Why do I see "Trends" in Randomly Generated Data?

[marketsurfer]

Quote from pneuma:

I have mathematically defined what a trend is for me, and based on the algorithm I have statistically proved that the trend as I have defined will continue. Thus I developed a trading system.

Why can't you get it through your thick head that there is no prediction there – it’s just a numbers game. For actual trading I don’t even look at charts, I just run algorithms and generate statistically relevant trading signals.

Let me put it in terms you might understand. If you ask enough chicks one is bound to sleep with you. As you practice you can "pick" those chicks that are more likely to sleep with you. If you can objectively define the variables that frame your "pick" you can statistically understand possible future performance.

And before you suggest that “you know statisticians that don’t believe in TA”, I’ll let you in on a secret – I am a statistician. I don't think your friends understand TA, you obviously don’t.

pneuma

you have statistically proved that the trend you have defined will continue? using what formula and how? mind sharing your insight? or at the very least, elaborate a little on what everyone else has missed.

if you have found an edge that everyone else has missed, what can i say, its possible. however, if you are able to actually put the probabilities of success above the rest of the players ( and not be curve fitting or deluded) soon you will be the next sim@ns or at the least SAC . nice meeting you

surf

 

[Jossan]

Of course there are trendlines in randomly generated signals....

A completely random generated signal is called brownian motion and it is very hard to distinguish brownian walks with financial signals which carry information.

Applying technical analysis to these random walks and drawing conclusions of parameters also walks. You can easily optimize parameters for a random walk and generate returns higher than a buy-and-hold strategy. BUT if these signals would be real you would never be able to make a profit on average with random signals.

One way to estimate the randomness of financial time series is the Hurst Exponent. For a complete random sequence the Hurst exponent is 0.5. Higher values, for example 0.6 indicate that there is information in the time series which you can use for prediction.

 

[makloda]

Quote from marketsurfer:
if you have found an edge that everyone else has missed, what can i say, its possible. however, if you are able to actually put the probabilities of success above the rest of the players ( and not be curve fitting or deluded) soon you will be the next sim@ns or at the least SAC .

Or the next David Harding... whooops scrap that, he's a trend follower and according to Surf's philosophy shouldn't even exist

 

[marketsurfer]

Quote from makloda:

Or the next David Harding... whooops scrap that, he's a trend follower and according to Surf's philosophy shouldn't even exist

no they exist, just the ones who are succesful are randomly generated.

 

[kiwi_trader]

surfer's just like his can't trade writer buddy.

a noisy randomly generated twat.

 

[makloda]

Quote from marketsurfer:
no they exist, just the ones who are succesful are randomly generated.

Of course. And markets are efficient. LMAO

 

[Rahula]

Lets assume a random market assumption. Here are the three ways that I believe the market can be profitably traded:

1. Buy and hold. In this case if you buy and hold stocks then at the very least stocks will act as an inflation hedge + added benefit of avoiding paying taxes every year which allows greater compounding + profit from long term earnings growth.

2. Market neutral relative value arbitrage. This is very high frequency market making.

3. Undiscounted News/Earnings. I'll give 3 examples. (1) Remember when CME when down 100 points when government requlators threatened to seperate the exchange and clearing operations. (2) GRMN collapsed over 10% 2 days ago when CEO on the conference call said prices will fall and competition will increase. (3) Yesterday's philly fed number came in way below expectations and treasuries took off without ever looking back.

 

[QQQShort]

Quote from Rahula:

Lets assume a random market assumption. Here are the three ways that I believe the market can be profitably traded:

1. Buy and hold. In this case if you buy and hold stocks then at the very least stocks will act as an inflation hedge + added benefit of avoiding paying taxes every year which allows greater compounding + profit from long term earnings growth.

If buy and hold is a valid strategy, then the random market assumption is invalid.

 

[Rahula]

Quote from squeeze:

I generate and use random data quite a lot to provide a base-line when testing new analytics.

Superficially, i.e. by casually observing charts it is difficult to tell the difference between real and artificial random data, especially on a daily time frame.

Analysed in more detail they are not the same. If you can find the differences then you are well on your way to having a tradeable edge.

Markets are much closer to random than most traders would like to admit and much more deterministic than most economists believe.

The difference is that randomly generated data gives you "normal" return that follow fit neatly on a bell curve.

Whereas the market gives you returns that follow a power-law distribution.

 

[newbunch]

Quote from marketsurfer:

yes, this is fact--people WANT to believe in the face of overwhelming evidence to the contrary. if you flip a coin 10 times and you get 10 heads in a row, are you in a heads trend??

If I knew nothing else about the coin, I'd bet on heads. Maybe the coins has two heads and no tails. Maybe it is weighted somehow and more likely to land on heads than tails.

In worst case, I bet on heads and the odds are 50-50, so I break even in the long run.