Harvard, Oxford, Stanford... It doesn't make it any better. I wonder why anyone would do those tests and publications. I can not see any useful information in it. Ok, I also admit I do not understand some of those graphs and I really don't want to. MatLab must be able to be better in presenting results? What about rollovers for those futures. They can have some effect.

Of course rollovers are factored it. Quite a straw man...

You cannot use a single indicator on a single chart as a trading strategy - no matter the timeframe or the RSI period.

Soren, one should research each piece independently, not a cocktail of nonsense indicators (your argument is typically used by developers of some weak tools). Regardless of this obvious fact, the strategy has its developer -- L. Connors. The purpose of the test is to verify what is in a public domain. The purpose is not to argue about anything. It is what it is.

Stefan, I will chose to ignore your personal comment - looking closer at the stragtegy, it is a long only strategy, in the stock market. Which mean the above analysis by Mr. Conners is a waste of his time, and of the readers time.

Soren, nothing personal. I meant all developers of indicators mainly those on Bloomberg + TA folks. It is a very old song ("this is an indicator not a strategy"). Regarding long-only equity strategies, I can see large CTAs diversifying in this direction (e.g. Winton, etc.). Where there is demand, there is supply.

I say, developers should consider to move to currencies, instead of trading the stock market and the related futures. Stocks and stock indices are all highly correlated, difficult to test the robustness of a strategy across the stockmarket - the stock markets have been in a rally since 2009, in fact since1950 so this market cant be backtested for years ahead.

A strategy can be tested against 10 different non-correlated currency pairs - if not similar results across non-correlated currency pairs, then the strategy is most probably a curve fit.

Besides, currency pairs have no expiry, are traded 24/5, plenty of liquidity, easy access to historical data, low spreads and commissions.

@Stefan

I wonder if this code/description from 2014 fits yours: https://www.wealth-lab.com/Strategy/Details/296?

If so I can run the test and present the results here.

How did you factor in rollovers?

Which data did you use?

@Soren

I have not yet seen a strategy that fits all markets, currencies, oil, stocks, gold... Etc. IMO each market/investment class has different participants and underlying background which act different to the other class, based on passed experience and other influences.

When I replied to Stefan I was not questioning the purpose of the model or the quality of the test but the way it is presented. It feels like DOS ages. It feels like: I use MatLab because I am a mathematician. A simple proof of concept (or not) can be done easier, quicker and presented better. But maybe I couldn't call it "Oxford" anymore. ;)

Hi Volker, in simple tests where nobody cares about the term structure in futures (i.e. contract selection with rules to mitigate contango or beckwardation), continuous futures contracts are fine (as long as one does not use % in entry/exit rules -- a typical error made by some developers switching from stocks/fx to futures). Data: CSI is ok. Your strategy definition in the link is ok.

Regarding testing, I usually like to see sensitivity of parameters on a portfolio level. This is the core purpose of some tests on the blog. Most commercial SW cannot do sensitivity tests on portfolios -- they show just ONE dot on a map (SR=1.35, Sortino=1.5...). Magazines are full of these useless tests. Maybe your SW can do it. I do not know. I quit commercial SWs a decade ago because they usually limit a user in some way.

I apologize if you find outputs dense or boring. I admit I chose a style that fits more my research preferences. It is like private research notes made public.

@stefan You say you quit commercial solutions a decade ago - much has happened since then. I say when you were limited a decade ago by the limited platform offerings and their features, you are today limited by trying to develop everything your self.

Today is about to handle big data in the platform, about cross market aware strategies and portfolios, and most importantly its about time to market.

I say today its about to find the latest development platforms, and adapt to what they offer of features. There are many different types of strategies to start focus at, why not chose the ones which has the shortest time to market.

@Soren. The market is the ultimate decision-maker. In-house development is still preferred by those who can (#c/matlab/python/etc.). Retail traders not able to code could be interested in some alternatives. I can see you have developed a non coding platform which looks great. It will be interesting to observe in 5 years who are the clients. The time will show. But I am staying conservative in this area -- I think all people serious in research should learn coding; and maybe they will become better researchers in the process:-)

@Stefan - our clients are institutional clients, since 2010 - we have no retail clients, but will have soon. This is both a non-coding and a coding platform - if you want to code everything, you can do that.

Today you develop a strategy from scratch, every time - you do not do that in the future, where development platforms nativly have 99% of the features you need as visual development - the 1% custom code you add once, extending the visual environment. This is how the future looks like.

@Stefan - if you were using our development platform, you would never post something like this simple single chart RSI strategy - you would post examples of cross market aware portfolios of strategies, using probabilities in strategies etc. You would be focused at the conceptual aspects of a strategy, not at math.

@Soren. I'm am not interested. Thank you. Let's stop it.

@Stefan - you started mentioning our platform, not me.

Thanks Stefan for the private message - I always try not to mention what we do, but you posted some wrong information which had to be corrected.

@Stefan I stumble over your money management rule. 1. I wonder why you want ot use it on your test. You are eliminating some trades and the MM rule becomes a factor that might turn the system into a winner. When I test I want to see all the trades first and maybe later use a filter. 2. It seems you want your "Fixed Fractional" money management rule to be 1 % defined as: The average of the past 20 days time 6. In my short test that alone eliminates all or most the trades in SP500. I must be doing something wrong. Unfortunately I can not post screenshots here (it seems) otherwise I could show you that for example the 1% rule cancells the trade on 2003-06-24... And many others. Please help me. ;)

On a different note, I am not sure what you mean by: sensitivity tests on portfolios. May be you mean what I call: Parameter Stability Testing - on a portfolio level. I quit using none commercial software since Long-Term Capital Management went down. For the younger ones among us LTCM was run by finance veterans, PhDs, professors, and two Nobel Prize winners. Didn't that teach you a lesson?

Have you guys ever heard of the Best Market Timer in the U.S.? Me!!!