Here's a statement of the obvious: The opinions expressed here are those of the participants, not those of the Mutual Fund Observer. We cannot vouch for the accuracy or appropriateness of any of it, though we do encourage civility and good humor.
Support MFO
Donate through PayPal
Morningstar Ratings To Expand Six-Fold, To 10K Funds
I get the feeling that M* is flinging around terms to make things sound more impressive. The FA article, like M*'s FAQ, speaks of a rules-based system. But M*'s paper on this system (linked to in the FA article) talks about statistical models, and says it makes minimal use of rules - just to avoid anomalies (two different share classes of one fundhaving different Process or Parent ratings, taking Analyst Parent rating if it exists over the statistical rating in case they differ).
The FA article talks about "how the automated methodology performs with out-of-sample data, that is, in real time". Out of sample and real time are simply different concepts. The former deals with testing a model using data other than that used to build the model; the latter deals with speed, e.g. is the analysis batched overnight (likely, since MF data is reported daily), or is it processed as it comes in.
Whatever. A drawback of statistical models is that they're lousy at explanations. That's what analyst reports do, and why I am skeptical of the value in these "quantitative" ratings. Single numbers such as star ratings convey fairly limited information. Here where the information is already speculative (forward looking) to begin with, "Gold", "Silver" and "Bronze" ratings without explanation likely won't help much.
Comments
https://www.fa-mag.com/news/morningstar-ratings-to-expand-six-fold--to-10k-funds-37760.html?print
I get the feeling that M* is flinging around terms to make things sound more impressive. The FA article, like M*'s FAQ, speaks of a rules-based system. But M*'s paper on this system (linked to in the FA article) talks about statistical models, and says it makes minimal use of rules - just to avoid anomalies (two different share classes of one fundhaving different Process or Parent ratings, taking Analyst Parent rating if it exists over the statistical rating in case they differ).
The FA article talks about "how the automated methodology performs with out-of-sample data, that is, in real time". Out of sample and real time are simply different concepts. The former deals with testing a model using data other than that used to build the model; the latter deals with speed, e.g. is the analysis batched overnight (likely, since MF data is reported daily), or is it processed as it comes in.
Whatever. A drawback of statistical models is that they're lousy at explanations. That's what analyst reports do, and why I am skeptical of the value in these "quantitative" ratings. Single numbers such as star ratings convey fairly limited information. Here where the information is already speculative (forward looking) to begin with, "Gold", "Silver" and "Bronze" ratings without explanation likely won't help much.
Regards,
Ted