Capital Available for Traders

[mickson]

Nonlinear, excellent questions indeed.

See my comments next to each point.

Mike

Michael/mickson, thanks for making this service and website available. I signed up and uploaded my IB Flex statement by following the instructions, and it was straightforward, except a few minor things where your instructions referred to certain menu options in IB which have apparently changed. I made it to the "top 15" in the leader-board, actually. Not much competition there yet.

Yes your comments on our tutorial are fair, this is certainly an area for improvement in 2013 on the issue of competition I like your confidence but I wouldn't be so quick to dismiss your competition these returns are pretty impressive and your return is just 2 months.

I have some questions:

1. The account performance metrics include standard quantities (such as PnL, Sharpe, Sortino, Max DD, stddev, alpha, beta, and expectancy). Are these metrics used by the RAPA score? Are there any other quantitative measure which affect the RAPA score? What's the point of keeping the RAPA score "proprietary"? I mean, if it's truly a good metric of trading performance which incorporates the risk and reward, it would be impossible to manipulate, right?

Most of our competitors and other social trading websites apply a very basic ranking approach focusing on % performance. Ours is a very risk adjusted focus. Part of our value add is our ability to efficiently Score a traders performance. Our algorithm is dynamic process which we re-calibrate from time to time against the database we used for optimizing. There are other factors to the algo than the ones you mentioned, some of them have not yet been fully "turned on". For instance size of account plays a role, as not all account sizes are equal the same applies to track record duration. There are other factors that also measure deviation of performance characteristics, clearly this factor has less effect on a shorter than a longer track record. Lastly, we have done some more work on option writing type Sharpe ratio measurement, where we have adapted one of the latest new formula's in this type of measurement. We have not yet updated our RAPA Scoring algorithm since our launch, we plan on doing an update in early January or perhaps later this month. We will notify our members when this happens.

2. Since the Flex query generated by IB is just an XML file, it can be easily edited by the user and uploaded to your site as a stellar trading record. How does your service prevent this from happening?

Yes this is a definite possibility hence our request for admin rights so that we can fetch the XML file directly from IB. To make things as flexible as possible (excuse the pun) in the early stage we are allowing some leeway. In order for us to make an allocation one of the criteria will be to verify that the performance updated corresponds to actual broker notes. We may from time to time ask members on the leaderboard to send us something to help us "cleanup" in case of tampering.

3. How do you plan to keep in sync with the IB XML schema which drives the Flex report? It can change at any time, which means that your parser is destined to fail when this happens.

We will have to deal with this on a case by case basis. I am sure we will sort out any issues very quickly if it comes up.

4. If I opt out of the option which gives you the admin right to my account (yes, I understand that it only has access to reports), can I upload the Flex report periodically (such as weekly)? Would these incremental updates be correctly processed by your RAPA site?

Yes they will work fine.

Thanks.
Mike

 

[mickson]

Quote from bRobbins72:

If you are all traders what do you need more traders for?

We number 5 traders, perhaps you haven't heard Harry Markowitz's comment that the only free lunch in financial markets is diversification.

That is why?

 

[nonlinear5]

Quote from mickson:
on the issue of competition I like your confidence but I wouldn't be so quick to dismiss your competition these returns are pretty impressive and your return is just 2 months.

I agree. What I found surprising is that I made it to the top, even though my record is so short. In my own ranking equation which rates my trading systems, I use (among other things) the term sqrt(number of trades) as a multiplier, which essentially rewards statistical significance, or to put it differently, it punishes large standard errors which originate from a small number of trades. Just wondering if this term is in your equation.

Another related question is, do you use the Omega functions by Shadwick and Keating (http://www.effas-ebc.org/fileadmin/projects/Convenants/OmegaBasic_-_Con_Keating_.pdf) in any shape or form? These two fellows proclaim that the Omega metrics are far superior to Sharpe, Sortino, stdev, and any other metrics which make an assumption of the normal distribution of returns.

 

[CT10Gov]

Quote from nonlinear5:


Another related question is, do you use the Omega functions by Shadwick and Keating (http://www.effas-ebc.org/fileadmin/projects/Convenants/OmegaBasic_-_Con_Keating_.pdf) in any shape or form? These two fellows proclaim that the Omega metrics are far superior to Sharpe, Sortino, stdev, and any other metrics which make an assumption of the normal distribution of returns.

Arg.... this kind of subtle incorrectness annoys me every time... there's no normality assumptions behind sharpe, sortino, or std. For christ's sake, there's an explicit standard deviation (or the second moment) for EVERY freaking distribution (even in the pareto family, it's just infinite in those cases). Likewise, there's an mean for every single distribution too. So how can the sharpe, being a ratio of mean over std, require any distributional assumptions?

Now, the sharpe might be a bad measure if the returns are due to certain type of distribution. But it's a bad measure because it makes normality assumptions (because it doesn't).

And the omega ratio isn't exactly a cure all either... it requires that you estimate the full empirical distribution. Different task, same problems.

 

[vkrouglov]

Quote from nonlinear5:

I agree. What I found surprising is that I made it to the top, even though my record is so short. In my own ranking equation which rates my trading systems, I use (among other things) the term sqrt(number of trades) as a multiplier, which essentially rewards statistical significance, or to put it differently, it punishes large standard errors which originate from a small number of trades. Just wondering if this term is in your equation.

Another related question is, do you use the Omega functions by Shadwick and Keating (http://www.effas-ebc.org/fileadmin/projects/Convenants/OmegaBasic_-_Con_Keating_.pdf) in any shape or form? These two fellows proclaim that the Omega metrics are far superior to Sharpe, Sortino, stdev, and any other metrics which make an assumption of the normal distribution of returns.

On the RAPA Cap Intro website we are using the Sharpe ratio as a proxy for our performance measure. This is a 'gold standard' in the industry and we feel that our rating has to be intuitive in a sense that performances that are considered good according to the Sharpe ratio will get decent RAPA Score. There is also a proprietary part of the score where we believe we have an edge over the usual Sharpe ratio. For example, we are adjusting the Sharpe ratio to the possible manipulations with options and/or "martingale type strategies".

In our equation, instead of trades we are using the number of days in the trading history. IMHO, one should use equally spaced in time observations to assess the performance.

Concerning the Omega - it is a function and you cannot assign a single number to it. We could make a subjective decision as of what is the level of return we are interested in and compare according to Omega(this return) but this would once again be misleading as we will force the traders to step on a quite dangerous route of extra leveraging their strategies to get better Omega scores as all things being equal Omega prefers more leveraged returns.

 

[oldtime]

if I was young just starting out I would be eating this up. Start with just 10k with a chance at managing 100k. Even though at 1% that is only 1k a year, you get your name out there, and people start talking about you. You can be sullen and withdrawn, and you don't have to kiss anybody's ass, as long as you are trading well.

If you get lucky and get hooked up, that is a big jump early on.

if not, just plod on, by by the time you get old you will have lost all interest in OPM. You already have enough to pay the bills and you don't need the aggravation.

I worry about people that care more about their reputation than their ROI.

 

[Jack_Larkin]

Quote from marketsurfer:

http://www.elitetrader.com/vb/showthread.php?s=&threadid=118174&highlight=Mt4+scam

MT4? Wasn't this platform exposed years ago as being a scam?

I guess if you use this software you can catch them at their own game---LOL!!

http://4xtrader.net/forex-tools/

Kinda spreading FUD with that comment.

Evil brokers will be evil despite the software they use.

MT4 Server has a bunch of extensions by third parties to do various tasks, dealing desk brokers who want to screw with clients have tools at their disposal to do so..

..but you'd be kidding yourself if you think that other brokers who don't use MT4 are automatically innocent of such actions themselves. If they built their own platform and wish to act shady, you bet your ass they have the same "evil" functionality coded in.

There are a lot of decent brokers using MT4 that don't do anything shady to their quote feed or clients. You just need to stick with decent, well regulated, brokers who pride themselves on clean execution.. they aren't hard to find.

 

[Jack_Larkin]

Quote from mickson:

Today we launch beta support for MT4 traders.

All you need is an html MT4 file to upload it onto the RAPA platform.

While it's great that you've made progress with supporting FX brokers, I'd be very cautious about going the 'report upload' method of tracking performance.

The report HTML files that MT4 generates can be doctored quite easily, and there's no way to verify if it's a real money account or demo.

--

Take a look at what MyFXBook is doing; they have a setup where they use the investor password for an account to directly pull the account's info from the broker's server, which makes doctoring the records much harder and the broker verifies the real money status of the account.

 

[nonlinear5]

Quote from vkrouglov:

Concerning the Omega - it is a function and you cannot assign a single number to it. We could make a subjective decision as of what is the level of return we are interested in and compare according to Omega(this return) but this would once again be misleading as we will force the traders to step on a quite dangerous route of extra leveraging their strategies to get better Omega scores as all things being equal Omega prefers more leveraged returns.

Right. Shadwick and Keating themselves warn against using a single point in the Omega function to rank the strategies. Rather, they insist on the comparison of the entire function. They apparently have something what they call the "F2 score", which ranks the entire thing. I was going to use it for my own ranking system, as I compare literally millions of trading strategies, but I can't find the description of how to calculate the F2 score anywhere.

With regards to the verification of the trading record, I am thinking the best way would be to have an arrangement with IB to provide what I'd call an "auditing API". The trader would add the user to the IB account with the access to the auditing (and nothing else), and you guys use the API to do your thing. This would be a trivial thing to do on the IB's part, but convincing them to implement it is another thing, of course.

 

[nonlinear5]

Quote from vkrouglov:
In our equation, instead of trades we are using the number of days in the trading history. IMHO, one should use equally spaced in time observations to assess the performance.

I thought about this for quite a while even before this discussion has started. It certainly makes sense to include the length of the trading history in the ranking equation. However, I think that the number of trades is equally (if not more) important than the length of the history.

Consider two trading strategies, A and B.
Strategy A has a history of 10 years, and made 10 trades.
Strategy B has a history of 1 year, and made 100 trades.

Everything else being equal (i.e, all the other performance metrics are the same), which one do you trade? According to your RAPA score, and assuming that you use the term sqrt(history length), strategy A will get the score which would be 3.16 times higher than that of strategy B. But is A really 3.16 times more attractive? I'd say "no". The sample size of 10 trades is too low, and thus the probability that strategy A accomplished its performance simply by chance is too high.

I'd say that if you take into consideration both the length of the trading history and the number of trades, it would be more fair:

Strategy A: sqrt(10) * sqrt(10) = 10
Strategy B: sqrt(1) * sqrt(100) = 10

What do you think?