Actually it is your previous post (to which I reply now) that made me suspicious. You added italics at the wrong place: the word “subjective” should have been emphasized. If Aronson is familiar with research of cognitive biases (and it looks like he is) then he refers to the phenomenon of illusory correlation. That is, if you estimate an correlation intuitively, without actually computing it, then the estimates are often biased by cognitive heuristics such as the representativeness heuristic (which causes insensitivity to sample size), the availability heuristic (which causes overweighting vivid data and recency effects), and the anchor heuristic (which causes persistence and insufficient adjustment of prior though false theories). E.g. read Nisbett & Ross, 1980. But the poster to which you responded obviously wasn’t talking about intuitively estimated correlations, he was talking about properly established correlations. In summary, I suspect that you didn’t understand what you were quoting.
Anyhow, there is an easy way to falsify my hypothesis that you are Pal or his twin brother: Please tell us how you think about open drawdowns.
I think the point is that it pays off to trade a portfolio of systems. If you have 5 systems with identical expected return and variance, all you need is to have at least one system that does not correlate 100% with one other, to be better off trading a portfolio rather than a single system.
In other words, if the 5 systems are fairly equal in terms of reward/risk, you expect to almost always be better off trading all 5 of them, rather than just one. The lower the correlations, the better. But low correlations are not a necessity.
On the other hand, when using correlations for technical trading rules, you can usually only benefit if they are reliable and stable (or predictable).
Eu: Thanks for bringing up my blog. I hope it’s helpful to others.
It’s thanks for nothing. I think it’s helpful, because hmm… your approach is correct (imho, of course) The second you provide your data real-time with your opinion. The third you’re not using gambling approach. (of course you might be wrong in your estimations, but you don’t gamble with your money at C2) So… as educational material it’s very good 
Eu
>In other words, if the 5 systems are fairly equal in terms of reward/risk, you expect to almost always be better off trading all 5 of them, rather than just one.
Agree. The benefits of diversification will hold only if the systems in the portfolio are not perfectly correlated, for eg., System 1 trading stocks, System 2 trading Commodities and System 3 trading Forex. All the better if these systems are exceptional (Annualized return atleast twice the DD) to begin with, which was my point. But it is amazing what diversification can do to your porfolio on average, yielding higher returns and posing lower risk.
>On the other hand, when using correlations for technical trading rules, you can usually only benefit if they are reliable and stable (or predictable).
Agree. Correlations do change (become unstable), which makes following the shift in correlations even more important. Sentiment and global economic factors are very dynamic and can even change on a daily basis. Strong correlations today might not be in line with the longer-term correlation between two instruments. That is why taking a look at the six-month trailing correlation is also very important. This provides a clearer perspective on the average six-month relationship between the two instruments, which tends to be more accurate. Correlations change for a variety of reasons, the most common of which include diverging monetary policies, a certain instrument pairs sensitivity to commodity prices, as well as unique economic and political factors.
Thanks all for the feedback so far. There has been some really useful info.
"Recent research indicates that the ‘average expert’ (producing around 25% annualized return with around 20% DD) has little to offer investors/traders."
And how does this 25% compare to an S&P index fund, Brian (bought and held, of course, with no stoploss)?
