCITA STOP ENDIFYour strategy uses STOP and TARGET orders so you must test it with tick by tick on. You are not doing this. Tick by tick on the DJI only allows us to test from 2010 onwards.
Also you are using a spread of 2 for the DJI. 3.8 is the norm I believe for end of day trades when positions are opened. Spread might however be lower if STOP or TARGET orders are hit during the trading day.
Even with these changes the equity curve is quite pretty though.
Just for interest I decided to add my ‘Strategy Every Bar’ code to your strategy as it is a prime candidate as trades can be on the market for quite some time and so misses other possible trades. My first test was with just 31 tests (one for each day of the month). This increased the number of trades tested to 102 from 65 out of a maximum of 120 possible trades. So we went from testing only 54% of all trades to testing 85% of all trades. Despite this the total profit only increased by 1.24%. The win rate dropped by 3.94% and the average gain per trade dropped by £97.16
So although profitable performance is not as good as it first appears.
This is the code I used. dn must be optimised from 0 to 31.
// Strategy SR Dow J. Day indicator by roberto gozzi DEFPARAM CumulateOrders = false defparam preloadbars = 0 tradeon = 0 if day = dn or dn = 0 then tradeon = 1 endif indicator1 = ADX[10] indicator2 = ADXR[14] c2 = (indicator1[1] > indicator2[1]) indicator11 = ADX[10] c1 = (indicator11[1] >= 15) ONCE PrezzoC = 0 ONCE PrezzoB = 0 ONCE PrezzoA = 0 Prezzo = CALL SRtest IF Prezzo > 0 THEN PrezzoA = PrezzoB PrezzoB = PrezzoC PrezzoC = Prezzo ENDIF If longonmarket and (barindex-tradeindex)>=43 then SELL AT MARKET ENDIF LossMonth =200 once StopMonth=1 IF Month <> Month[1] THEN ProfitMonth=strategyprofit StopMonth=1 endif If strategyprofit<=ProfitMonth-LossMonth then StopMonth=0 endif Cond = PrezzoB > PrezzoA AND PrezzoB > PrezzoC IF tradeon and StopMonth=1 and close CROSSES UNDER PrezzoB AND C2 AND C1 AND Cond > 0 THEN trades = trades + 1 if not onmarket then BUY 1 Contract At Market SET TARGET PROFIT 5* averagetruerange[1](close) SET STOP %LOSS 8 endif ENDIF Trailinglong = 2* averagetruerange[5](close) Trailingsteplong = 5* averagetruerange[1](close) TGL = Trailinglong STPL = Trailingsteplong if not onmarket then MAXPRICE = 0 PREZZOUSCITA = 0 ENDIF if longonmarket then MAXPRICE = MAX(MAXPRICE,close) if MAXPRICE-tradeprice(1)>=TGL then PREZZOUSCITA = MAXPRICE-STPL ENDIF ENDIF if onmarket and PREZZOUSCITA>0 then EXITSHORT AT PREZZOUSCITA STOP SELL AT PREZZOUSCITA STOP ENDIF graph tradesI just realised that I forgot to add DEFPARAM PRELOADBARS = 0 to the above code so the possible trade count is wrong. There were actually only 100 possible trades. I’m not sure how separating it into 31 strategies then created 102 trades! Anyway we tested all possible trades and the return from the strategy was barely better so it seems your back test just got lucky and hit all the good trades and missed all the bad trades.
I have edited the code in my post to add the correction.
I’m also not a massive fan of the bit where it closes any trade after 43 days as it has a strong chance of curve fitting. I checked it with other values than 43 and got the attached chart. It seems closing any trades over 31 days returns a similar result. So I did a SEB test with that condition to close after 43 days removed. The results were much worse so it seems that giving up on trades that have been open for a long time is worthwhile in this strategy but it worries me that this exit condition makes such a difference especially as so few trades are actually closed by it.
I added a “TR” filter …. First Chart without Time Stop (43 Days) the second with Time Stop inserted.
Always Tick By Tick Optimized to 2017.
// Strategy SR Dow J. Day indicator by roberto gozzi DEFPARAM CumulateOrders = false indicator1 = ADX[10] indicator2 = ADXR[14] c2 = (indicator1[1] > indicator2[1]) indicator11 = ADX[10] c1 = (indicator11[1] >= 15) c30 =TR(close)>25 ONCE PrezzoC = 0 ONCE PrezzoB = 0 ONCE PrezzoA = 0 Prezzo = CALL SRtest IF Prezzo > 0 THEN PrezzoA = PrezzoB PrezzoB = PrezzoC PrezzoC = Prezzo ENDIF If longonmarket and (barindex-tradeindex)>=43 then SELL AT MARKET ENDIF LossMonth =200 once StopMonth=1 IF Month <> Month[1] THEN ProfitMonth=strategyprofit StopMonth=1 endif If strategyprofit<=ProfitMonth-LossMonth then StopMonth=0 endif Cond = PrezzoB > PrezzoA AND PrezzoB > PrezzoC IF StopMonth=1 and close CROSSES UNDER PrezzoB AND C2 AND C1 AND C30 AND Cond > 0 THEN BUY 1 Contract At Market SET TARGET PROFIT 5* averagetruerange[1](close) SET STOP %LOSS 8 ENDIF Trailinglong = 2* averagetruerange[5](close) Trailingsteplong = 5* averagetruerange[1](close) TGL = Trailinglong STPL = Trailingsteplong if not onmarket then MAXPRICE = 0 PREZZOUSCITA = 0 ENDIF if longonmarket then MAXPRICE = MAX(MAXPRICE,close) if MAXPRICE-tradeprice(1)>=TGL then PREZZOUSCITA = MAXPRICE-STPL ENDIF ENDIF if onmarket and PREZZOUSCITA>0 then EXITSHORT AT PREZZOUSCITA STOP SELL AT PREZZOUSCITA STOP ENDIFI ran SEB tests on the code with the new TR filter.
Without the time based selling the code performs terribly in the SEB test. With time based selling there is not much to see because the normal backtest sees almost all the possible trades due to the low number of trades.
What this tells us is that the strategy is very dependent on buying and holding and selling after a certain amount of time in what has been a very up trending market during the test period. Buying and holding will always work in this market!
Without the time based selling it just gets lucky in the normal back test and hits all the good trades and misses all the bad trades because it is already on the market.
- You must be logged in to reply to this topic.
Author
Summary
This topic contains 18 replies,
has 4 voices, and was last updated by
5 years, 8 months ago.
Topic Details
| Forum: | ProOrder: Automated Strategies & Backtesting |
| Language: | English |
| Started: | 06/27/2020 |
| Status: | Active |
| Attachments: | 14 files |
