Sharpe ratio >3 ? Seriously ?

• 06/05/2020 at 9:41 AM #134659
  zilliq

  Hi Guys,

  I play with the new backtest results, but it’s strange, because on Backtest we can have Sharpe Ratio of more than 2 (Here 4.64 ! on 100 000 unit EUR/USD 1 mn, only on 40 trades and so on..)

  Remember that the best trading algo have a Sharpe ratio of 3, so I think a Sharpe ratio of more than 4 is impossible (I’m not stupid and I know we haven’t fin the graal, or the @Grahal 🙂 )

  What’s your Point of view ?

  Cheers

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  06/05/2020 at 9:48 AM #134662
  Nicolas

  Low quantity of orders means not much data to average a significant ratio. With a high profitability strategy divided by a low standard deviation of gain, you obviously get a high Sharpe Ratio.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  06/05/2020 at 10:29 AM #134674
  zilliq

  Sharpe ratio is on mean gain Nicolas, not total gain

  “…The Sharpe ratio reveals the average investment return, minus the risk-free rate of return, divided by the standard deviation of returns for the investment…”

  But agree with you on a small trades who makes interpretation difficult (That’s why there is a correction factor on SQN/Van Tharp notably ;,-) )

  Cheers

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  06/05/2020 at 1:53 PM #134720
  zilliq

  As you can see, with quite the same number of trade, we can have very good sharpe ratio even with a small gain (+34 % only)

  Have a nice day

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  06/05/2020 at 4:22 PM #134742
  Nicolas

  Orders are grouped as a daily return:

  dailyReturn[day] = sum(perf)[day] / nb_trades[day]

  Then the Sharpe ratio is calculated:

  sharpe = (average(dailyReturns) – risk free interest rate between start and end of the report) / standard-deviation(dailyReturn)

  The risk free interest rate is automatically calculated by the software with a geometric interpolation.

  1 user thanked author for this post.

  zilliq

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  06/05/2020 at 4:57 PM #134747
  zilliq

  We are together agree, that’s derived/based on mean gain, that’s why even with a small total gain we can have a beautiful Sharpe ratio (Interest is may be for the “ego” 🙂

  Have a nice day Nicolas

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  06/11/2020 at 7:54 AM #135534
  zilliq

  FYI (very bad) results after some days even with a sharpe ratio of 4.64 ! (And remember 90 % winning trades, 6 win/loss and so on…EUR/USD 1 mn 100 000 units)

  Sharpe ratio don’t do all the job 😉

   

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  06/11/2020 at 10:16 AM #135555
  Nicolas

  Nothing new if the sharpe ratio is calculated on bias and over-fitted results 🙂

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  06/11/2020 at 11:28 AM #135571
  zilliq

  It wasn’t overfit Nicolas (Only 2-3 parameters as usual), 100 000 units and so on

  It’s jut that Sharpe Ratio isn’t a standalone and need to be interpret with other indicators as we previous seen

  Have a nice day

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
• Author
  Posts