Z-Score – closer to zero is good?

• 06/13/2021 at 9:39 AM #171675
  Monobrow

  Hi

  Ive been looking at Z-Score in the “statistics of closed trades” dialog and am still not entirely sure whether high/low or close to zero is what I want?

   

  I think… closer to zero means there is a lower liklihood of a streak / variance, which is good if you want a solid algo… but im really not sure.

   

  Can someone shed any light please?

   

  thanks

  

  Monobrow

  Participant

  Topics: 15

  Replies: 157

  Been thanked: 23 times
  06/13/2021 at 5:36 PM #171688
  Nicolas

  Sequence of losses/gains are random or not, the sign simply tells us whether the dependence is positive or negative, i.e. :

  Negative Z-Score – the dependency is positive.
  Positive Z-Score – the dependency is negative.

  Positive dependence – a benefit will be followed by a benefit and a loss by a loss.
  Negative dependence – a loss will be followed by a gain and a gain by a loss.

  And how sure can we be that this will happen? That’s the percentage that you are told in the z-score value of your system. So, for example, if the z-score = 2.17 (97.0%), it means that there is a 97% confidence level that a loss will be followed by a profit and a profit by a loss (negative dependence). In this case, you would want to miss a trade after a profit and take a trade after a loss to maximize profits and minimize losses.

  1 user thanked author for this post.

  Monobrow

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
  06/14/2021 at 6:41 AM #171696
  Monobrow

  Thank you Nicolas, a few more questions if I may…

   

  1. How do you calculate the value 97% from 2.17 z-score? What percentage is a 0 (zero) z-score for example? Is it 100% or 0%?

  2. in terms of a “perfect” system, where would you want the z-score? I understand what you explained above, but if possible id like a target to aim for to add to other stats, like low drawdown for example.

  3. over what period (i.e how long) would you expect be required to get an indicative z-score? Obviously the more the better, but in general – 3 months? 6 months? Etc. If not clear, just your best guess, or what period would you look for?

   

  thanks!

   

  

  Monobrow

  Participant

  Topics: 15

  Replies: 157

  Been thanked: 23 times
  06/14/2021 at 8:29 AM #171701
  Nicolas

  1. Z-score measure the distance from the mean (average), think of it as a standard deviation. 100% is the highest Z-score.
  2. It obviously depends of the strategy and the way you manage your PnL with money management or any other magic
  3. Statistical calculation is always better with more population, it depends of trades quantity first, so the more the better 🙂

  1 user thanked author for this post.

  Monobrow

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
  06/14/2021 at 1:38 PM #171745
  Monobrow

  I think I am almost there! Thank you..

  so..for question 1.  Does a ZScore of zero mean there is a 100% chance the next trade will go the other way? If so.. surely that just isn’t possible!

   

  thanks!

   

  to add.. if a z score of zero is not 100%, what would be? A score of 100?

  

  Monobrow

  Participant

  Topics: 15

  Replies: 157

  Been thanked: 23 times
  06/15/2021 at 8:10 PM #171820
  Monobrow

  Does anyone else know the answer to this?

   

  thanks!

  

  Monobrow

  Participant

  Topics: 15

  Replies: 157

  Been thanked: 23 times
  06/27/2021 at 6:24 AM #172548
  JS

  Z-score is a tricky interpretation.

  A winning system can have both a positief or negative Z-score.

  When the Z-score is zero, all of your trades are winners because the total profit divided by your (winning) trades is the average profit (mean), so there is no deviation.

  Example negative Z-score: if you have 6 trades and 1 is a big winner than your Z-score will be negative because the other 5 will be below your average profit (mean).

  

  JS

  Participant

  Topics: 22

  Replies: 1423

  Been thanked: 557 times
• Author
  Posts