I quite enjoy reading your posts. Very good and very thought out. As you mentioned, you are not saying it is the best way, but at least it is well thought out and therefore, if there is an issues, it should be easy to find where it went wrong based on how you are thinking through things.

Can you help answer these questions from other members on futures io?

Thanks. Of course, if anyone thinks there is a better way for me to do any of us (better position sizing, more reliable and cheaper broker, etc), please feel free to contribute. Such suggestions will make this thread even better, as we test, evaluate and compare the alternatives.

In a previous post, I presented some results with fixed fractional position sizing. Basically, the results say that in an "average" year (meaning, 50% of years will be worse, and 50% of years will be better), I expect to make $30,735 profit in that year, and hit a maximum drawdown of 38.1% sometime during the year.

That profit number seems a little too good to be true...and my motto is "if something seems too good to be true, it probably is." And that profit number does seem too good to be true. 362% rate of return in that first year. Seems high...

Remember, though, the actual rate of return could be just about anywhere on the spectrum. It is just that the 362% is the median value.

Below is a histogram of possible returns. It will be interesting to see if the Year 1 results is anywhere close to the median ending equity (black vertical line). If it is, I will be very happy. I'll still be happy if I even hit the 25% mark, which is a final equity of about $21000, which is 147% return for the year. Still in "too good to be true" region.

The following 2 users say Thank You to kevinkdog for this post:

Is that Monte Carlo? If so, then that is not quite what it is saying though it is a similar statement. That technically is saying that 50% of the time, final balance is that amount or less.

Yes, it is based on Monte Carlo. Can you explain the statement "that is not quite what it is saying..." - I am not sure I understand.

Are you referring to my statement: "Basically, the results say that in an "average" year (meaning, 50% of years will be worse, and 50% of years will be better), I expect to make $30,735 profit in that year, and hit a maximum drawdown of 38.1% sometime during the year." I know I am using the term "average" incorrectly in this sentence (hence the quotes around it), as it is really the median value - the true average would be much higher, due to the spike at the far right of the histogram.

@kevinkdog: I'm an attentive follower of your thread and very appreciate the granularity of your analysis. In another thread we have discussed the measuring of drawdowns und i've wrote my primary style of drawdown analysis here https://futures.io/traders-hideout/24308-how-compare-drawdown-metrics.html#post279260. Recently i've found a term for my Kind of analysis "Start-Trade DrawDown". It's mentioned in Keith Fitschen newest book "Building Reliable Trading Systems" which i can strongly recommend.

With a default monte carlo analysis you ignore the result dependency of subsequent trades. I refer to "Winning strikes" and "Loosing strikes". It could be a huge difference in your money management if you have an strongly trade-to-trade correlated system. In one of my books i have an algorithm to test that. With an "Start-Trade DrawDown" analysis you can see what your largest drawdown would be, if you had started your system at every day in the last x years.

I'm courios to see the difference between both methods.

Koepisch

Edit: "result dependency of subsequent trades" means "serial correlation" - thanks kevinkdog, sometimes it's hard for a non native speaker

If there is serial correlation in your data (where the result of Trade X is related to the result of Trade X-1), the Monte Carlo analysis is not 100% appropriate.

To test this, I usually use a Durbin-Watson statistic. I find, for most of my systems, they do not have this serial or auto correlation issue.

When I get a chance, I will try to run the "start trade drawdown" - I think it would be useful. If it looks a lot different than the Monte Carlo, then that will be VERY interesting!

The following user says Thank You to kevinkdog for this post:

NOTE: This post scratches the surface of a complicated issue...proceed at your own risk...

OK, I ran the analysis. For everyone reading this, the issue is that, depending on your trade data, Monte Carlo analysis is not always a correct tool to use.

So, to first test this (is Monte Carlo OK to use?), I run something called the Durvin-Watson statistic. It checks for positive and negative autocorrelation. I think you can google search to find this for an Excel spreadsheet, what it means, etc. I'm no mathematician, so I will defer from getting into detail on this calculation.

Suffice it to say, for my NGEC system, the analysis says there is no autocorrelation issue, so I can use Monte Carlo analysis.

So to answer Koepisch's excellent question - how does Monte Carlo analysis compare with the "Start Trade Drawdown" analysis he mentions a couple of posts ago?

Again, without getting into too much detail, here are the results:

With the uncertainty of this whole analysis, it says that the 2 methods give very similar values for drawdown. The Monte Carlo gives more "worst case" scenarios - higher drawdowns than the Start Trade method, and this shows up as a greater standard deviation.

Thanks, Koepisch, for helping me add to the analysis here. The "Start Trade Drawdown" approach is verty interesting and useful.

The following 3 users say Thank You to kevinkdog for this post:

Excellent and fast analysis, great @kevinkdog! Due to the similarity of the results i would trust even more in the system. Furthermore you can trust your monte carlo analysis related money managment decisions - because it's BACKED with practical "What-if" "Start Trade Drawdown" analysis.

The following user says Thank You to Koepisch for this post: