Simulated Trading
-
As Nicolas has posted a great new blog on using simulated trading to improve strategy performance at:
https://www.prorealcode.com/blog/learning/how-to-improve-a-strategy-with-simulated-trades-1/
I thought that I would post something a little similar (but not as good!) that I wrote a while back. I have never used it in anger in a live or demo strategy as I felt the value of the average period was too critical as to whether it worked well or not. Maybe a better filter can be used?
x = 20 //average period positionsize = 5 //tradesize if equitycurve >= average[x](equitycurve) or tradecount <= x then trading = 1 endif if equitycurve < average[x](equitycurve) and tradecount > x then trading = 0 endif if trading and not onmarket then buy positionsize contracts at market buyprice = close tradecount = tradecount + 1 endif if trading = 0 and not onmarket then buyprice = close tradeonmarket = 1 endif if onmarket and close > positionprice or onmarket and close > high[1] then sell at market equitycurve = equitycurve + (close - buyprice) endif if tradeonmarket and close > buyprice or tradeonmarket and close > high[1] then tradeonmarket = 0 equitycurve = equitycurve + (close - buyprice) endif graph equitycurve graph average[x](equitycurve) graph tradingI tested it on the SP500 weekly. Obviously there is no real strategy involved as it just buys every week if it is not on the market and sells at the close of a week if it is in profit or if the week has closed higher than the previous weeks high.
As with all uses of an average it will be curve fitted dependent on the average period selected. Also just like Nicolas’s example the filter does not come into play until x number of trades have passed. It would be possible to alter the code so that these are either real trades or simulated trades depending on your appetite for risk I guess.
Here are the results with a period of 20:
[attachment file=80483]
Maybe a discussion on similar codes and the use of simulated trades will be of interest to someone?
Nicolas thanked this postHere’s the same strategy entry and exit criteria but this one keeps a simulated record of the last 10 trades. If the last 10 trades add up to an overall loss then real trading stops and simulated trades are recorded. Real trading does not resume until the last last ten trades have returned to an overall profit.
Once again the filter is not applied to the first 10 trades.
x = 10 positionsize = 5 lasttradestotal = trade1 + trade2 + trade3 + trade4 + trade5 + trade6 + trade7 + trade8 + trade9 + trade10 if lasttradestotal >= 0 or tradecount <= x then trading = 1 endif if lasttradestotal < 0 and tradecount > x then trading = 0 endif if trading and not onmarket then buy positionsize contracts at market buyprice = close tradecount = tradecount + 1 endif if trading = 0 and not onmarket then buyprice = close tradeonmarket = 1 endif if onmarket and close > positionprice or onmarket and close > high[1] then sell at market trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = close - buyprice endif if tradeonmarket and close > buyprice or tradeonmarket and close > high[1] then tradeonmarket = 0 trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = close - buyprice endif //graph trading graph lasttradestotal graph 0[attachment file=80499]
I thought it would be nice to see if the ten day floating equity curve as a filter would improve a random strategy. So I wrote a strategy that bought the SP500 after a three day drop and sold if the price went over an average – nothing else – no stops or take profits and no more than one position held at a time. I used a spread of 1.6.
Here are the results:
[attachment file=80578]
Then I added the ten trade floating equity curve filter to see if it improved it at all. Here are the results:
[attachment file=80579]
The conclusion was that using the filter resulted in more profit, slightly better gain per trade, slightly better win ratio, less time in the market but most importantly a big improvement in draw down.
This sort of equity curve filtering may have some merit it seems but further testing is needed on the effect of such things such as number of trades used to calculate the floating equity curve before a full conclusion can be reached.
GraHal thanked this postAnother idea I am testing is comparing a five trade floating equity curve with a ten trade floating equity curve and only trading live if the ten trade curve is above the five trade curve.
lasttradestotal = trade1 + trade2 + trade3 + trade4 + trade5 + trade6 + trade7 + trade8 + trade9 + trade10 lasttradestotal5 = trade1 + trade2 + trade3 + trade4 + trade5 if lasttradestotal >= lasttradestotal5 then trading = 1 endif if lasttradestotal < lasttradestotal5 then trading = 0 endif if trading and not onmarket then buy 5 contracts at market buyprice = close endif if trading = 0 and not onmarket then buyprice = close tradeonmarket = 1 endif if onmarket and close > positionprice or onmarket and close > high[1] then sell at market trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = (close - buyprice) * positionsize endif if tradeonmarket and close > buyprice or tradeonmarket and close > high[1] then tradeonmarket = 0 trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = (close - buyprice) * positionsize endifMags67 thanked this postYou have almost rewrote “Dual momentum” theory from scratch
Almost… but with errors! I just noticed that I am comparing the first 5 trades in the 10 recorded trades instead of the last 5 trades. Here is the corrected code:
lasttradestotal = trade1 + trade2 + trade3 + trade4 + trade5 + trade6 + trade7 + trade8 + trade9 + trade10 lasttradestotal5 = trade6 + trade7 + trade8 + trade9 + trade10 if lasttradestotal >= lasttradestotal5 then trading = 1 endif if lasttradestotal < lasttradestotal5 then trading = 0 endif if trading and not onmarket then buy 5 contracts at market buyprice = close endif if trading = 0 and not onmarket then buyprice = close tradeonmarket = 1 endif if onmarket and close > positionprice or onmarket and close > high[1] then sell at market trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = (close - buyprice) * positionsize endif if tradeonmarket and close > buyprice or tradeonmarket and close > high[1] then tradeonmarket = 0 trade1 = trade2 trade2 = trade3 trade3 = trade4 trade4 = trade5 trade5 = trade6 trade6 = trade7 trade7 = trade8 trade8 = trade9 trade9 = trade10 trade10 = (close - buyprice) * positionsize endifMags67 thanked this post
- You must be logged in to reply to this topic.
Summary
This topic contains 52 replies,
has 7 voices, and was last updated by 
5 years, 1 month ago.
Topic Details
| Forum: | ProOrder: Automated Strategies & Backtesting |
| Language: | English |
| Started: | 09/14/2018 |
| Status: | Active |
| Attachments: | 11 files |
