Help with Monte Carlo Drawdown analysis

[MustPlayOptions]

I've been resisting using equity curve calculations. The reason for this is due to the various number of trades per day that can occur. It's starting to look like it might be necessary. I tried adding in the max excursions and the drawdown was something like 130k, which is double what the sim showed. I changed the spreadsheet btw to look at the actual cash drawdown since the trades are of a fixed size so the same max drawdown could theoretically happen at any time.

Interestingly, the method I've been playing with showed a capital requirement of 72k to put the ROR at 0.1%. The problem is that most of the values I'm getting for bankroll needed seem way too high relative to the straight sim values.

That's why I wanted a reality check with the Monte Carlo sim.

I'll try looking at the equity drawdowns tomorrow.

Thanks

 

[Volker Butzlaff]

@MustPlayOptions,

if you are interested, I can give you a mcs for your system results and can post it here.

I don't care about your system implementation or your single trades, but I need the system report information of a back test, which is described in the following document:

http://www.zentrader.de/checklist_blackboxsystems_e.pdf

bye,
zentrader

 

[Vince1]

Quote from MustPlayOptions:

I've never done a Monte Carlo analysis so I don't know what's needed.

http://www.schindlertrading.com/include/content/uploads/drawdown analysis.pdf
http://office.microsoft.com/en-us/excel/HA011118931033.aspx?pid=CL100570551033

These might also help++

 

[NoWorries]

Quote from Volker Butzlaff:

@MustPlayOptions,

if you are interested, I can give you a mcs for your system results and can post it here.

I don't care about your system implementation or your single trades, but I need the system report information of a back test, which is described in the following document:

http://www.zentrader.de/checklist_blackboxsystems_e.pdf

bye,
zentrader

I wouldn't recommend this, as you will lose a lot of valuable information by not using the individual trade data.

 

[Volker Butzlaff]

@NoWorries,

imho, wrong.

For a Monte Carlo Simulation based chance evaluation and risk analysis of a trading system concerning the existing back test results(!) - a so called "system simulation" - there is no need to look at the single trades...

bye,
zentrader

 

[NoWorries]

Quote from Volker Butzlaff:

@NoWorries,

imho, wrong.

For a Monte Carlo Simulation based chance evaluation and risk analysis of a trading system concerning the existing back test results(!) - a so called "system simulation" - there is no need to look at the single trades...

bye,
zentrader

How exactly does your simulation work then. If you don't use individual trades, you have to make at least some assumptions about their distribution (normality?), right?

 

[MustPlayOptions]

Here's the data from the performance. Thanks, I'm curious to see what you get for comparison:

Long + Short
Starting Capital $100,000.00
Ending Capital $323,032.07
Net Profit $223,032.07
Net Profit % 223.03%
Annualized Gain % 10.16%
Exposure 2.73%

Number of Trades $2,469.00
Avg Profit/Loss $90.33
Avg Profit/Loss % 1.84%
Avg Bars Held 2

Winning Trades 1,645
Winning % 66.63%
Gross Profit $444,500.79
Avg Profit $270.21
Avg Profit % 5.48%
Avg Bars Held 1.34
Max Consecutive 33

Losing Trades 824
Losing % 0
Gross Loss ($221,468.72)
Avg Loss ($268.77)
Avg Loss % -5.43%
Avg Bars Held $2.00
Max Consecutive $12.00

Max Drawdown ($16,689.03)
Max Drawdown Date 9/21/2001
Max Drawdown % -9.98%
Max Drawdown % Date 8/31/1998

Wealth-Lab Score 335.4942
RAR 372.6825
Profit Factor 2.0071
Recovery Factor 13.364
Payoff Ratio 1.0104
Sharpe Ratio 1.815
Ulcer Index 1.0874
WL Error Term 798%
WL Reward Ratio 1.2734
Luck Coefficient 8.9287
Pessimistic Rate of Return 1.9012
Equity Drop Ratio 0.2791

Winning Trades 1,645
Losing Trades 824
Max Consec Wins 33
Max Consec Losses 12
Max Winners Held 23
Max Losers Held 36

 

[Volker Butzlaff]

@NoWorries,

for such a mcs model the relation of winning/losing trades and the average profits/losses are enough...and (important!) you should run enough simulations to get a useable result!

You can analyse the mcs results themselves also with statistical reliabilty ("value at risk" concept), but I prefer the absolute results of the mcs runs.

bye,
zentrader

 

[Volker Butzlaff]

@MustPlayOptions,

my assumption: 10 year EOD data...so the profits are shown as yearly profits.

So a MCS with 10,000 simulation runs (each run with 2,469 trades) can show the following result in the gif attachment.

bye,
zentrader

 

[MustPlayOptions]

Thank you very much for running that. You'll notice though that you're drawdown max is only about 3k but the actual was 16k... I'm not sure why that is but that's not enough of a measure for me then...

Vince thanks for the links.

I've modified my algorithm slightly and now have a max risk listed of about 31k which seems more reasonable.

NoWorries: I did try the Excel MonteCarlo file again using the maximum excursions only and the last one that was run had a max drawdown of 151k which would mean a very large number of losses in a row (over 500) which doesn't seem to make sense for a strategy with 67% winners...

I've attached the data of the min and max excursions (net trade ignored since it's always within those bounds and including would just artifically decrease the variance).

The problem I have right now is that I don't know what the right answer is so while my latest adjustment probably makes sense, I have no way of knowing if it's correct...

I'm trying to nail the risk down because if I underestimate it's an obvious problem and if I overestimate then I'm not taking full advantage of the capital...

Thanks again to everyone for the suggestions/answers so far and for any further information.