Evaluation of the systems performance

• 07/23/2019 at 8:56 AM #103198
  winnie37

  I often wondered how I could quickly compare the performance of models that I launched in demo or live, whatever the time frame.

  The way I found is: At the same lot size comparison (1), I multiply the W/L ratio by the frequency of the daily trades. This makes it possible to compare an index of absolute performance, taking into account the frequency of the occurrences.

  Of course, other elements then come into play, such as DD.

  Example: Dayopen straddle
  R W/L = 4.19
  Average executed orders: 1,29
  Index: 4.12 (4.19 * 1.29)

  Vectorial dax
  R W/L = 3.66
  Average executed orders: 0,73
  Index: 1.94 (3.66 * 0.73)

  Dayopenstaddle has a better absolute performance than Vectorialdax in the same time (this one could have been reevaluated by a higher trade frequency, which is not here).
  And you, have you developed some indicators for a quick comparison of the performance in the same time?

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 9:21 AM #103202
  Vonasi

  These might be of interest:

  https://www.prorealcode.com/topic/average-gain-per-1-code-snippet/

  https://www.prorealcode.com/topic/annual-percentage-returns/

   

  On a separate note I’m not sure of the maths in your examples:

  Index: 4.12 (4.19 * 1.29)             surely 4.19*1.29 = 5.41?

  and

  Index: 1.94 (3.66 * 0.73)            surely 3.66*0.73 = 2.63?

  1 user thanked author for this post.

  winnie37

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 9:57 AM #103207
  winnie37

  Sorry Vonasi, a bit tired 😉 You’re right 🙂

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 10:34 AM #103215
  winnie37

  Thanks a lot. I’m gonna add the first code to my systems, and monitoring. However, how to take the frequency of the trade into account?  Maybe by mixing the average gain per 1€ with the trades/day?

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 11:52 AM #103220
  Vonasi

  It shouldn’t be too difficult to count the number of days by adding 1 to a variable every time date <> date[1] and to also count trades by adding 1 to a variable every time that strategyprofit <> strategyprofit[1].

  Then use these in your index calculation.

  1 user thanked author for this post.

  winnie37

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 1:23 PM #103239
  winnie37

  Sorry Vonasi but could you developp more? With an exemple?

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 2:02 PM #103252
  Vonasi

  Perhaps something like this:

  if countofposition > countofposition[1] then trades = trades + (countofposition - countofposition[1]) endif if strategyprofit <> strategyprofit[1] then averagegain = strategyprofit/trades endif if opendate <> opendate[1] then daycount = daycount + 1 endif tradesperday = trades / daycount index = averagegain * tradesperday graph averagegain as "average gain" graph index as "index"

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

  if countofposition > countofposition[1] then trades = trades + (countofposition - countofposition[1]) endif   if strategyprofit <> strategyprofit[1] then averagegain = strategyprofit/trades endif   if opendate <> opendate[1] then daycount = daycount + 1 endif   tradesperday = trades / daycount index = averagegain * tradesperday   graph averagegain as "average gain" graph index as "index"

   

  1 user thanked author for this post.

  winnie37

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 4:00 PM #103274
  GraHal

  Interesting and useful, I’ll read the Topic again later.

  Wouldn’t it be sooo much better if there is a tool in PRT than enabled some degree of System Performance Comparison.

  Even being able to  sort the performance stats in columns would be a start … I believe / hope V11 has this?

  1 user thanked author for this post.

  winnie37

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  07/23/2019 at 4:12 PM #103275
  winnie37

  Thanks a lot Vonasi. It will be very precious for me to allocate the ressources. To be sure, what mean for exemple a index value of 10 compare to 5? 2 x more effective  for the same comparaison?

   

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 4:15 PM #103276
  Vonasi

  Yes GraHal it would be great. Imagine if we could do a personal quality index calculation in our strategy and then have the results shown in the optimisation results table – then we could sort the results based on that personal index calculation. We could compare drawdown to run up or to total gain or to win loss ratio. The options would be endless and the results far more useful.

  1 user thanked author for this post.

  GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 5:10 PM #103280
  GraHal

  Would be worth sending as a suggestion direct to PRT Vonasi … they’ll listen to you more than me! 🙂

  https://www.prorealtime.com/en/contact?suggestion=1

  Love your avatar Winnie … is that little piggie yours!? 🙂

  btw Vonasi … my wife’s IG Account was deleted about 2 weeks ago … Reason Given –  Inactivity for 2 Years!!

  It was not inactivity per se as we had 25 Systems on Demo Forward Test … now all Performance History is lost and the Systems themselves of course!!!!!

  My wife emailed IG and begged them to restore her account, saying she would fund Live immediately and trade via the IG New PLatform. She even said  PLEASE PRETTY PLEASE but as yet we have heard nothing!!!!!

  Lesson to be learnt … copy all Systems to PC /wherever outside of PRT Platform.

  Annoying thing is above action would still not preserve performance history.

  It needs like you suggested / voiced a while ago … all Systems from both SB and CDF Accounts to be stored / saved against an Account at PRT or some such method?

  Deleting an Account / Systems with no warning shows how little IG value our efforts to get good performing Systems so we can make money and make them money in the process??

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  07/23/2019 at 5:22 PM #103281
  Vonasi

  Sorry to hear about the loss of your account. I too had exactly the same thing a few years ago. In fact I would not be surprised if I was due an email from IG any day soon as I have not funded a couple of my live accounts for quite some time now.

  A good reminder for us all to back up regularly!

  I don’t understand why IG have these account clear out policies unless an idle account is costing them money in some way. Surely it is easier for a user to get back into spread betting if they have an open account than if they have to sign up for a new one?

  Strangely just the other day I was thinking that maybe IG were keeping idle accounts open when I read a warning saying that 75% of their account holders lost money. Surely if 95% of their accounts were idle then this figure would be far lower!

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 5:31 PM #103283
  Vonasi

  GraHal – I’ve sent a suggestion to PRT via the link that you provided as I’ve never done it that way before so it will be interesting to see if anything happens. I think I’ll use it every time I have a suggestion from now on just to double my chances of something that I actually want appearing in the platform!

  1 user thanked author for this post.

  GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  07/23/2019 at 5:59 PM #103284
  winnie37

  No, Grahal, it’s not my little pig, just my avatar 🙂
  Regarding the loss of your systems, I am sorry. I used to archive backtests and every .itf strategy in Evernote. it’s a reflex that I took and that may serve one day in the light of experience … Concerning the code foruni by Vonasi, I still hesitate in the understanding. Is the index the batch comparison element, and equivalent time value? What to remember to enjoy my systems and their allocations, index or average gain? I’m a little lost … sorry
  Nevertheless happy that it allows to emit suggestions relevant and profitable to all if Prt listens to us 🙂

  

  winnie37

  Participant

  Topics: 27

  Replies: 165

  Been thanked: 16 times
  07/23/2019 at 7:15 PM #103286
  Vonasi

  winnie37 wrote:

  what mean for exemple a index value of 10 compare to 5?

  Having an index is one thing – knowing what it means is another thing altogether!

  For example you could have a strategy that has one trade a day with a return of £1000 for every £1 staked and then another strategy that has 1000 trades a day with a return of £1 for every £1 staked. The index score would be the same so which is better? The answer I guess is that both are strategies that each of us would love to have but the one we personally use is down to our own individual trading style.

  Only by comparing lots of strategies can we understand what an index score really means and what minimum score level we are happy to accept.

  1 user thanked author for this post.

  winnie37

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
• Author
  Posts