Back testing strategies

• 06/29/2018 at 7:26 AM #74679
  Vonasi

  Another important point is that if you spend a whole day developing a strategy idea that turns out to be rubbish then do not be depressed about it. Sit back and ask yourself why it does not work (which you should now know thanks to your tireless analysis) and then take that reason why it does not work and store it in your brain somewhere. Your next strategy that you develop will not have any time wasted on that idea or method and the process becomes easier and faster day by day. Slowly you will hone in on a method of trading that you understand and that works for you. Every failure is a good lesson and not a failure at all.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  06/29/2018 at 7:58 AM #74686
  Vonasi

  Just to give you an idea of how after all that work and analyse your strategy still might not work. Here is my demo trades since mid May:

  

  Looks great! But here is the equity curve for the last 750ish trades:

  

  Clearly not every idea worked straight out of the box even after so much testing and analysis and I have been slowly removing the failed strategies and replacing with more proven ideas in the last month or so. One or two of these strategies might get to go live after a few more months of forward testing.

  You have to look at that last chart and say to yourself ‘this is why I forward test’.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  06/29/2018 at 8:43 AM #74700
  ChrisNYE

  This may sound odd. But my best strategies are my basic ones with basic and simple strategy code.

  I also find that when I have a good strategy I want to optimize it so much that the profits go up but the trades go to a small number.

  Someone told me years ago not to data mine and the most important this is your exit AND NOT MY ENTRY. I am finding no one stray customization works for all fx pairs.

  

  ChrisNYE

  Participant

  Topics: 27

  Replies: 125

  Been thanked: 4 times
  06/29/2018 at 8:56 AM #74703
  Vonasi

  Entry is the easy bit. Look at any chart and point at any candle and you can see a reason to open a trade. Knowing when it has gone wrong far enough to get out or when it is at a profit peak to get out is the hard bit. You have to accept that you will always leave money on the table and you will always have losers bigger than you want.

  If you have optimized a strategy to the point where it is a winner but with very few trades then put it in the bin and start again.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  06/29/2018 at 9:19 AM #74714
  GraHal

  ChrisNYE wrote:

  I want to optimize it so much that the profits go up but the trades go to a small number.

  Don’t choose / go forward with the top profit opto-result. Keep a sense of balance about which opto-result you select with regard to win % and number of trades etc.

  Click on several rows in the opto-results table and let the equity curve load for each result row and make a judgement on the equity curve. Also check the performance stats for these equity curves you load up so you can make a judgement on drawdown and average gain per trade etc.

  For example … is $460 average gain per trade on a 5 min TF likely to be repeatable or might we have curve fitted just a tad too much!?? 🙂

  

  GraHal

  Participant

  Topics: 204

  Replies: 8428

  Been thanked: 2046 times
  06/29/2018 at 9:19 AM #74715
  ChrisNYE

  As I predominately trade 1 hr charts I want around 30 trades per annum. Then I thought it depends on your strategy and if you ride the trend for months. My personality likes the profits to come in often and I am not interested in still being in a trade one month later. I prefer to take profits and look for another opportunity or is this I correct?

  

  ChrisNYE

  Participant

  Topics: 27

  Replies: 125

  Been thanked: 4 times
  06/29/2018 at 9:24 AM #74717
  GraHal

  ChrisNYE wrote:

  I prefer to take profits and look for another opportunity or is this I correct?

  Yes deffo if it suits you and it sure suits me!

  I get bored with Systems that don’t trade a few times per week at least, preferably minimum once per day. I have had Systems in the past rejected as they traded more than the IG limit of 50 times per day! 🙂

  PS / EDIT
  As an example, I started a System on Monday  25 June and it lost £493 over 3 trades by yesterday 28 June so it is history and out it goes!!!! I never throw the baby out with the bathwater however … the System will get exported to my PC into a folder for no hopers! 🙂 🙂

  Above System may have gone on to be my biggest winner  of all time if I gave it long enough (as it was on 4H Tf) , but I know I wouldn’t have the patience to live with such a big drawdown so why wait around … plenty more fish in the sea, just need to catch them / code them!!!!  

   

  

  GraHal

  Participant

  Topics: 204

  Replies: 8428

  Been thanked: 2046 times
  06/29/2018 at 9:31 AM #74719
  ChrisNYE

  Grahal what time frame are you trading for 50 trades a day?

  I will stick to 1 hour charts. I am guilty on clicking on the top % return where as I know to be more consistent and realistic I should be looking down a few more. My aim is % wins and avge gain per trade and low drawdown

  

  ChrisNYE

  Participant

  Topics: 27

  Replies: 125

  Been thanked: 4 times
  06/29/2018 at 9:40 AM #74722
  GraHal

  ChrisNYE wrote:

  time frame are you trading for 50 trades a day?

  That System was 10 seconds TF.  I had been manual trading on 1, 5 and 10 second TF and so I coded up my strategy into an Auto-System.  It got binned (sorry, exported to my PC)  after a while as I had curve fitted it to much and results dropped off etc.

  Smaller TF the better for me consistent with increasing profit, but 5 mins TF is a good compromise that can give repeat / consistent-ish results.

   

   

  

  GraHal

  Participant

  Topics: 204

  Replies: 8428

  Been thanked: 2046 times
  06/29/2018 at 9:47 AM #74724
  ChrisNYE

  With 5 min time frames you must only trade during certain hours and not 24/7?

  I am finding as I can only back test 4 yes that a few systems are long only. Is this ok? For example Nasdaq is purely long only on my back testing optimization.

  

  ChrisNYE

  Participant

  Topics: 27

  Replies: 125

  Been thanked: 4 times
  06/29/2018 at 10:05 AM #74727
  GraHal

  ChrisNYE wrote:

  With 5 min time frames you must only trade during certain hours and not 24/7?

  I can’t see where you are coming from with this? Most of my 5 M TF Systems are 24 hours, but I may restrict them when I go Live with them after I analyse the times where the most losses are occurring.

  I have currently disabled my Live PRT Account so I can open a beer and get on with life (Vonasi style) during the long summer days.

  ChrisNYE wrote:

  I am finding as I can only back test 4 yes that a few systems are long only. Is this ok? For example Nasdaq is purely long only on my back testing optimization.

  This could be because your short entry and exit strategy are not good and the optimiser is telling you this by not giving you any shorts for results near the top of the results list? If you could see all results of the optimiser then you would see shorts losing money with each trade?

  

  GraHal

  Participant

  Topics: 204

  Replies: 8428

  Been thanked: 2046 times
  06/29/2018 at 10:21 AM #74733
  Vonasi

  GraHal wrote:

  but I may restrict them when I go Live with them after I analyse the times where the most losses are occurring.

  Sounds to me like you do all the testing and then curve fit the strategy with an extra condition just before you put it live!

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  06/29/2018 at 10:25 AM #74734
  ChrisNYE

  My problem was that I was looking at the top % return and it had no shorts in it. Need to look down the list and look at all the performance stats as you advised me to do.

  

  ChrisNYE

  Participant

  Topics: 27

  Replies: 125

  Been thanked: 4 times
  06/29/2018 at 10:28 AM #74735
  Vonasi

  GraHal wrote:

  For example Nasdaq is purely long only on my back testing optimization.

  Not sure why anyone would want to go short on a market like this – you are fighting the overall trend that most indices have. Just stay out of the market during recessions (or short only then if you want to short at all) and you will already have a big advantage.

  

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  06/29/2018 at 10:38 AM #74742
  GraHal

  Vonasi wrote:

  Sounds to me like you do all the testing and then curve fit the strategy with an extra condition just before you put it live!

  Mmm maybe, by cutting out trades between say 20:00 and 07:00 because the majority are losers. But, anyway I kinda just said it that to illustrate  to  Chris re his thinking that 5 min TF had to be restricted / not 24/7.

  Also you know what it’s like … by the time I’ve spent hours coding and optimising then itterating, I just want to set my baby free in the real world and see her grow and grow (I wish!!!! 🙂 )  So maybe some analysis gets left for another day!

  But I agree lossy clock times should be seen in the opto-results before doing Demo Forward Testing.

  

  GraHal

  Participant

  Topics: 204

  Replies: 8428

  Been thanked: 2046 times
• Author
  Posts