Create strategies codes with ChatGPT for ProRealTime

• 05/31/2023 at 9:20 AM #215486
  TraderGlyn

  GraHal wrote:

  TraderGlyn it be good to start a new Topic for Bard.google.com coded Algo’s etc?

  Have you got an Algo you can post (in a new Topic), it be very interesting to see the results from google AI?

  I’m not a skilled coder and don’t have my own algo unfortunately.

  Bard, at least for now, appears unable to write ProRealCode even after giving it a sample of correct code to learn from. I was surprised as Bard AI is described as a complex language model capable of “machine learning”. In this respect it seems way behind ChatGPT 4 currently.
  With that said, this technology surely holds incredibly useful potential for us traders.

  1 user thanked author for this post.

  GraHal

  

  TraderGlyn

  Participant

  Topics: 3

  Replies: 12

  Been thanked: 5 times
  05/31/2023 at 7:35 PM #215517
  Monochrome

  Hi all, I am back after a long break. Chatgpt got me interested back into experimenting with strategies. It makes starting up a workflow so much more easier.

  Here’s a simple strategy I asked it to make with Nichola’s prompt(with many thanks)

  Seems to do pretty well. Alteast to help in formalising the initial strategy

  defparam cumulateOrders = false defparam flatafter = 235959 // Prevent any overnight trades // Define opening period startHour = 80000 // DAX opens at 8:00 local time endHour = 110000 // Ends at 11:00 local time // Variable to keep track of the high and low since 1am once highestSince1am = high once lowestSince1am = low // Update the high and low since 1am every bar if time >= 10000 and time < startHour then highestSince1am = max(highestSince1am, high) lowestSince1am = min(lowestSince1am, low) endif // If current time is within the opening hours and price crosses over the highest price since 1am, then buy if time >= startHour and time <= endHour and close > highestSince1am then buy at market set stop ploss abs(low[1]-close) / pointsize set target pprofit 50 * pointsize endif // If current time is within the opening hours and price crosses under the lowest price since 1am, then sellshort if time >= startHour and time <= endHour and close < lowestSince1am then sellshort at market set stop ploss abs(high[1]-close) / pointsize set target pprofit 50 * pointsize endif // Reset the highest and lowest variables for the next day if time >= 235959 then highestSince1am = high lowestSince1am = low 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

  defparam cumulateOrders = false defparam flatafter = 235959  // Prevent any overnight trades   // Define opening period startHour = 80000  // DAX opens at 8:00 local time endHour = 110000  // Ends at 11:00 local time   // Variable to keep track of the high and low since 1am once highestSince1am = high once lowestSince1am = low   // Update the high and low since 1am every bar if time >= 10000 and time < startHour then highestSince1am = max(highestSince1am, high) lowestSince1am = min(lowestSince1am, low) endif   // If current time is within the opening hours and price crosses over the highest price since 1am, then buy if time >= startHour and time <= endHour and close > highestSince1am then buy at market set stop ploss abs(low[1]-close) / pointsize set target pprofit 50 * pointsize endif   // If current time is within the opening hours and price crosses under the lowest price since 1am, then sellshort if time >= startHour and time <= endHour and close < lowestSince1am then sellshort at market set stop ploss abs(high[1]-close) / pointsize set target pprofit 50 * pointsize endif   // Reset the highest and lowest variables for the next day if time >= 235959 then highestSince1am = high lowestSince1am = low endif

  

  Monochrome

  Participant

  Topics: 17

  Replies: 116

  Been thanked: 56 times
  06/01/2023 at 7:53 AM #215526
  Nicolas

  Thank you @monochrome

  and what was your prompt to get a so nicely coded strategy please?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  06/01/2023 at 8:08 PM #215562
  Monochrome

  This was my initial ask –
  Code me a strategy for DAX in the 5 min timeframe. During the opening 3 hours buy when price crosses over the highest price since 1am and vice versa for sell. Profit target is 5pips and stop loss is the low of the previous bar

  I had to correct it a few times by saying for eg –
  – there is no when function, recode the dax strategy
  – this code does not work

  I think we can make the prompt better by giving it most commonly used functions and a few simple examples.

  Still testing

  this is what it gave me initially –

  defparam cumulateOrders = false defparam flatafter = 235959 // prevent any overnight trades TimeFrame = 5 // define opening period startHour = 080000 // DAX opens at 8:00 local time endHour = 110000 // ends at 11:00 local time priorHour = 010000 // 1:00 am as per your request // variables for highest price and lowest price since 1 am highestPrice = max[BarIndex](high) when time >= priorHour and time < startHour lowestPrice = min[BarIndex](low) when time >= priorHour and time < startHour // if current time is within the opening hours and price crosses over the highest price since 1am then buy if time >= startHour and time <= endHour and close crosses over highestPrice then buy at market set stop ploss abs(low[1]-close) / pointsize set target pprofit 5 * pointsize endif // if current time is within the opening hours and price crosses under the lowest price since 1am then sellshort if time >= startHour and time <= endHour and close crosses under lowestPrice then sellshort at market set stop ploss abs(high[1]-close) / pointsize set target pprofit 5 * pointsize 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

  defparam cumulateOrders = false defparam flatafter = 235959  // prevent any overnight trades TimeFrame = 5   // define opening period startHour = 080000  // DAX opens at 8:00 local time endHour = 110000  // ends at 11:00 local time priorHour = 010000 // 1:00 am as per your request   // variables for highest price and lowest price since 1 am highestPrice = max[BarIndex](high) when time >= priorHour and time < startHour lowestPrice = min[BarIndex](low) when time >= priorHour and time < startHour   // if current time is within the opening hours and price crosses over the highest price since 1am then buy if time >= startHour and time <= endHour and close crosses over highestPrice then     buy at market     set stop ploss abs(low[1]-close) / pointsize     set target pprofit 5 * pointsize endif   // if current time is within the opening hours and price crosses under the lowest price since 1am then sellshort if time >= startHour and time <= endHour and close crosses under lowestPrice then     sellshort at market     set stop ploss abs(high[1]-close) / pointsize     set target pprofit 5 * pointsize endif

  

  Monochrome

  Participant

  Topics: 17

  Replies: 116

  Been thanked: 56 times
  06/02/2023 at 8:18 AM #215570
  Nicolas

  Monochrome wrote:

  I think we can make the prompt better by giving it most commonly used functions and a few simple examples.

  Yes, that’s what I’m looking to improve, the intial prompt. That’s why we all should share our experience with wrong results and how we have resolved them with other specificitions given to ChatGPT.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  06/04/2023 at 4:02 AM #215625
  crankbuddy

  I asked ChatGPT  “I want sell when the price is 20% above the startpoint” and listed my code and got my result. I then asked “can I improve the macd” I was given a few improvements to my code and the end result. I then asked “show me code for divergence analysis” based on one of the end results. I cleaned up the code and listed below. I’ll back test and test on demo this week.

  // Definition of code parameters DEFPARAM CumulateOrders = False // Cumulating positions deactivated // Calculate MACD values MACDLinePeriods = 12 SignalLinePeriods = 26 MACDAveragePeriods = 9 MLine = MACDline[MACDLinePeriods, SignalLinePeriods, MACDAveragePeriods](close) SignalLine = MACDSignal[MACDLinePeriods, SignalLinePeriods, MACDAveragePeriods](close) // Calculate price and MACD divergence PriceDivergence = close - MACDLine // Conditions to enter long positions Indicator1 = SignalLine Indicator2 = MLine Indicator3 = (Indicator2 CROSSES OVER 0) DivergenceCondition = (PriceDivergence[1] < 0 AND PriceDivergence > 0) c1 = (Indicator1 < Indicator2) c2 = Indicator3 StartPoint = TRADEPRICE IF c1 AND c2 AND DivergenceCondition THEN BUY 1 CONTRACT AT MARKET set stop ploss abs(low[1]-close) / pointsize set target pprofit 50 * pointsize ENDIF // Conditions to exit long positions Indicator4 = MACD CROSSES UNDER 0[3] ExitCondition = (PriceDivergence[1] > 0 AND PriceDivergence < 0) c3 = (Indicator1 > Indicator2) c4 = Indicator4 c5 = (close >= StartPoint * 1.2) // Price is 20% above the trade price DivergenceExitCondition = (c4 OR c5) AND ExitCondition IF c3 AND DivergenceExitCondition THEN SELL AT MARKET set stop ploss abs(high[1]-close) / pointsize set target pprofit 50 * pointsize 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

  // Definition of code parameters DEFPARAM CumulateOrders = False // Cumulating positions deactivated   // Calculate MACD values MACDLinePeriods = 12 SignalLinePeriods = 26 MACDAveragePeriods = 9   MLine = MACDline[MACDLinePeriods, SignalLinePeriods, MACDAveragePeriods](close) SignalLine = MACDSignal[MACDLinePeriods, SignalLinePeriods, MACDAveragePeriods](close)   // Calculate price and MACD divergence PriceDivergence = close - MACDLine   // Conditions to enter long positions Indicator1 = SignalLine Indicator2 = MLine Indicator3 = (Indicator2 CROSSES OVER 0) DivergenceCondition = (PriceDivergence[1] < 0 AND PriceDivergence > 0)   c1 = (Indicator1 < Indicator2) c2 = Indicator3 StartPoint = TRADEPRICE   IF c1 AND c2 AND DivergenceCondition THEN BUY 1 CONTRACT AT MARKET set stop ploss abs(low[1]-close) / pointsize set target pprofit 50 * pointsize ENDIF   // Conditions to exit long positions Indicator4 = MACD CROSSES UNDER 0[3] ExitCondition = (PriceDivergence[1] > 0 AND PriceDivergence < 0)   c3 = (Indicator1 > Indicator2) c4 = Indicator4 c5 = (close >= StartPoint * 1.2) // Price is 20% above the trade price DivergenceExitCondition = (c4 OR c5) AND ExitCondition   IF c3 AND DivergenceExitCondition THEN SELL AT MARKET set stop ploss abs(high[1]-close) / pointsize set target pprofit 50 * pointsize ENDIF

   

  2 users thanked author for this post.

  Meta Signals Pro, Nicolas

  

  crankbuddy

  Participant

  Topics: 2

  Replies: 3

  Been thanked: 4 times
  06/09/2023 at 11:29 AM #215899
  GraHal

  Anybody tried to get ChatGPT to create a Machine Learning Algo.

  

  GraHal

  Participant

  Topics: 211

  Replies: 8754

  Been thanked: 2172 times
  06/10/2023 at 2:39 PM #215951
  bertrandpinoy

  indeed I.A allows to progress the personal coding, on the other hand I am surprised that the remarks which are made to him on the errors start again the next day. it doesn’t seem to save corrections or bad elements from languages outside of PRT.

  

  bertrandpinoy

  Participant

  Topics: 24

  Replies: 121

  Been thanked: 8 times
  06/10/2023 at 3:23 PM #215952
  GraHal

  bertrandpinoy wrote:

  it doesn’t seem to save corrections or bad elements

  Let’s call your Chat – Chat No 1.

  ChatGPT saves the corrections and bad elements you advised about and got sorted, but only remembers within Chat No 1.

  If you start another Chat – Chat No 2, then there is no link across / memory from Chat No 1.

  1 user thanked author for this post.

  bertrandpinoy

  

  GraHal

  Participant

  Topics: 211

  Replies: 8754

  Been thanked: 2172 times
  06/15/2023 at 3:02 PM #216285
  ProRealAlgos

  I made an algo today, approximately 6 months after making my first algo with ChatGPT. I used the prompts provided here in the forum thread, but it’s still not really making plug-and-playable code for ProRealTime.
  The first issue with ChatGPT was that it insisted to use underscores in the variable names, which is not accepted by ProBuilder language. I think it’s an easy instruction to add – but I have also noticed that ChatGPT seems to have issue when providing too many instructions.

  Here’s the video of me making it and the full code it generated.
  https://youtu.be/YciwTN4ZquU

  Here’s the full code

  defparam cumulateorders = false // Define RSI parameters RSIPeriod = 2 RSIOversold = 15 RSIOverbought = 20 // Calculate RSI RSIValue = rsi[RSIPeriod] //Day-of-the-week filter DayOfWeekFilter = dayofweek FilterDay = 5 // Friday // Monthly filter if month = 0 or month = 1 or month = 2 or month = 3 or month = 4 or month = 10 or month = 11 or month = 12 then LongMonths = 1 else LongMonths = 0 endif if month = 5 or month = 6 or month = 7 or month = 8 or month = 9 then AvoidMonths = 1 else AvoidMonths = 0 endif // Entry condition if RSIValue crosses under RSIOversold and DayOfWeekFilter = FilterDay and Longmonths then buy at market endif if RSIValue crosses over RSIOverbought and DayOFweekFilter = Filterday and not AvoidMonths then sell at market 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

  defparam cumulateorders = false   // Define RSI parameters RSIPeriod = 2 RSIOversold = 15 RSIOverbought = 20   // Calculate RSI RSIValue = rsi[RSIPeriod]   //Day-of-the-week filter DayOfWeekFilter = dayofweek FilterDay = 5 // Friday   // Monthly filter   if month = 0 or month = 1 or month = 2 or month = 3 or month = 4 or month = 10 or month = 11 or month = 12 then LongMonths = 1 else LongMonths = 0 endif   if month = 5 or month = 6 or month = 7 or month = 8 or month = 9 then AvoidMonths = 1 else AvoidMonths = 0 endif   // Entry condition if RSIValue crosses under RSIOversold and DayOfWeekFilter = FilterDay and Longmonths then buy at market endif   if RSIValue crosses over RSIOverbought and DayOFweekFilter = Filterday and not AvoidMonths then sell at market endif

   

  1 user thanked author for this post.

  GraHal

  

  ProRealAlgos

  Participant

  Topics: 6

  Replies: 74

  Been thanked: 28 times
  06/16/2023 at 7:07 PM #216365
  ProRealAlgos

  I actually see now that the initial prompt had instructions not to have underscores in the variable names. It really proves that ChatGPT still have issues handling too many instructions in the same prompt.

  1 user thanked author for this post.

  GraHal

  

  ProRealAlgos

  Participant

  Topics: 6

  Replies: 74

  Been thanked: 28 times
  06/17/2023 at 12:01 PM #216381
  gonzocorp

  Thank you Nicholas for this post, it’s just incredible and all informations shared here are just real gold mine. I’m starting my own test to improve my own PRC strategies ! I will probably share my feedback in future.

  2 users thanked author for this post.

  Nicolas, GraHal

  

  gonzocorp

  Participant

  Topics: 0

  Replies: 1

  Been thanked: 2 times
  07/05/2023 at 12:04 PM #217304
  AMQ

  Hi all!

  I would like to convert code for trailing stop, based on parameters, in code based on action price. To get that, I used de first prompt and then:

   

  “trailing stop” is a stop loss that is moving near the price, based in rules included in the code, in order to exit the operation if the price turns back against the position. Answer “YES” if you can understand the meaning of “trailing stop”

  ChatGPT

  YES

   

  

  I will paste “trailing stop” code and i will ask you to write probuilder code based in the minimum and maximum of prior candlesticks in timeframe 5 min. Answer “YES” if you are ready me to paste the “trailing stop” code

  ChatGPT

  YES, you can go ahead and paste the “trailing stop” code.

  The result is different to the objetive, so i would like to ask you if you suggest a better prompt.

  Trailing stop Code is the next one for 2 min TF:

  // — 0. Variables

  pp = (positionperf*100)

  // — 1. Trailing stop

  once enablets = 1 // activar trailing stop
  // trailing stop

  if shortonmarket then
  ts1= 0.5 // es mayor que este porcentaje, entonces utiliza trailingstop ts1
  switch= 0.55 // es mayor que este porcentaje, entonces utiliza trailingstop ts2
  ts2= 0.35
  switch2= 0.7 // es mayor que este porcentaje, entonces utiliza trailingstop ts3
  ts3= 0.3
  endif

  if longonmarket then
  ts1= 0.48//0.55//0.48 // es mayor que este porcentaje, entonces utiliza trailingstop ts1
  switch=1.03//1.05//0.4 // es mayor que este porcentaje, entonces utiliza trailingstop ts2
  ts2=0.35//0.85
  switch2= 1.3//1.75 // es mayor que este porcentaje, entonces utiliza trailingstop ts3
  ts3=0.25//0.2
  endif

  pp = (positionperf*100)
  if enablets then

  trailingstop1 = (tradeprice(1)/100)*ts1
  trailingstop2 = (tradeprice(1)/100)*ts2
  trailingstop3 = (tradeprice(1)/100)*ts3

  if not onmarket or ((longonmarket and shortonmarket[1]) or (longonmarket[1] and shortonmarket)) then
  maxprice1=0
  minprice1=close
  priceexit1=0

  maxprice2=0
  minprice2=close
  priceexit2=0

  maxprice3=0
  minprice3=close
  priceexit3=0

  a1 = 0
  a2 = 0
  a3 = 0

  pp=0

  endif

  if onmarket then
  if pp >= ts1 then
  a1=1
  endif
  if pp >= switch then
  a2=1
  endif
  if pp >= switch2 then
  a3=1
  endif
  endif

  // setup long

  if longonmarket then
  maxprice1=max(maxprice1,close)
  maxprice2=max(maxprice2,high)
  maxprice3=max(maxprice3,high)
  if a1 then
  if maxprice1-tradeprice(1)>=(trailingstop1)*pointsize then
  priceexit1=maxprice1-(trailingstop1)*pointsize
  endif
  endif
  if a2 then
  if maxprice2-tradeprice(1)>=(trailingstop2)*pointsize then
  priceexit2=maxprice2-(trailingstop2)*pointsize
  endif
  endif
  if a3 then
  if maxprice3-tradeprice(1)>=(trailingstop3)*pointsize then
  priceexit3=maxprice3-(trailingstop3)*pointsize
  endif
  endif
  endif

  // setup short

  if shortonmarket then
  minprice1=min(minprice1,close)
  minprice2=min(minprice2,low)
  minprice3=min(minprice3,low)
  if a1 then
  if tradeprice(1)-minprice1>=(trailingstop1)*pointsize then
  priceexit1=minprice1+(trailingstop1)*pointsize
  endif
  endif
  if a2 then
  if tradeprice(1)-minprice2>=(trailingstop2)*pointsize then
  priceexit2=minprice2+(trailingstop2)*pointsize
  endif
  endif
  if a3 then
  if tradeprice(1)-minprice3>=(trailingstop3)*pointsize then
  priceexit3=minprice3+(trailingstop3)*pointsize
  endif
  endif
  endif

  // exit long

  if longonmarket and priceexit1>0 then
  sell at priceexit1 stop
  endif
  if longonmarket and priceexit2>0 then
  sell at priceexit2 stop
  endif
  if longonmarket and priceexit3>0 then
  sell at priceexit3 stop
  endif

  // exit short

  if shortonmarket and priceexit1>0 then
  exitshort at priceexit1 stop
  endif
  if shortonmarket and priceexit2>0 then
  exitshort at priceexit2 stop
  endif
  if shortonmarket and priceexit3>0 then
  exitshort at priceexit3 stop
  endif
  endif

  Many thanks.

  Alfonso

   

  

  AMQ

  Participant

  Topics: 8

  Replies: 18

  Been thanked: 2 times
  07/05/2023 at 2:30 PM #217326
  GraHal

  What part of the above code (if any?) was provided by ChatGPT?

  If you use the Insert PRT Code button (far right top line above each new message – blue button) then Line Numbers will show in posted code … makes it easier for us all. 🙂

   

  

  GraHal

  Participant

  Topics: 211

  Replies: 8754

  Been thanked: 2172 times
  07/05/2023 at 3:32 PM #217328
  AMQ

  Ok, thanks and sorry: Next post I Will use It as said.

  This code works in Backtest and real systems.

  What I would need is Code for trailing stop based on price action instead of optimizable variables, so I asked GPT to write Code for that. I am afraid that the prompt used is not working well, so I need more ideas to write the Code based on Price Action.

   

  Thanks a lot.

  

  AMQ

  Participant

  Topics: 8

  Replies: 18

  Been thanked: 2 times
• Author
  Posts