Entry signal reliability

• 07/21/2016 at 10:15 AM #10750
  absent1980

  Dear All,

  Today i would like to discuss entry signals reliability in general and how we could test them. First of all its worth to mention that entry isn’t that crucial after all and risk management with good exits are far more important. Nevertheless, having a reliable entry would give you some extra edge in trend following systems.

  Q: How reliable it should be?

  A: Literature suggests that consistent 55% and above win ratio is likely to be due to a good entry signal rather that just random occurrence. This should not be confused with the average trend following systems performance which is lower, I will come back to it later.

  Q: What about random entry performance?

  A: Some of you may be surprised to know that random entry can provide anything between 45% and 50% success ratio.

  Q. How do I test my entry signal?

  1. Delete everything from your code apart of entry signal.
  2. Do NOT set any stop loss or target profit.
  3. Set exit to a number of days/bars after entry. For example, you can use optimization and try 5, 10, 15, 20 and 30 days/bars. I cannot suggest the exact number, it depends on a market and time frames.
  4. Backtest your system, ideally on several markets.

  Q: How about results?

  A: If you consistently get more than 55% win ratio then your signal is likely to be more reliable than just a random entry. It will drop after you introduce stops and money management, which is expected. Trend following systems are not based on high success ratio and 35% to 40% is a good result.

  N.B. Bare in mind that testing it on time frames lower than 4H will involve a lot of noise and unlikely to of any use.

  Please share your opinions and results.

   

   

   

   

  

  absent1980

  Participant

  Topics: 11

  Replies: 36

  Been thanked: 3 times
  07/28/2016 at 6:18 PM #10960
  absent1980

  I will continue this topic by adding some objective parameters to test your system which also allow you to compare different systems.

  • System Expectancy = (Win % * Average Win) – (Loss % * Average Loss)
  • SQN (system quality number) = root(n) * expectancy / StDev(R)
  • root(n) – the square root of the number of all trades (you need at least 50 trades for statistically sound results)
    expectancy – as shown above and measured in R multiples
    stdev(R) – the standard deviation of your profit/loss R multiples

  Usually a SQN score of between 1.6 – 1.9 is considered poor but tradable. 2.0 – 2.5 is average. 2.5 – 2.9 is good and anything above 3.0 is deemed excellent.

  SQN and system expectancy are mostly research products of Dr Van Tharp. I would strongly recommend to read some of his books.

  1 user thanked author for this post.

  Bard

  

  absent1980

  Participant

  Topics: 11

  Replies: 36

  Been thanked: 3 times
  07/28/2016 at 6:57 PM #10966
  Nicolas

  Thanks a lot absent1980,

  These calculations can be easily made directly into a strategy which is currently running. The Average win/loss and % can be computed within a loop through the past orders.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  07/28/2016 at 8:55 PM #10975
  absent1980

  Thats a good idea. Perhaps it could be built into backtesting? Saying that it takes only few minutes to caclculate.  I wonder if SQN can be used to monitor ongoing system performance? For example, if you start with 2.5 and the it drops after some time to less than 2, it may mean that market condition changed or system needs to be adjusted.

  

  absent1980

  Participant

  Topics: 11

  Replies: 36

  Been thanked: 3 times
• Author
  Posts