Risk of Ruin (Monte-Carlo

My question to every one is how much do you take this into account when reviewing a possible new strategy to subscribe to? I know there is a lot that goes into choosing a new strategy draw done, capital, length since started, what it trades, etc. But I am not real educated on the Monte-Carlo system and wanted some of you to chime and tell me your thoughts. Thank you in advance for sharing.

Given the length of history of most strategies I usually discount the risk of ruin figures. Unless the system has a high trade frequency and/or multiple years of history then the significance of the ROR figures should be taken with a pinch of salt.

Ask them do they use guaranteed stops and what is the maximum they will loss before cutting a trade. They problem with stats is some managers learn from their mistakes. Some of the best managers either learn from their mistakes or die so duration of strategy is key. 3 or 4 years but 5 is better.

Sorry my spelling is awful lose instead of loss.

I’m familiar with Monte-Carlo simulation and have written my own utilities using it. Intuitively I like to think of MC simulation as presenting ranges of potential returns/drawdowns and their likelihood for a system, based on actual past trade data and presuming the system continues performing roughly as it has in the past. The real equity curve of a system shows you what the system has done, but not what the system would’ve done if (for example) several bad trades (or good trades) occurred more clustered together, in a different order, etc. MC simulation shows you returns/drawdowns that could’ve happened if trades had happened slightly differently, but with the same general tendencies of how the system has performed so far.

MC simulations have limitations. They presume a system will perform roughly as it has performed in the past, but in reality a system is not constrained to the past and can do anything. If a system suddenly starts acting differently than it has before, a MC simulation on previous data will not be accurate. Also MC simulations on small amounts of data are likely not very accurate.

But despite the significant limitations I find MC useful. In particular I like to see MC simulation estimates on drawdown and return that roughly match backtest data to give me further confidence in the backtest data being a good representation of how the system will perform. And in the absence of backtest data I like to see the drawdown estimations from MC simulations as the live max drawdown of a system is often not representative.


If system has long equity, then this approach is useful. You can see all alternative scenarios, and find out the worst case scenario and its probability. For one month age systems with 1000,000%+ annual return and 10% max dd it is useless.

This observation is applicable for almost all criteria (max dd, profit factor, calmar ratio etc) here in C2.

thank you all for your thoughts