>> Ray Dill>
>
>>> Maybe we should calculate a coefficient of information against,
>>> say, 30 day post facto results?
>>>
>>> I have my doubts about the merits of technical analysis,
>>> especially automated technical analysis, but a least-squares
>>> regression line does not lie.
>
>
>>>
>
>
>>> Shall we run the experiment?
>>>
>
>
>> Who are we? You and who else? What is your relation to the original >> poster, if any?
>
>
> I have no relation to the original poster; the "we" I referred to
> was all interested parties; I can do the math, you can check my
> code if you like. I'm a Knowledge Engineer who's been working
> independently on stock tracking and market information extraction
> (among other things) since the dot-bomb.
>
> As a KE, I have math background and some software tools for
> measuring the value of predictions given. And I've just gotten
> to a point where I want to start measuring the value of my own
> results against the value of predictions made by others.
>
>> As far as I remember he did not rank them, just gave categories and
>> sets of shares that fitted them.
>
>
> Meh. That's a ranking, of sorts. He sorts them into categories;
> ranks can be calculated from there, although it will mean assigning
> the same "rank" to every item in any given category. The numeric
> value of the rank is the number of things ranked lower, plus half
> the number given the same rank.
>
>> I posted this note weeks ago. Why have you only responded now?
>
>
> 'Cause in a few days the 30-day window will be up and we can measure
> the a priori predictions against post facto results. :-)
>
>> If you can get some sort of performance criterion out of his daily
>> recommendations then by all means have a go, but the OP is probably
>> impervious to any comment as he ignored my fairly provocative one. >
>
> Hell, I don't care what the OP thinks. I just think it'd be fun
> to apply the math and see whether this is real, random, or (as I
> suspect) shilling in an attempt to to wilfully mislead investors. >
> Ray
>
Fine. What does your analysis show then? I would expect the results to be random with chance fluctuations.