Seasonality or just the ultimate curve fit?

• 03/07/2021 at 5:03 PM #163370
  Vonasi

  I got to thinking about seasonality again and this got me thinking that I should test to see if there are any sections of a year that if you had been long or short in every year would have returned good results and a nice equity curve. So I coded two very simple strategies that count week numbers and and buy on one week number and sell on another week number. Optimise the two strategies with the variables startweek and endweek with values between 1 and 53 and it will tell you which blocks of weeks have given the best return during the historical data.

  Then I created a third strategy that you can enter the best returned values as startweeklong, endweeklong, startweekshort and endweekshort so as to see what the equity curve would have looked like had you been short during the best weeks and long during the best weeks. On my tests I found that the best long weeks rarely clashed with the best short weeks but if you find an instrument where they do then you can’t put those values in the ‘Both’ strategy.

  The strategies only work on the weekly chart.

  The question up for general discussion is….. is this seasonality or is it just the ultimate curve fit?

  I have attached images of Live Cattle and Soybeans. No spread was used in the tests.

  Long

  defparam cumulateorders = false if openyear <> openyear[1] then wn = 0 endif wn = wn + 1 if wn = startweek then buy 1 contract at market endif if onmarket and wn = endweek then sell at market endif

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

  defparam cumulateorders = false   if openyear <> openyear[1] then wn = 0 endif   wn = wn + 1   if wn = startweek then buy 1 contract at market endif   if onmarket and wn = endweek then sell at market endif

  Short

  defparam cumulateorders = false if openyear <> openyear[1] then wn = 0 endif wn = wn + 1 if wn = startweek then sellshort 1 contract at market endif if onmarket and wn = endweek then exitshort at market endif

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

  defparam cumulateorders = false   if openyear <> openyear[1] then wn = 0 endif   wn = wn + 1   if wn = startweek then sellshort 1 contract at market endif   if onmarket and wn = endweek then exitshort at market endif

  Both

  defparam cumulateorders = false startweekshort = 17 endweekshort = 39 startweeklong = 39 endweeklong = 17 if openyear <> openyear[1] then wn = 0 endif wn = wn + 1 if wn = startweeklong then buy 1 contract at market endif if onmarket and wn = endweeklong then sell at market endif if wn = startweekshort then sellshort 1 contract at market endif if onmarket and wn = endweekshort then exitshort 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

  defparam cumulateorders = false   startweekshort = 17 endweekshort = 39 startweeklong = 39 endweeklong = 17   if openyear <> openyear[1] then wn = 0 endif   wn = wn + 1   if wn = startweeklong then buy 1 contract at market endif   if onmarket and wn = endweeklong then sell at market endif   if wn = startweekshort then sellshort 1 contract at market endif   if onmarket and wn = endweekshort then exitshort at market endif

   

  2 users thanked author for this post.

  Paul, GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/07/2021 at 5:15 PM #163376
  Vonasi

  Here is another example. This time it is Heating Oil.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/08/2021 at 12:35 AM #163392
  Paul

  interesting. I put it also on a few.

  maybe a month is interesting too and you can expand parameters to enter again long after short i.e. more easily. Not sure yet if usefull.

  Here are a few pics with use long&short

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  03/08/2021 at 2:45 AM #163400
  Paul

  It’s is just more curvefitting and defeats the goal of the code. This is a bit what I had in mind.

  now i.e. 4 turnarounds a year are possible, or long only twice a year certain periods.

  monthy bars, 25 units

  defparam cumulateorders = false //batchtype=1 then set a3 & a4 not to optimise //batchtype=2 can comment out the buy/sellshort rows as variation in batch2? //optimise a1 t/m a4 range 1-12 once tradetype=1 //1 both, 2 long, 3 short once batchtype=2 //batchtyp 1 solo with only parameters a1,a2 //first batch startweeklong = a1 endweeklong = a2 startweekshort = a2 endweekshort = a1 // second batch startweeklong2 = a3 endweeklong2 = a4 startweekshort2 = a4 endweekshort2 = a3 if openyear <> openyear[1] then wn = 0 wn2 =0 endif wn = wn + 1 wn2= wn2+ 1 // BATCH1 if batchtype=1 or batchtype=2 then if tradetype=1 or tradetype=2 then if wn = startweeklong then buy 1 contract at market endif if onmarket and wn = endweeklong then sell at market endif endif if tradetype=1 or tradetype=3 then if wn = startweekshort then sellshort 1 contract at market endif if onmarket and wn = endweekshort then exitshort at market endif endif endif // BATCH2 if batchtype=2 then if tradetype=1 or tradetype=2 then if wn2 = startweeklong2 then buy 1 contract at market endif if onmarket and wn2 = endweeklong2 then sell at market endif endif if tradetype=1 or tradetype=3 then if wn2 = startweekshort2 then sellshort 1 contract at market endif if onmarket and wn2 = endweekshort2 then exitshort at market 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 73 74 75 76 77 78 79 80 81 82 83

  defparam cumulateorders = false   //batchtype=1 then set a3 & a4 not to optimise //batchtype=2 can comment out the buy/sellshort rows as variation in batch2? //optimise a1 t/m a4 range 1-12   once tradetype=1 //1 both, 2 long, 3 short once batchtype=2 //batchtyp 1 solo with only parameters a1,a2   //first batch startweeklong  = a1 endweeklong    = a2   startweekshort = a2 endweekshort   = a1   // second batch startweeklong2  = a3 endweeklong2    = a4   startweekshort2 = a4 endweekshort2   = a3   if openyear <> openyear[1] then wn = 0 wn2 =0 endif   wn = wn + 1 wn2= wn2+ 1   // BATCH1 if batchtype=1 or batchtype=2 then   if tradetype=1 or tradetype=2 then   if wn = startweeklong  then buy 1 contract at market endif   if onmarket and wn = endweeklong then sell at market endif endif   if tradetype=1 or tradetype=3 then   if wn = startweekshort then sellshort 1 contract at market endif   if onmarket and wn = endweekshort then exitshort at market endif   endif endif   // BATCH2 if batchtype=2 then   if tradetype=1 or tradetype=2 then   if wn2 = startweeklong2 then buy 1 contract at market endif   if onmarket and wn2 = endweeklong2 then sell at market endif endif   if tradetype=1 or tradetype=3 then   if wn2 = startweekshort2 then sellshort 1 contract at market endif   if onmarket and wn2 = endweekshort2 then exitshort at market endif endif endif

   

   

   

  1 user thanked author for this post.

  Vonasi

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  03/08/2021 at 2:09 PM #163451
  Vonasi

  Thanks for that code Paul. I had considered something similar but decided that (as you suggest) it was just increased curve fitting. I also felt that if a commodity was to be seasonal then it was likely to only have two periods – one where being short the commodity was wise and the other where being long the commodity was wise due to seasonal supply and demand caused by weather and/or crop growing seasons.

  I think that on some truly seasonal commodities the equity curves highlight nicely what a good return one could have got just by being long and short in time with the seasons. What I don’t like the look of is the fact that a lot of the equity curves seem to have negative dips on the most recent data which makes me think that perhaps seasonality as a market force is declining and also that perhaps the nice previous equity curves are just quite simply a lovely curve fit.

  Still if one was to be trading a commodity that the test shows to produce a nice equity curve then having only long strategies at one part of the year and only short on another part of the year seems like a sensible filter that should improve win rates and performance/returns.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/08/2021 at 3:34 PM #163460
  Paul

  Yes, but even with a nice equitycurve as you describe in you lowest part, there are not many trades to have it rely on.  Long one part of the year, short the other had me make this variation because sometimes it just holds to a position while the market is trending against it, perhaps eventually becoming winning.

  I took another approach.  I changed it to a daily reset, mainly to have more trades.

  Now the focus is 2-3-4 hour charts and optimising starts now in range at 0 if i’am correct.

  As example I took the nasdaq. Added spread 2 & tickbytick.

  There are always dips in the market and here’s a way to minimise the effect it seems.

  I.e. buys if the high>high[1] next bar at market, else it exit’s shorts, which is still a buy but not holding a new position against the trend.

  Results are pretty good.

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  03/08/2021 at 3:52 PM #163465
  Vonasi

  I’m struggling a little to fully understand what you are now doing and see that you have switched to the Nasdaq which is a market that has not a lot to do with seasonality. Is it possible to post the code so that I can try to understand more easily what you mean?

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/08/2021 at 3:59 PM #163466
  Paul

  here’s eurusd with spread

  defparam cumulateorders = false //defparam flatbefore =080000 //defparam flatafter =200000 //batchtype=1 then set a3 & a4 not to optimise //batchtype=2 can comment out the buy/sellshort rows as variation? //optimise a1 t/m a4 range 1-12 once tradetype=1 //1 both, 2 long, 3 short once batchtype=2 //batchtyp 1 solo with only parameters a1,a2 //first batch startweeklong = a1 endweeklong = a2 startweekshort = a2 endweekshort = a1 // second batch startweeklong2 = a3 endweeklong2 = a4 startweekshort2 = a4 endweekshort2 = a3 //if openyear <> openyear[1] then if openday<>openday[1] then wn = 0 wn2 =0 endif wn = wn + 1 wn2= wn2+ 1 // BATCH 1 if batchtype=1 or batchtype=2 then if tradetype=1 or tradetype=2 then if wn = startweeklong then if high>high[1] then buy 1 contract at market else exitshort at market endif endif if onmarket and wn = endweeklong then sell at market endif endif if tradetype=1 or tradetype=3 then if wn = startweekshort then if low<low[1] then sellshort 1 contract at market else sell at market endif endif if onmarket and wn = endweekshort then exitshort at market endif endif endif // BATCH2 if batchtype=2 then if tradetype=1 or tradetype=2 then if wn2 = startweeklong2 then if high>high[1] then buy 1 contract at market else exitshort at market endif endif if onmarket and wn2 = endweeklong2 then sell at market endif endif if tradetype=1 or tradetype=3 then if wn2 = startweekshort2 then if low<low[1] then sellshort 1 contract at market else sell at market endif endif if onmarket and wn2 = endweekshort2 then exitshort at market endif endif endif //set stop %loss 10 //set target pprofit 250

  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 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

  defparam cumulateorders = false //defparam flatbefore =080000 //defparam flatafter  =200000   //batchtype=1 then set a3 & a4 not to optimise //batchtype=2 can comment out the buy/sellshort rows as variation? //optimise a1 t/m a4 range 1-12   once tradetype=1 //1 both, 2 long, 3 short once batchtype=2 //batchtyp 1 solo with only parameters a1,a2   //first batch startweeklong  = a1 endweeklong    = a2   startweekshort = a2 endweekshort   = a1   // second batch startweeklong2  = a3 endweeklong2    = a4   startweekshort2 = a4 endweekshort2   = a3   //if openyear <> openyear[1] then if openday<>openday[1] then wn = 0 wn2 =0 endif   wn = wn + 1 wn2= wn2+ 1   // BATCH 1 if batchtype=1 or batchtype=2 then   if tradetype=1 or tradetype=2 then   if wn = startweeklong then if high>high[1] then buy 1 contract at market else exitshort at market endif endif   if onmarket and wn = endweeklong then sell at market endif endif   if tradetype=1 or tradetype=3 then   if wn = startweekshort then if low<low[1] then sellshort 1 contract at market else sell at market endif endif   if onmarket and wn = endweekshort then exitshort at market endif   endif endif   // BATCH2 if batchtype=2 then   if tradetype=1 or tradetype=2 then   if wn2 = startweeklong2 then if high>high[1] then buy 1 contract at market else exitshort at market endif endif   if onmarket and wn2 = endweeklong2 then sell at market endif endif   if tradetype=1 or tradetype=3 then   if wn2 = startweekshort2 then if low<low[1] then sellshort 1 contract at market else sell at market endif endif   if onmarket and wn2 = endweekshort2 then exitshort at market endif endif endif   //set stop %loss 10 //set target pprofit 250

   

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  03/08/2021 at 4:20 PM #163474
  Vonasi

  OK but now seasonality has completely gone out of the window and because we are on faster time frame charts we truly have entered the world of finding the best possible curve fit to the data that we have.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/08/2021 at 4:52 PM #163485
  Paul

  haha, yeah seasonality is not longer the target. One thing let to another.

  This code analyses the market and decides based on a lot of data if it’s best to go long, short or both.

  Having 1-2k trades and a curve like that on different markets with spread is intriguing. It’s also on the dow, nikkei btw in lesser extend.

  Here’s eur/zar. Spread 1. Don’t know if that’s correct spread

  Al what we try is curvefitting. you can have 2 parameters for 50 trades in say 8 years, or 4 parameters for 2000 trades in 8 years. I’am leaning then to 4 parameters.

  I’ve to look into it a bit more. Thnx for posting the original idea. It gives a lot of new options!

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  03/08/2021 at 4:59 PM #163491
  Vonasi

  I think the only thing these curve fitted equity curves do is give us a best performance datum against which to compare other strategies to get some sort of idea of how good or bad they are. We can compare them to ‘buy only the best bits’ as well as ‘buy and hold now’!

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/08/2021 at 5:16 PM #163497
  Nicolas

  It reminds me the Pathfinder Trading System and Pathfinder swing TS

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  03/08/2021 at 5:30 PM #163500
  Vonasi

  Yes that used historical seasonality as a filter for future trades I seem to recall. I also believe that it didn’t work out too well as the future refused to do the same as the past even if the in sample equity curves looked great!

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/09/2021 at 12:13 PM #163589
  snucke

  ive been dabbeling alot with seasonality later too.

  i have been using the pathfinder as a basecode and ive been using this site https://charts.equityclock.com/live-cattle-futures-lc-seasonal-chart

  to look for seasonality patterns.

   

  but as you say, it might just be a little bit to curvefit..

  

  snucke

  Participant

  Topics: 9

  Replies: 143

  Been thanked: 21 times
  03/09/2021 at 3:44 PM #163627
  Vonasi

  You might also be interested in this indicator that I coded a while back then:

  https://www.prorealcode.com/prorealtime-indicators/seasonality-curve/

  1 user thanked author for this post.

  snucke

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
• Author
  Posts