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

[Jerry030]

Quote from highlifejoker:

ok, whatever you think.

how about the billion dollar fund manager who had a fight with his wife the night before and decides to liquidate just to burn her ass?

what about seth tobias types? think you can anticipate their moves?

emotion still plays big part, but its now individuals emotion/ideas who control the large capital pools, not the masses of participants----if it ever was the masses. the very premise of TA is destroyed yet the TAers keep coming back for more punishment.

can't predict it, dont try.


HLJ


ps. by the way seth tobias movie/documentary on tonight CNBC.

enjoy!

Who can't predict? .... you? That doesn't mean those who trade profitability with automated predictive models, myself among them, can't do so. Keep in mind that it is a logically fallacy to prove a negative premise based on a limited set of observations.

My comments were partially satire on highlifejoke’s rather silly statement that computerized trading eliminates the ability to predict future market moves.

In actuality in predictive models you can detect a boundary in the patterns that have predictive power in about 1995 and 1996. I suspect this is around when automated trading became a significant driver of market moves. This complicates model training in that what is learned on historical data prior to that point has more erratic results when applied in post boundary zone testing or current live trading. For those engaged in predictive modeling I'll share my solution.

1) Model and test OOS on data prior to the mind 90's. Save predictive signal sets with decent performance.

2) Model and test OOS on data after the mind 90's. Save predictive signal sets with decent performance.

3) Compare 1) and 2) above and tag those that give the same performance in both time groups. Those are good as they haven't changed.

4) Combine 3) with 2) above and you have what will work in the market in a computerized trading environment.

Jerry030

 

[ProfLogic]

Quote from highlifejoker:

ok, whatever you think.
how about the billion dollar fund manager who had a fight with his wife the night before and decides to liquidate just to burn her ass?
what about seth tobias types? think you can anticipate their moves?
emotion still plays big part, but its now individuals emotion/ideas who control the large capital pools, not the masses of participants----if it ever was the masses. the very premise of TA is destroyed yet the TAers keep coming back for more punishment.
can't predict it, dont try.
HLJ
ps. by the way seth tobias movie/documentary on tonight CNBC.
enjoy!

After 09/11/01, the talking heads were saying that the SEC was calling for an investigation into anyone SHORT the markets on 09/11, when in fact the major Indices had all made lower lows days before the attacks. The oscillation resistance tops had confirmed almost a month earlier on the faster fractal charts that we were heading farther down at that time.

The London bombings were telegraphed as a strong pullback (drop) setting up in the markets the night before it happened.

Over the last couple years there have even been a few "fat finger" trades that have been telegraphed in the Intraday charts before they were executed.

This is just to name a few but there are many many more examples.

Uncanny? Yes.
100% accurate. Absolutely not but the percentage of accuracy available to us that can set up our charts fractally to see those internal movements of price, see it as a real benefit.

Typical TA is limited at best as to its real-time usefullness and readability. Indicators lag and the way one needs to gain a high level of comfort with price action is to eliminate the need to rely on them (indicators) as decision triggers. Using indicators as confirmation tools, for me, is the only acceptable use for them.

One can't condemn all TA just because the TA that one is familiar with is inconsistent, that is a surfism. That is like saying:
Chiropractors are all worthless because the few I went to were worthless.
or
Fish all have a bad smell because most of the fish I have smelled had a bad odor.

Neither statement is logical, just like it is illogical stating at ALL TA is worthless and inconsistent.

 

[Jerry030]

Quote from fluttrader:

Jerry030,



What you define as trade accuracy ? only PF ?

Measures include:

1) Win Loss Ratio in points: win/(win + loss)
2) Win Loss Ratio in trades: win/(win + loss)
3) Profit Factor in points: win/loss
3) Profit Factor in trades: win/loss
4) Average of winning trades in points
5) Average of loosing trades in points
6) Ratio of average winning to average loosing trade
7) Max drawdown
8) Distribution of losses (random is better than clusters)
9) Ratio of MFE captured by trades to theoretical maximum MFE
10) Ratio of MAE captured by trades to theoretical maximum MAE

There are a few others, but these are the major ones.


Jerry030

 

[Kevin Schmit]

Quote from Jerry030:

you can detect a boundary in the patterns that have predictive power in about 1995 and 1996. I suspect this is around when automated trading

The change in state in 1995/96 was probably due to a collapse in transaction costs (mostly narrowing bid/ask spreads). Algorithmic trading really did not take off until early in this decade.

 

[Jerry030]

Quote from Kevin Schmit:

The change in state in 1995/96 was probably due to a collapse in transaction costs (mostly narrowing bid/ask spreads). Algorithmic trading really did not take off until early in this decade.

That could explain it as well.

 

[TraderSystem]

Quote from Jerry030:

TraderSystem,

You won't find any real life stuff on this thread. Sadly, it's mostly theory and dueling quotes. The debate could have ended long ago by simply posting a price history data set and taking the following steps:

1) Let the "Randomites" test it and certify that all sequences and relationships are random and have no causative relationship or predictive potential.

2) Let the "Predictors" take the certified random data set and apply their algos and models to find predictive elements and see what kind of Profit Factor can be obtained

3) Then the Ramdomites can offer proof that such finding are illusionary.

4) Then the Predictors can trade what they found in real time and post the signals along with their brokerage statements.

5) The real life results would then end the discussion one way or another.

Heck, this gives me an idea for a fresh thread on Elite Trader!

Jerry030

Thanks Jerry030, for the explanations. Please tell us when you start the new thread, it will be interesting.

 

[TraderSystem]

Quote from kut2k2:

Here's one:

http://elitetrader.com/vb/showthread.php?s=&postid=1827852#post1827852

My point is that trends can occur randomly or nonrandomly. Don't assume it has to be one or the other.

But the only trends you can rely on are the nonrandom ones. Once you figure out how to spot them, you're on your way to profits.

Thanks kut2k2, for the explanations. Also, the link you gave was an answer to my first comment, regarding the coin. It looks clearer now.

 

[highlifejoker]

<b>4) Then the Predictors can trade what they found in real time and post the signals along with their brokerage statements.</B>



may i suggest reading "the predictors" first??


HLJ

 

[highlifejoker]

<b>Quote from kut2k2:

Here's one:

http://elitetrader.com/vb/showthrea...852#post1827852

My point is that trends can occur randomly or nonrandomly. Don't assume it has to be one or the other.

But the only trends you can rely on are the nonrandom ones. Once you figure out how to spot them, you're on your way to profits. </b>


this makes no sense whatsoever. if you catch a trend, it doesn't matter if its random or non random. in addition, your entry into the alledged non random trend is random per definition so it's a mute point regarding the randomness or non randomness of a trend. no wonder the market is so RICH....


remember--seth tobias tonight CNBC 9PM

 

[Jerry030]

Quote from highlifejoker:

<b>4) Then the Predictors can trade what they found in real time and post the signals along with their brokerage statements.</B>



may i suggest reading "the predictors" first??


HLJ

Could you suggest what you mean to look for in reading the book?

I have it. So if I know what you comment is directed to it would help.