New: Correlation with S&P 500

You’ll see that in the system statistics box (viewable on each systems “Details” page) we now display the trading system’s correlation with the S&P500 Index. Soon we’ll also display those numbers in The Grid. Finally, within a week or so, we’ll begin to show intra-system correlations (i.e. we’ll allow you to find trading systems not correlated with each other).



Matthew

would be nice if you offered different instruments to be the correlation point.



Forex could track against the US dollar… it means not so much how forex tracks the S&P 500…

Fantastic addition.



While sophisticated C2 members probably figure this out on their own prior to subscribing to a system, this number should frankly be very visible to all potential subscribers.



Once we can easily view the correlation of returns of the various C2 systems, I think C2 will have reached the next significant level of real-world usefulness.



It is also important to note what period (daily, weekly, monthly) is being represented. Ideally it would be daily, although I can understand if the C2 back-end has to make due with monthly. Honestly though, monthly figures will be of very limited use on C2 – you’d have to wait two years at least to get a decent correlation representation.

All figures use daily data.

Also, and this is definitely nitpicking, I don’t recall ever seeing correlation go to the third decimal place. It just seems odd. It also isn’t keeping with the C2 general conventions. Tracking to the 1 thousandth decimal point? Again, I’m nitpicking – but I think it just serves to confuse people to have such strange discrepancies – as if the thousandth-place calculation is important here, but not on other measure.



Again though, kudos.

Well, Dustin MK is doing search for babies milk of system vendors harder and harder :frowning:



Once we can easily view the correlation of returns of the various C2 systems, I think C2 will have reached the next significant level of real-world usefulness.

I would agree.

Eu

I guess you never opened the advanced statistics? There the correlations are in three decimals too. Or try SPSS: this reports correlations default in three decimals too. Nevertheless, I agree that two decimals is often enough and that there are usually not enough data to attach any weight to the third decimal. The exception is if you want to compute something else with it and if you want that outcome also in two decimals.

Yes, the benchmark is a problem for forex systems. I believe a common way to define a benchmark is to base it on factor analysis. A simple approximation in that line would be to define the benchmark as a weighted average of a subset of C2 forex systems, and to recalibrate it at regular intervals (e.g. each year). The result will indicate what happens to your account if you blindly subscribe to all these forex systems and put an equal amount of money on each of them.



But the details must be right. Which systems should be included in the subset? I think this should be systems that have a reasonable chance to get subscribers, so I suggest



- only active systems

- only systems that are older than 3 months

- only systems in the top 20 with the Sharpe ratio (or another statistic?) in order to avoid that the benchmark will be loosing all the time :slight_smile:



Another possibility would be to take the top 20 systems with respect to number of subscribers, but it will be hard to mimic that because nobody but MK knows those numbers.



Given the short life span of most system it may be good to recalibrate more frequently than once per year; say at the end of each month.



Similar benchmarks can be made for stocks and futures systems.



(Even more in line with factor analysis would be to include only systems with the highest average correlation with other forex systems, but that is more difficult to implement and I am afraid that it can lead to a selection of only loosing systems.)

I’m not sure what correlation with a “benchmark” tells you about a system. If I trade futures on the S&P or ETFs indexed to the S&P and I go long/short in my trading, I’d want as high a correlation as possible. A value of 1 would indicate I was right in the direction every day.



But if I only trade from the long side, my correlation coefficient is going to be less, even if I’m right every day the S&P is up.



To me the value of correlation is with the equity curves of other systems. If I find 5 positive expectancy strategies that are lowly correlated, and they suit my trading style, thats money.

True, but the point was, that the S&P has no bearing on the Forex systems.



I was just suggesting that the US Dollar (the futures contract traded in New York, & based on a number of different currencies, 19 or so) would be more appropriate for Forex, since it is also a currency.

I don’t understand what you mean by “A value of 1 would indicate I was right in the direction every day”. I think you can only achieve a correlation of 1 (between the returns of your system and the returns of the S&P) if you’d stay long all the time. If you’d be right in the direction every day I’d expect the correlation would be zero: E.g. on a day when the S&P would show a return of +3%, your system would also show a return of +3% (or a multiple, if you’re using leverage), but on a day when the S&P would show a return of -2%, your system (short in that case) would show a return of +2%.

The correlation with a benchmark tells you to which extent you could have obtained the same result with a blind stategy (the benchmark) like buy and hold S&P.



"A value of 1 would indicate I was right in the direction every day."



No. The value will be 1 if you are long all the time. If you go long and short at the right time, the correlation will be smaller and can be close to 0.



"But if I only trade from the long side, my correlation coefficient is going to be less, even if I’m right every day the S&P is up."



No. If you buy and hold a long position in S&P, then your correlation with the S&P will be 1.



"To me the value of correlation is with the equity curves of other systems."



That is also a valuable application of correlations, but it is not the only one. One application does not exclude the other. The concept of a correlation is just as general as the concept of a mean. I suppose that you don’t believe that there is only one set of numbers for which it is useful to compute the mean? :wink: Same with correlations.

Yes, I agree that the S&P has no bearing for forex.



I didn’t get that you meant that futures contract. I think that it is better than the S&P, but that my suggestion is even better (although a lot more work). The problem that I have with that future (if I understand what it is) is that it has one base currency. So it is comparable to something like the price of Microsoft shares divided by the S&P. A similar point is that for forex it is just as natural to go long as to go short.

If you’re referring to the Dollar Index (USDX) traded on NYBOT then it only represents 6 major currencies vs the dollar, not 19, and from what I remember the Euro has a huge weighting at about 60% with the Yen next at about 12% so I’m not sure how fair a benchmark the Dollar Index would be either though obviously for Forex it’s still better than the s&p.



An alternative suggestion might be the Stable Currency Benchmark (SCB) that Bob Prechter developed a few years ago (relax, it has no links with Elliott Wave theory) purely as something to accurately measure a market’s performance. It makes a lot of sense. http://www.stablecurrencybenchmark.com



the Dollar Index was once 19 currencies in years past. I had not reviewed its recent construction.

Yes, must have been before the Euro came into effect over 8 years ago.

Prechter has also used gold as a benchmark. My first though

was to use a gold / dollar (or maybe better gold / swiss) index hybrid. Just using the dollar index doesn’t do much, except tell you how the euro is doing vs the dollar.

Gold is the ideal benchmark as it is the ultimate commodity. Better suited than the S&P 500 because it can be better applied across financial instruments like stocks, commodities and forex.

> Gold is the ideal benchmark…



Still you must consider how to value gold… dollars, francs, ec’s?

GC and YG are dollar based…so if the dollar is going down it may

appear gold is rising. In this case it is only costing more dollars and

not actually changing in absolute value.



It does make sense to compare stock related investments to an

S&P benchmark.

>It does make sense to compare stock related investments to an

S&P benchmark.



No argument there. But, so does gold.