How to evaluate Monte Carlo Simulation

• 11/29/2019 at 10:00 AM #113903
  PastaPesto

  Hello,

  I have tested the Monte Carlo sometimes before but now I really want to have it included in my strategy evaluation procedure. I just run three of my strategies in the excelsheet from kjtrading and they all look good to me. The question is what to look for and when it would be a warning flag. I have attached a sample and I have some questions about this if anybody uses the MC simulation, please give your input.

  1. pic1 shows the result from my backtest in PRT. The drawdown is 278 USD and the result for 8 years is 2170USD. That gives me about 100% return/year compared to the worst drawdown. The MC simulation is showing return/DD of about 2, which should be 100% as well. I can also see that the median profit is close to 100% of my drawdown. I guess this is all good?
  2. The drawdown from PRT was as mentioned 278USD which is a drawdown in percent of 27,8% from my 1000 equity start. The medium drawdown in MC is showing 11,5% on the first row where the start equity is also 1000USD. It is the same for all three of my tested strategies, they had all a much lower medium drawdown in MC than I had in PRT. What does this mean, I think its strange that the difference is so big? I guess that if the MC shows a much higher DD than my backtest, that should indicate that I was “lucky” with the distribution of my trades in backtest and I could expect a higher Drawdown if going live, but here it seems to be the opposite?!
  3. The risk of ruin seems to be calculated from the “stop trading if equity…” in cell B3. I think that is a bit strange, shouldnt it be calculated from my equity start? The ruin should accur when my equity is gone, not when I decide to stop trading my system?

  I just started writing things down, so I apologies for not being so clear with my questions. I would like to use the MC as a another step in the strategiy evaluation, I just need to decide what rules to set up. From what I am thinking now the Medium drawdown is important to have pretty close to my DD in PRT and the return/DD shouldnt be to low (guess that goes hand in hand). Risk of ruin is ofcourse important but I dont think I dont need MC for that. So my question is really if any of you are using the MC and how you evaluate the numbers?

  Btw, I am aware that if I use the Monte Carlo with curvfitted numbers it wont help me (garbage in, garbage out).

  

  PastaPesto

  Participant

  Topics: 13

  Replies: 29

  Been thanked: 6 times
  11/30/2019 at 11:21 PM #113978
  deletedaccount100622

  Hi Pasta Pesto

  For me the distribution of returns and time frame may be skewing the model, looking at the back testing results the model does not look to be as effective in the last three years. The simulation on the face of it appears to show that the risk of a 30% draw down is next to none however if the 0.44 trades per day was valid in the last three years then the average return per trade would be about £1 on one contract which looks like an unfavourable risk to reward ratio.

  Personally I would run the back tests and montecarlo simulation for each year at a time. I don’t know the specifics of you system but if you are holding positions overnight then you should also consider this if the system returns are marginal

  Thanks

  Ruark

  1 user thanked author for this post.

  PastaPesto

  

  deletedaccount100622

  Participant

  Topics: 13

  Replies: 116

  Been thanked: 60 times
  12/02/2019 at 3:51 PM #114077
  PastaPesto

  Hi Ruark,

  Thank you for your input. I will take each year and look at them seperately, that seems like a good idea. I dont know though how reliable that test will be as the number of trades is about 50/year. Maybe Ill do two years at a time and at least have about 100 trades to look at. I can see how the result is going down during the last three years and it seems to be the average win that is going down a bit as well as some fewer trades. The % of winners are constant. I think I will run all my strategies with the Monte Carlo just to get a feeling of how to use it.  I am sure it is a good tool if used correct.

  Are you using it in your evaluation? If yes, what would be a warning flag for you? How do you use it?

  Anyone else in the forum using the MC simulation?

  Thanks

  

  PastaPesto

  Participant

  Topics: 13

  Replies: 29

  Been thanked: 6 times
  12/02/2019 at 7:29 PM #114083
  GraHal

  Did you see / read this Topic @PastaPesto

  https://www.prorealcode.com/topic/monte-carlo-analysis/

   

  1 user thanked author for this post.

  PastaPesto

  

  GraHal

  Participant

  Topics: 204

  Replies: 8418

  Been thanked: 2043 times
  12/03/2019 at 12:53 PM #114111
  PastaPesto

  Hi GraHal,

  Yes I read it all 🙂 thanks.

   

  

  PastaPesto

  Participant

  Topics: 13

  Replies: 29

  Been thanked: 6 times
• Author
  Posts