Matthew,
I’m not sure the risk of losing 20% on account stat is working properly or as intended.
Can you please explain how “trending futures” can show a 30% chance to lose 20% of an account, when other systems this one for example:
http://www.collective2.com/cgi-perl/systems.mpl?want=publicdetails&systemid=15829169&
shows a 0% to lose 20% on account with “EXTREME” draw downs in its stats…
Thank you for your time,
at
The stat is not perfect, nor should it be treated as a guaranteed prediction. The stat you are asking about is the result of a Monte Carlo simulation. Essentially, C2 runs tens of thousands of simulations on your system, replicating the system’s intraday gain/loss profile many, many times. The stat in question reports how often in the simulations a certain loss was achieved. The stat should be regarded as simply one weapon in a trader’s arsenal, and she be used in conjunction with other statistics and analysis provided by C2.
Thank you for the reply. I understand that “one stat” is to be taken as a “grain of salt” so to speak, but it appears that that one “simulated” stat alone takes the TF system out of being one of the “best” on C2 … (risk of 20% loss must be < than 5% as stated on C2)
ok… so the system, and me as a trader are not perfect, but like the rest of us as a vendor, I am trying to run a good solid service / system that subscribers can rely upon, and for whatever reason - maybe one fluke trade early on, or what… as I am not sure exactly what has caused that stat to show up and give the impression that someone using the trending futures system has a 30% chance of 20% account loss.
Perhaps if the stat were weighted (which it may or may not be?) but if it were weighted on more recent trades / positions, I would think it would be a more realistic stat for observers, as well as I believe more weight should be put on the risk of 50% and / or 100% account loss for the “best systems” availability.
As a vendor, the biggest thing we want to be able to do first and foremost is to provide a “winning” system for our subscribers. One that will make money over the long haul. And also, we need to gain the “trust” over time in being able to show that in fact that is what we are doing. I don’t know about the rest of you, but if I did a quick search here on C2 of “the best” systems on C2, I would be weary of any system that didn’t show up in that list. It leads me to believe to have a “best” system is one that should show a potential subscriber that these are the vendors that would likely to be one that may have a higher “trust” factor than others. Am I correct on this assessment?
IMHO more weight on the “best of systems” could also be included on other “factual stats” than a “simulated / perhaps / maybe” type of stat which is what it appears to me the “possible risk of account loss” stat is showing.
Let’s take a quick look at what the factual stats / results that TF has demonstrated to date:
5 star independent sub rating
> 90% realism
> 2:1 profit factor
Avg Win > 2x Avg Loss
Avg Win = $1,975; Avg Loss = $878
> 90 days in existence (121 days)
> 1 APD (1.23)
> 1 Sharpe (2.78)
> 90% keep after worse case slippage (96.9%)
> 100% annual return (479.8% over 121 days)
Risk of 50 or 100% account loss = 0%
What do all of these stats taken as a whole mean to the casual observer?
A pretty good system when taken collectively. One of the best? You be the judge.
But then we have the “Risk of 20% account loss = 30.0%” (which > 5% takes it out of the “best” system C2 listing) … which seems very odd to me to take the system out of the “cut” for this one oddball stat.
Thank you for your consideration to review what makes a system “one of the best” here on C2.
sincerely,
at
Aspire Trader
The question is whether the 36% drawdown shown for your system is accurate. If so, you can’t question the probablity of a 20 percent drawdown being significantly greater than zero.
What always bothered me about the “chance of 20% DD”, is when you see a system that may already have 1-2 DDs > 20%, but it might only list a 30% chance of DD.
Too often, we see systems that later blow up, when they were listed with a less than 50% chance of 20% DD.
Although you may not be able to factor in a vendor’s future carelessness, there are often clues. I think your chance of 20% DD needs to consider its existing history and its leverage factor, not just the Monte Carlo simulation.
Maybe I can explain why the outcome is reasonable.
According to the advanced statistics (daily all, quartiles of return rates), there was one day with a return rate of 0.839, which means a loss of 16.1% of your account value. Your system is 122 days old. Suppose that someone trades 122 days and on each day he receives a return that is randomly selected from the 122 daily returns of your system. That would be a system with properties very similar to yours. The probability that he gets at least one day with a loss of 16.1% is 63.4%. This can be computed easily with a binomial distribution. This loss of 16.1% is already close to 20% account loss. There were also 7.6% outliers in the return rates, and the probability that at least one of these outliers happens in the next 7 days is 42.5%. On average these outliers were 0.949, which means a 5.1% loss. If there are one or more days with a 16.1% loss of which at least one is followed within 7 days by another outlier loss (average 5.1%), then you will probably have a drawdown of about 20%. The probability that this happens is about 63.4% * 42.5% = 27%, which is close to the 30% probability that C2 reports on basis of simulations. My computation is only a crude approximation (because not every outlier is 5.1%, in these 7 days you can have some profits, and there are other scenarios in which your account loss exceeds 20%). But this approximation suggests that the outcome of the C2 simulation is correct.
This computation also tells you which factors influence the outcome of the simulation: The number of days, the presence of days with large losses, and the percentage days with moderate losses (which can build up).
In other words, what the C2 simulation tells the subscriber in this case is "In its relatively short history of 122 days, this system had already one day with a rather large loss, and several days with moderate losses. If you trade this system 122 days, then there is a substantial risk that drawdowns exceeding 20% will happen."
I think this is important to know and to use. And it is based on facts, namely the fact that you had a 16% drawdown within 122 days.
However, I am not sure how the reported statistic is influenced by the number of days. Suppose that two systems both have a 1% probability of a day with 20% loss, and profits on all other days. Suppose system A has a history of 100 days and system B has a history of 1000 days. Using a binomial distribution, the probability that system A has at least one day (in 100 days) with a loss of 20% is 63%, but the probability that system B has at least on such day (in 1000 days) is 99%. This would be unfair, because systems actually have the same chance per day. I do not say that this is what C2 computes; I don’t know (MK told me some details, but I don’t remember them). If this (63% and 99%) is basically what C2 would report then I suspect a bias against older systems. Then I suggest that it is better to annualize the simulation statistics, i.e. base all simulations on a period of one year. Or, if this is not feasible, I would second the motion not to use them in rankings of the systems.
AT,
Right below the chart, where it says ‘showing equity’, you should click on ‘intraday’. You’ll notice there is a massive intra-day bar on March 13, indicating your equity swung between $40K and $70 on a single day. I don’t know if that is an error or not, but I could imagine such a swing could have a large impact in a Monte Carlo simulation.
Hi ST et al,
I have a simple question: Is the risk of 20% DD (<5%) stat with referring to intraday or the system (typical) DD?
Gilbert
[LINKSYSTEM_30873120]
Any point-to-point drawdown (intraday included).
Thanks for pointing that out Science Trader! Yes, I see what you are talking about now. Maybe this is the issue all along? In review of the records, that was a CRUDE OIL trade, April 08 contract. March 13 was the day the position was STC for a profit. Here is the order data as showing up in the track record:
BTO 0.658 QCLJ8 104.43 3/5/08 15:50
STC 0.658 107.96 3/13/08 18:15
draw down & risk: ($1,039) High
profit: $2,319
Matthew, can you please verify the accuracy of this trade should be causing the issue in question here? I don’t think there was that large a draw down specifically on the March 13 day on this trade?
