I'd suggest that all these (mostly anecdotal) monster success stories are inconclusive, at best. To me, they only serve as yet more evidence of survivorship bias (not that any more evidence were needed)...
Can linear regression analysis really predict the future?
- Thread starter tradrejoe
- Start date
Ok - at what point does survivorship bias end and what is your level of confidence.
How did you arrive at it.
After how many profitable trades will you say skill must have been involved.
Or - that the markets are organized at times.
In your mind does front running orders count as an edge?
How about knowing where soros or the fed or fidelity will be buying because of experience or knowledge.
How about where the floor likes to sell in an ag pit.
I wonder how the heck people can ignore that there is information in the market place and the markets do not respond to that information immediately.
How did you arrive at it.
After how many profitable trades will you say skill must have been involved.
Or - that the markets are organized at times.
In your mind does front running orders count as an edge?
How about knowing where soros or the fed or fidelity will be buying because of experience or knowledge.
How about where the floor likes to sell in an ag pit.
I wonder how the heck people can ignore that there is information in the market place and the markets do not respond to that information immediately.
Exploration:
if the Markets are truely random, then no "edge" shall exist.
Not in the past (back-testing), not in the present (live trading), not in the future (forward simulation).
Any gain or loss is the sheer (by)product of pure luck - good or bad. Period.
Corollary:
all these numerous books describing how to get an "edge" on the Markets shall be burnt at once.
My take: some of these books definitely deserve such a tragic end. Not ALL of them though.
if the Markets are truely random, then no "edge" shall exist.
Not in the past (back-testing), not in the present (live trading), not in the future (forward simulation).
Any gain or loss is the sheer (by)product of pure luck - good or bad. Period.
Corollary:
all these numerous books describing how to get an "edge" on the Markets shall be burnt at once.
My take: some of these books definitely deserve such a tragic end. Not ALL of them though.
what makes you think edge cannot exist in a random walk environment? In fact, I'd prefer a more truly (gaussian) random walk environment to be frank. Markets are actually worse than random in the sense that most perceive (gaussian/IID).
Take 1000 IID normal distributions with a positive offset (markets as you know have pos drift) as your cumsum components. Do you believe it would be impossible to make some type of betting strategy, such that expectation would be in your favor?
Take 1000 IID normal distributions with a positive offset (markets as you know have pos drift) as your cumsum components. Do you believe it would be impossible to make some type of betting strategy, such that expectation would be in your favor?
My personal null hypothesis is that, where tall tales of Joe Retail who started with $10K and is now worth $200MM are concerned it's all survivorship bias. If you have some evidence to the contrary, I'd love to see it, but my view, simplistically, is that it's the case of proverbial monkeys and Shakespeare. This view is based on a personal "local" sampling of the data. Needless to say, I may be biased, but I have not seen anything to disabuse me of my views.
As to your second question, I don't know. If I knew, I'd be making a killing working for a Fund of Funds (FoF). One thing's clear: judging by their performance last year, lots of very intelligent FoF people, who thought they knew the answer to your question, actually had no clue. As an aside, this whole thing reminds me of the story of the pseudoprimes in number theory (sometimes called Chinese Pseudoprimes). To find the first counter-example to what seemed like a bona fide primality test, you had to compute 2^341 - 2. Obviously, people thought they figured it out, but all it meant was that they needed to keep looking.
Now, as to actual advantages that involve non-public or semi-public information or behavioral biases, that's a completely different story. These may be real effects, in fact. But a natural question I'd ask then is what these things have to do with TA and chart patterns etc?
Well, when hedge funds major banks and what have you decide to dump any and all asset they have in store, pushing down all indices for days - if not weeks on, where having an "edge" merely consisted in entering short at 10:00am (have a look at the charts in the Sept-Nov 2008 and Jan-Feb 2009 periods), would that constitute a valid counter example?
Thing is, there a many examples of this type (not THAT blatant however, admittedly).
I suppose the DJ went from 12000 to 6500 in the span of a few months in a random fashion?
Thing is, there a many examples of this type (not THAT blatant however, admittedly).
I suppose the DJ went from 12000 to 6500 in the span of a few months in a random fashion?
so you're saying "random walk" and "edge" (ie "pattern") are NOT mutually exclusive...?
Quote from cfd_trader:
so you're saying "random walk" and "edge" (ie "pattern") are NOT mutually exclusive...?
I am saying random walk (in the gaussian sense) and edge are not mutually exclusive. A positive offset is an edge. Forget about patterns, just think about probability, expectation, and central tendency with large numbers... what does that mean to you?
There were some earlier threads were maestro talked about arcsine law. Maybe some clues lurk there.
I am also saying, that unfortunately markets are not perfectly gaussian, making it (in some ways) much more difficult than my toy problem. Too many posters seem to throw around the notion of randomness without truly clarifying what it really means, and how there are different models to express and classify it.
http://paul.kedrosky.com/archives/2007/07/02/a_practical_les.html
ah Ok, sorry u're right, let's stick to the mathematical meaning of random distribution and "skewness", which is maybe what u're referring to as "positive offset"?
http://mvpprograms.com/help/mvpstats/distributions/SkewnessKurtosis
Now since we're on a trading forum (or so I was told anyway
) we're still to think in terms of how to take advantage of skewness in random distribution like that of a stock/futures price...Via a statistical method maybe...?