Algo against the trend…ouch

• 05/24/2017 at 8:11 PM #36593
  Inertia

  Dear Traders,

  Kindly allow me to share the following:

  My 2 very simple strategies (Long only – UT 10 sec and UT 10 mn) showed relatively great results on the back test, tick by tick, 200.000 bars and 1 pt as per the spread (9:00 to 17:30) on PRT 10.3, no 0 bars on the report … so I’d put it on real on the DAX 1EUR contract, since May 9th 2017…And guess what ? It doesn’t work as expected and I lost a bit of money… well 22% of my small portfolio. No worries, I will recover 😉

  Here is my deduction – Correct me if I am wrong PLEASE: “Algos”, what ever it is or complicated they are, should be started in the same direction as the major trend. Yeah I was probably to excited to refrain myself no to give it a go !

  So today I have changed a bit my desktop (attached to this message) to manage this/ my issue.

  You can see the Dax + the chart of Siemens (most weighted stock of the Dax – all UT on daily + HA candle style…)

  The result is pretty clear: I pressed the STOP button today on Pro order Auto Trading before my account explodes…

  I red in this forum that multiple UT would be available ? This could be the solution to my issue I suppose…

  Any thoughts would be more than welcome !

  Thank you and have a good WE !

  And great forum by the way.

   

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/24/2017 at 8:29 PM #36595
  Eric

  next time?

  backtest > demo money > real money

  1 user thanked author for this post.

  Inertia

  

  Eric

  Participant

  Topics: 18

  Replies: 441

  Been thanked: 84 times
  05/25/2017 at 4:13 AM #36603
  Wing

  That is a low timeframe you are using, and you are placing many trades per day. Higher timeframes will tend to have more signals versus noise, and be less ‘competitive’.

  I don’t know what your backtest looked like, or how you optimized your systems, but that is where the culprit could be. You must also factor in randomness. With 9 wins and 18 losses, can you really judge the system yet? If chance of winning trade is around 50 %… flip a coin 27 times and you may very well end up with at least 18 heads. In fact, the chance of that happening is 6%+. Since markets and systems do not behave entirely random, wins or losses can tend to come in groups, so the chance is even higher. Double the amount of test trades/coin flips to 54, and the chance of 36 or more heads is below 1%

  Of course, as mentioned above, it is good to first test systems on a demo account.

  And yes, systems are usually improved if you introduce some trend filters so you have the market in a good direction.

  1 user thanked author for this post.

  Inertia

  

  Wing

  Participant

  Topics: 6

  Replies: 159

  Been thanked: 76 times
  05/25/2017 at 5:24 AM #36604
  Inertia

  Dear Wing,

  Hereafter are the back tests.

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/25/2017 at 5:42 AM #36607
  Inertia

  Dear Wing,

  It absolutely personal but I do believe that you can make monthly money in trading only in using UTs equal or under 15mn. Higher than that, it might be mostly consider as an “investor”. Why not ? no worries about that…;) Maybe I am wrong…But yes, perhaps 10sec it’s quiet small…let’s see.

  The strategy is with 2 ema 150 and 450 (filters) 1 osc that is suppose to trigger that strategy above a specific level, 1 stoploss and 1 target…that’s it. Both strategies are mostly the same…

  Cheers,

  DJ

   

   

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/25/2017 at 10:19 AM #36612
  GraHal

  Hi DJ

  I trade 1 min timeframe with entries / exits on 6 sec TF, but on manual trading. I’ve tried bots on 1 min and 10 seconds but got similar results to yourself. I put it down to my inability to code everything in that I take into account with my eyes.

  My entries / exits are dependent on market structure … hh, hl, ll, lh, reversal patterns, support. resistance, elliott waves, retracement etc … all the factors that are very difficult to code up?

  So often when analysing bot trades I think, as an example  … if only the bot knew that short was opened a few points off a 61.8 fib reversal (from a high)  … and so I manual close the bot with a small loss (£2 – £3) to prevent a bigger loss,

  What is the max number of bars and min number of bars that your 10 second bot trades … better still do a screen shot of No of Bars sorted max top and min top.

  Or better still add your code on here using the <> icon. I’d love to try it … I’m a keen short timeframe trader … rightly or wrongly, it suits my personality! 🙂

  Cheers
  GraHal

  PS  You could even email me your Bot via Nicolas if you wish? It’s okay I try some ‘whacky stuff’ in my Bots and often I don’t want to defend my whackiness publicly! 🙂 Maybe we are a couple of Mavericks?? 🙂 🙂 🙂

  1 user thanked author for this post.

  Inertia

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  05/25/2017 at 3:57 PM #36641
  Inertia

  Thank you GraHal for your kind answer ;). Appreciated.

  I do like your idea for entries and exits…and a chance that all “traders” do not use the same TF…

   

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/25/2017 at 5:30 PM #36645
  GraHal

  Looking at your screengrabs of max and min number of bars it would seem to suggest that if your bot is not in profit within a few 10 sec bars (few = ??) then you may as well exit anyway? You have loads of £9 wins within 4 or 5 mins of running.

  I run of 1 min bot off here (I tweaked some variables) and if it takes off in profit within a few seconds then it does well, if not then it drifts down to eventually close at  a loss … like a lot of yours are doing?

  Okay maybe on Backtest max profit is seen by stop loss settings you (and I) are running, but I’ve come round to thinking I prefer less stress, even if less profit overall! Quick wins put me in a good mood for the day! 🙂 🙂

  Just a few thoughts anyway
  GraHal
  PS I’ll (try) and do what I suggest above with the 1 min bot I run and let you know.

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  05/25/2017 at 7:20 PM #36652
  Inertia

  // Définition des paramètres du code DEFPARAM CumulateOrders = False // Cumul des positions désactivé DEFPARAM Preloadbars = 10000 // Annule tous les ordres en attente et ferme toutes les positions à 0:00, puis empêche toute création d'ordre avant l'heure "FLATBEFORE". DEFPARAM FLATBEFORE = 090000 // Annule tous les ordres en attente et ferme toutes les positions à l'heure "FLATAFTER" DEFPARAM FLATAFTER = 170000 // Empêche le système de placer des ordres pour entrer sur le marché ou augmenter la taille d'une position avant l'heure spécifiée timeEnterBefore = time >= 090000 // Empêche le système de placer des ordres pour entrer sur le marché ou augmenter la taille d'une position après l'heure spécifiée timeEnterAfter = time < 170000 // Empêche le système de placer de nouveaux ordres sur les jours de la semaine spécifiés daysForbiddenEntry = OpenDayOfWeek = 6 OR OpenDayOfWeek = 0 // Conditions pour ouvrir une position acheteuse indicator1 = Williams[5](close) c1 = (indicator1 CROSSES OVER -50) indicator2 = Williams[5](close) c2 = (indicator2 > indicator2[3]) indicator3 = ExponentialAverage[450](close) c3 = (close > indicator3) indicator4 = ExponentialAverage[150](close) indicator5 = ExponentialAverage[150](close) c4 = (indicator4 > indicator5[2]) IF (c1 AND c2 AND c3 AND c4) AND timeEnterBefore AND timeEnterAfter AND not daysForbiddenEntry THEN BUY 1 CONTRACT AT MARKET ENDIF // Stops et objectifs SET STOP pLOSS 19 SET TARGET pPROFIT 9

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

  // Définition des paramètres du code DEFPARAM CumulateOrders = False // Cumul des positions désactivé   DEFPARAM Preloadbars = 10000   // Annule tous les ordres en attente et ferme toutes les positions à 0:00, puis empêche toute création d'ordre avant l'heure "FLATBEFORE". DEFPARAM FLATBEFORE = 090000 // Annule tous les ordres en attente et ferme toutes les positions à l'heure "FLATAFTER" DEFPARAM FLATAFTER = 170000   // Empêche le système de placer des ordres pour entrer sur le marché ou augmenter la taille d'une position avant l'heure spécifiée timeEnterBefore = time >= 090000   // Empêche le système de placer des ordres pour entrer sur le marché ou augmenter la taille d'une position après l'heure spécifiée timeEnterAfter = time < 170000   // Empêche le système de placer de nouveaux ordres sur les jours de la semaine spécifiés daysForbiddenEntry = OpenDayOfWeek = 6 OR OpenDayOfWeek = 0   // Conditions pour ouvrir une position acheteuse indicator1 = Williams[5](close) c1 = (indicator1 CROSSES OVER -50) indicator2 = Williams[5](close) c2 = (indicator2 > indicator2[3]) indicator3 = ExponentialAverage[450](close) c3 = (close > indicator3) indicator4 = ExponentialAverage[150](close) indicator5 = ExponentialAverage[150](close) c4 = (indicator4 > indicator5[2])   IF (c1 AND c2 AND c3 AND c4) AND timeEnterBefore AND timeEnterAfter AND not daysForbiddenEntry THEN BUY 1 CONTRACT AT MARKET ENDIF   // Stops et objectifs SET STOP pLOSS 19 SET TARGET pPROFIT 9

   

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/25/2017 at 7:41 PM #36657
  Inertia

  Correct…The idea behind this (scalping) strategy is to capture the “explosion”. Only small TF can sufficiently break down the data to catch the acceleration.

  Correct again…after couple of bars, thank you to your analysis, better exit for new opportunities …

  I just took the liberty to attached the DAX of today in 3 different time frame ’10sec, 1mn and 10 mn. As we are all working on “BarClose”, and we try to enter and exit like highly qualify neurological surgeon doc 😉 in the market, yes indeed I do prefer 10 sec up to 1 mn bar chart ;). It feets my style too…Woohoo.

  So from my side I’ll see what I can add to this strategy…SAR or EMA or SuperTrend to exit nice and clean…Let’s see.

  Thx Grahal

  1 user thanked author for this post.

  GraHal

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/25/2017 at 8:41 PM #36662
  GraHal

  You may like a change and spend time trying to get clean swift exits after explosions on a new bot? Import attached .itf file and see results attached for 100,000 bars.  Even if set DEFPARAM CumulateOrders = False so as not to accumulate orders it is still respectable performance.

  I have tried to get the Author to cough up 🙂 his Short version of the code, but he’s not on here much these days. If you are any good maybe you can make a short version?

  I am thinking of moving towards two versions generally rather than  long and short sides in same version. Reason – when in a clear uptrend I can launch the Long version and vice versa … I think it be less stressful?? 🙂

  Just maybe Nicolas (or another skilled Developer) may spot this and weave his magic on GForton code attached to make a short version?? I have to admit, although I squeezed more profit out of it than Glen Forton, my head gets in a spin trying to understand it all enough to make a short version.

  Attached is on Dax 1 min btw and a 2 point spread as a worst case scenario,

  Cheers
  GraHal

   

  1 user thanked author for this post.

  Inertia

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  05/26/2017 at 4:24 PM #36718
  Inertia

  GraHal,

  I had this strategy copied in PRT from the forum – Thank you to the author – and just tried it in 1mn TF… instead of TH 12hr… Can you try it from your side please ???

   

  //----------------------------------------------------------------------------------------------- //TF a 12 h //dax min 1 Eur. //----------------------------------------------------------------------------------------------- //TAKE PROFIT NEAR ZERO IN ORDER TO NOT ACCUMULATE TOO MUCH //----------------------------------------------------------------------------------------------- //IF NOT ON MARKET TRY TO ENTER AGAIN AS SOON A CORRECTION APPEARS, TRYING TO FOLLOW //THE POTENTIAL CURRENT UPTREND //----------------------------------------------------------------------------------------------- //IF I BOUGHT THE PREVIOUS BAR DON'T BUY THE FOLLOWING BAR IN ORDER TO NOT ACCUMULATE TOO MUCH //----------------------------------------------------------------------------------------------- //MAIN PARAMETER IN ORDER TO CHANGE THE RISK LEVEL ARE: //acc----> % of decrease of the instrument compared to the last buy price (THE LOWER IS AND MORE // RISK AND CASH RETURN WILL APEEAR) //----------------------------------------------------------------------------------------------- //tp-----> % beyond which the entire position is closed (THE HIGHER IS AND MORE // RISK AND CASH RETURN WILL APEEAR) - below zero is safer ! //----------------------------------------------------------------------------------------------- //MONEY MANAGENT "ANTIMARTINGALA" STYLE - I SUPPOSE CLEARER RESULTS ON A LONGER PERIOD OF BACKTEST //----------------------------------------------------------------------------------------------- DEFPARAM cumulateorders = true TIMEWORK = 070000 TIMESTOP = 210000 acc = 5 if justone = 0 then capital = 10000 * pointvalue justone = 1 endif perccap = 0.5 margin = 0.5 * close * pointvalue / 100 period=5 stdev=STD[period](close) //LOOK FOR A LOCAL VOLATILITY DECREASE //I WANT TO BUY AGAIN ONLY WHEN BEAR MARKET SHOWS A LOCAL PAUSE if (stdev[0]+stdev[1]+stdev[2]+stdev[3]+stdev[4]) < (stdev[1]+stdev[2]+stdev[3]+stdev[4]+stdev[5]) then SWBUY=1 else SWBUY=0 ENDIF IF TIME > TIMEWORK AND TIME < TIMESTOP THEN if not onmarket then sizebuy = ROUND((capital *perccap /100)/ margin) tp = 0.25 IF CLOSE < CLOSE[1] THEN BUY sizebuy SHARES AT MARKET endif endif if LONGONMARKET then perf = (close - POSITIONPRICE)/POSITIONPRICE * 100 if perf > tp then SELl AT MARKET perfc = (close - POSITIONPRICE) * POINTVALUE * COUNTOFPOSITION capital = capital + PERFC ELSE IF LASTBARBUY = 0 THEN if (((CLOSE-TRADEPRICE)/TRADEPRICE)*100) < -acc AND SWBUY=1 then sizebuy = ROUND((capital *perccap /100)/ margin) BUY sizebuy SHARES AT MARKET LASTBARBUY = 1 ENDIF ELSE LASTBARBUY = 0 ENDIF ENDIF ENDIF ENDIF

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

  //----------------------------------------------------------------------------------------------- //TF a 12 h //dax min 1 Eur. //----------------------------------------------------------------------------------------------- //TAKE PROFIT NEAR ZERO IN ORDER TO NOT ACCUMULATE TOO MUCH //----------------------------------------------------------------------------------------------- //IF NOT ON MARKET TRY TO ENTER AGAIN AS SOON A CORRECTION APPEARS, TRYING TO FOLLOW //THE POTENTIAL CURRENT UPTREND //----------------------------------------------------------------------------------------------- //IF I BOUGHT THE PREVIOUS BAR DON'T BUY THE FOLLOWING BAR IN ORDER TO NOT ACCUMULATE TOO MUCH //----------------------------------------------------------------------------------------------- //MAIN PARAMETER IN ORDER TO CHANGE THE RISK LEVEL ARE: //acc----> % of decrease of the instrument compared to the last buy price (THE LOWER IS AND MORE // RISK AND CASH RETURN WILL APEEAR) //----------------------------------------------------------------------------------------------- //tp-----> % beyond which the entire position is closed (THE HIGHER IS AND MORE // RISK AND CASH RETURN WILL APEEAR) - below zero is safer ! //----------------------------------------------------------------------------------------------- //MONEY MANAGENT "ANTIMARTINGALA" STYLE - I SUPPOSE CLEARER RESULTS ON A LONGER PERIOD OF BACKTEST //-----------------------------------------------------------------------------------------------   DEFPARAM cumulateorders = true   TIMEWORK = 070000 TIMESTOP = 210000 acc = 5   if justone = 0 then capital = 10000 * pointvalue justone = 1 endif   perccap = 0.5 margin = 0.5 * close * pointvalue / 100 period=5 stdev=STD[period](close) //LOOK FOR A LOCAL VOLATILITY DECREASE //I WANT TO BUY AGAIN ONLY WHEN BEAR MARKET SHOWS A LOCAL PAUSE if (stdev[0]+stdev[1]+stdev[2]+stdev[3]+stdev[4]) < (stdev[1]+stdev[2]+stdev[3]+stdev[4]+stdev[5]) then SWBUY=1 else SWBUY=0 ENDIF   IF TIME > TIMEWORK AND TIME < TIMESTOP THEN if not onmarket then   sizebuy = ROUND((capital *perccap /100)/ margin)   tp = 0.25   IF CLOSE < CLOSE[1] THEN    BUY sizebuy SHARES AT MARKET   endif endif   if LONGONMARKET then   perf = (close - POSITIONPRICE)/POSITIONPRICE * 100   if perf > tp then    SELl AT MARKET    perfc = (close - POSITIONPRICE) * POINTVALUE * COUNTOFPOSITION    capital = capital + PERFC   ELSE    IF LASTBARBUY = 0 THEN     if (((CLOSE-TRADEPRICE)/TRADEPRICE)*100) < -acc AND SWBUY=1 then      sizebuy = ROUND((capital *perccap /100)/ margin)      BUY sizebuy SHARES AT MARKET      LASTBARBUY = 1     ENDIF    ELSE     LASTBARBUY = 0    ENDIF   ENDIF ENDIF ENDIF

   

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/26/2017 at 4:41 PM #36719
  Inertia

  GraHal,

  GForton: Mavericks or not Mavericks, that is the question… // I’ll give it a go next week if time permits but I will be near the screen just in case 😉

  Thx and have a good WE Sir !

  DJ

  

  Inertia

  Participant

  Topics: 40

  Replies: 223

  Been thanked: 52 times
  05/26/2017 at 5:46 PM #36723
  GraHal

  I think it keeps trades open (thro ups and downs) until it reaches some pre-set profit level?? The Strategy equity curve follows the Price Curve?? How would we sleep during the big drawdowns?? 🙂

  Should line 28 be corrected to this …

  If justone = 0 then

  1	If justone = 0 then

   

   

   

  1 user thanked author for this post.

  Inertia

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  05/27/2017 at 9:49 AM #36743
  JC_Bywan

  Hi Inertia,

  We have edited your previous message to format it for you with PRT code

  >> For clarity of messages on ProRealCode’s forums, please use the “<>” button (insert PRT code) to separate the text part from the code part! Thank you <<

  1 user thanked author for this post.

  Inertia

  

  JC_Bywan

  Moderator

  Topics: 20

  Replies: 2552

  Been thanked: 842 times
• Author
  Posts