recently i saw several system has a max DD < 20% while have a risk of 20% account loss > 0. can anyone help me understand that?
It can happen (risk > 0), but so far it didn’t happen (max DD < 20%).
More intruiging is that there are also cases where the risk is estimated 0 while the event has actually occured in the past.
Yes, and risk can never be zero anyhow, no matter how good the system. So I question the concept of estimating future risk as zero, or even close enough to round down to zero.
I believe this can happen because C2 calculates "risk of 20% account loss" by using Monte Carlo simulations, not actual trades.
thank you, CEO. now i’ve got the idea. so here comes another question. if only two factors, max DD and risk of x% account loss are considered, which system should get more credit, the one with more max DD and less risk, or the one with less max DD and more risk?
IMHO, there is more information needed to answer that question. For example, what if the system is 5 years old and the Max DD happened 4.5 years ago. I would give more leeway to that system compared to a system that is 1 month old and had the same Max DD. I suppose the "risk of x% account loss " in my example, might be more accurate. Another problem is I don’t know what exactly goes into the C2 calculation.
The former, as it suggests although the risk of that drawdown was low it’s already happened. Even more so if it’s been going longer. Generally speaking the older a system is, the more likely it will have already endured its maximum drawdown.
I also wish there was a way to factor in multiple max DDs. Which is worse, a system with a 45% max DD nad otherwise fine, or a system with four DDs which all exceeded 35% but not 40% ??
Good point Index.
I suppose you could look at the equity curve and manually appoximate it.
to me,. max DD is too sparse or of minor value. Even though widely used in the industry. Obviously, past max DDs are not really a reliable indicator of what will happen in the future. Regardless of the whistle and bells people try to add on to it (set aside 2x max DD, etc.). The market’s future performance is often rather independent of the past.
If someone has only 1 large max DD and the rest of its DDs are modest, and someone else has quite a few large DDs but are all smaller than the first, then the max DD for system A is very misleading. I would be much more interested in the protecting myself against a system with multiple large DDs, as it seems the vendor has a real problem with money management. A single large max DD could be just a learning-to-use C2 problem, or a single market spike.
Right – the "probability of loss" is based on weekly Monte Carlo simulations run on each system. The Monte Carlos use very fine-grained intra-day drawdown stats, and not big-grained trade-by-trade stats, which I think is a more realistic measure of probability of loss.
I personally wouldn’t subscribe to any calculation that doesn’t consider the past margin requirements for a trading system. In other words, the margin used during trading is money “at risk”. Margin requirement is often adjusted based an instruments volatility.
You can do Monte Carlo runs til the cows come home but until margin requirement is included in the calculation you won’t have a realistic assessment of % account loss.