Nice "new" advanced statistics option

Palsun Anand: C2 Rating: 4



There is no point arguing with this individual.



The rest of the world is wrong, he is right. Sharpe is wrong, he is right. C2 is wrong, he is right. Etc, ad infinitum.

Thanks, Eu. With respect to the Calmar ratio, I’m not sure why you say “sorry”. If you disagree with what I wrote, perhaps I didn’t express myself clearly. I mainly wanted to admit some confusion. Wikipedia says



"Calmar or MAR ratio’s of 1 are very rare in real world trading for an extended period of time. "



I don’t know what their source is, and I don’t believe that Wikipedia is always right, but I see a lot of C2 systems with a value that is much larger. So either Wikipedia is wrong, or there are many exceptionally good systems at C2, or we must expect that many systems with a Calmar > 1 will not be able to maintain that for an extended period of time (i.e. if you see such a system then Calmar ratio estimate is probably unreliable).

i.e. if you see such a system then Calmar ratio estimate is probably unreliable

In other words it is untradable in general. I apologized, because Calmar ratio is misleading on simulation data (from my point of view of course), but is very helpful on real data. You use as a base of calculation forward simulation data, but in any case of simulation the ratio should be adjusted or disregarded (just my 2c) So it’s something that cannot be fixed in theory until you have real data :frowning:

Eu



P.S. And I have no idea what is Calmar ratio is for Tango in the new stats, so I don’t promote the system. Still reading the guide :wink:

The section has the header “Theory”, so if you say that it is theory, then I agree. My objective was to give the reader somewhat more than the usual one-sentence explanation, by writing a (hopefully) understandable text that accurately reflects the theory of the original author. You understand what I wrote, and it agrees with what Sharpe wrote, so I conclude that my attempt was succesful. That you disagree with Sharpe is another point, but I don’t feel obliged to defend him.

You accuse MK for being non -objective, i.e, for being subjective. But you yourself is not being very different. You regard knowledge as a series of thunderbolts emanating from within human consciousness, your own or of others. This approach can’t identify the cognitive necessity of integration, of reduction, of proof. You do not feel the need of an "art of noncontradictory identification."



You took the field over, ignoring logic as a means of objectivity, reinterpreting its nature in accordance with their own premises. You stress sensations or percepts and detach logic from the world. To you, logic is the means of achieving consistency among arbitrary semantic conventions. The result is modern empiricism with its linguistic castles in the air, and its conclusions that sensory data can perhaps be described, but not understood. In this approach, logic is useless as a device of cognition, if “cognition” means a grasp of the facts of this world.



When men are deprived of their method of cognition, they have no means of validating their conclusions, no way to distinguish truth from error, fact from wish, reality from fantasy. The consequence is frustration and failure, the failure of their conclusions (including moral conclusions) to serve as reliable guides to action. This is the cause that explains the popularity of the notion that an idea may be "good in theory, but not work in practice."



This notion is impossible to an non-subjectivist, i.e, an objectivist. A theory is an identification of the facts of reality and/or guide lines for human action. A good theory is a true theory, one that recognizes all the relevant facts, including the facts of human nature, and integrates them into a noncontradictory whole. Such a theory has to work in practice. If a man’s course of action, thanks to his scrupulous use of logic, derives from a study of reality, then that course must be in harmony with reality. If so, what would prevent it from succeeding?



The theory-practice dichotomy is itself a theory; its source is a breach between concepts and percepts. Given such a breach, thought comes to be viewed as pertaining to one world of linguistic constructs, while action is viewed as pertaining to an opposite world (the world of concretes, or of things-in-themselves, or of empirical data). In this set-up, one expects an idea to be schizphrenic. One expects it to be good in one world, but not in the other, good in theory, but not in practice.



The consequence is to offer manking a monstrous choice. Practice theories that are impracticable, these theorists declare, or dismiss theory as a superluity and even a threat. This means remain loyal to concepts that clash with reality - or remain loyal to percepts by dispensing with concepts. The first is the intrincist choice; the second is the choice of the subjectivist.



No one departs from reality on the perceptual level; one can do so only on the volitional, conceptual level. In a primitive society (and in regard to a specific problem at any time), this departure can occur by default or simple error, through men’s ignorance of the proper methodology. In an advanced civilization, however, the only way such departure can be made to occur wholesale, with results disastrous for every problem and every branch of learning, is by means of a theory - a theory that subverts the conceptual level wholesale by detaching it from percepts. This is a disaster that only philosophers can create - or repair.

Jules,

How frequently are the daily advanced statistics updated? In some cases there’s a difference between N (daily values) and the length of a system’s history. E.g. extreme-os has been around for 709 days, but N is 677. Does that mean the advanced statistics do not include the most recent 32 days?

Science trader,

The 709 days is the number of calendar days, and 677 is the number of days for which there is an account value passed to the program. This is different because the markets aren’t open on all days. Actually I would expect the difference to be larger; I’m not sure why that is.



I don’t know the update frequency. It is not influenced by the part that I wrote. The program computes statistics from a list of account values, but it does not determine how often it is called. My impression is that the statistics are updated daily.

Yeah, it can’t be trading days as that number would be less than 500 for extreme-os. That’s why I assumed it would represent calendar days. I’m now a little confused. If 677 account values were passed to the program, I wonder about the values for those 177 non-trading days, perhaps zero returns? Wouldn’t that affect some other stats as well?

It depends on how you define a “trading day”. Some markets are open on a part of Sunday. So this can be viewed as a trading day, even though this particular system chooses to never trade on that day. This has the advantage that all systems have the same trading days. Doing that differently would produce a problem if a vendor of a stock system changes his mind and starts to trade forex, or conversely.



I’m not saying that this is the very best solution. I’m merely saying that this is to some extent a reasonable solution. In fact, it is the last of the aforementioned kinks that has to be ironed out. I’m not sure that any other solution is better than this one though.



Some returns will be 0, and this will bring the mean closer to 0, and the standard deviation will be smaller.