Science Advancements on trading

[fickletrader]

Quote from jerryz:

yesterday was a fat tail event. someone out there probably has a model that predicted it.

now we all know that there's only 1 reason why yesterday was a fat tail event. bernanke said something. so did the model predict accurately or did it get lucky?

years from now when someone is building another volatility model, he will process yesterday's data and find ways to predict yesterday's fat tail. then he'll find something and say, "look, my model predicted it." now there's only 1 reason why yesterday was a fat tail. bernanke said something. so can the model predict accurately or did the person inadvertently do a curve fit?

You miss the point. Just because Bernanke said something doesn't mean the nasdaq is going to move 2% in a few hours. People developing the kinds of quant models you're thinking of are simply trying to identify conditions where the market is vulnerable for a fat tail event. The idea is that if we can identify market conditions that have resulted in a fat tail event X% of the time, we can use a money management strategy to get us in when the vulnerable conditions arise, keep us in when a fat tail event happens, and get us out when it doesn't. After factoring in a lot of uncertainty, we can then estimate how many times we would have to trade the model before we could expect it to be profitable, etc.

The curve fitting problem happens when the market conditions you identify for entry and exit are overly specific and not robust. Curve fitting is a very different problem than predicting when Bernanke will say something.

 

[jerryz]

out of curiosity i did a quick and dirty test of yesterday's range compared to the range of 2006. i said huh? yesterday's move was normal? here's the data for the s&p500 index.

1/1/01 to 7/20/06
average range: 12.0
standard deviation: 5.2

so 2 standard deviations higher would be a range of 22.5.

yesterday's range was 25.07, more than 2 standard deviations from the mean.

Quote from nitro:

It was? I have seen several DOW +/- 200 point days in the last three months, and a slew of +/- 100 point days.

Markets are changing...

nitro

 

[dan05]

Hi Tyren,

They are currently offering a subscription for $650 per year through a propietary web site, www.tradingpro.com.

I'm also beta testing as a subscriber a new add-on for eSignal that looks cool so far as you can chart TA indicators on top of the prediction.

I'm not sure for how long they will offer the subscriptions, as I believe it will be for a limited number of traders.

Regards,

Dan

Quote from Tyren:

How much does their service cost ? (intraday forecast)

 

[nitro]

Quote from jerryz:

out of curiosity i did a quick and dirty test of yesterday's range compared to the range of 2006. i said huh? yesterday's move was normal? here's the data for the s&p500 index.

1/1/01 to 7/20/06
average range: 12.0
standard deviation: 5.2

so 2 standard deviations higher would be a range of 22.5.

yesterday's range was 25.07, more than 2 standard deviations from the mean.

I am not comparing yest range or even yest range to the mean, but to the tails. Read what I posted carefully. Most model the means differently from the tails...

nitro

 

[murray t turtle]

Partial Quote from Neodude:

LCTM failed mainly due ....... they over-leveraged and lastly they kept adding to losing positions.

According to what I've heard, if they weren't over-leveraged they could have salvaged much of their portfolio and actually made a profit.

-Neo

================
I understood , like you Neo ,as LTCM used insane leverage, & ignoring the Paul Tudor Jones principal ''losers average losers''.

Correct me if i am wrong Nazzdack; i took what you said as mostly a fun joke including 100% correlated to S& P 500
==========================================

Interesting T . Bass book reports on AMZN;
those Santa Fe scientists traders used quote '' trend glue' unquote '

 

[nazzdack]

Murray...........you got it right the first time.

 

[murray t turtle]

Quote from nazzdack:

Murray...........you got it right the first time.

=======================



Also did a bit of study on thier website of Predictors LLC;
catchy name,Predictors LLC .

However they werre advertising for a math PHD, with ''2 years estimation theory''.

ESTIMATION THEORY LLC, ''TREND GLUE'' probably more accurate,
Predictors LLC name does sound much more marketable

 

[fletch2]

Quote from nazzdack:

In your heart of hearts, you just know that if these geniuses formed a hedge fund, it would either be the next "LTCM" or correlate 100% to the S&P-500.

Right, like Renaissance Tech. All scientist chumps who can't trade their way out of a paper bag.

Fletch

 

[alesanti]

Quote from nononsense:

Science Advancements on trading

Could somebody tell me what is a "scientist" as related to market-trading? Must be some hybrid between our Jack and a Quant I guess.

More seriously: Does anybody around here know what is meant by "science", "scientist" and "scientific knowledge"?

nononsense, Ph.D.

A scientist related to market trading is essentially the same thing that a scientist related to any field of knowledge.
The scientist has to have an open minded and critical attitude and to use the artillery that more suitably fits the problem.
There exist a lot of possible candidates, say, fractals, nonlinear dynamics, complexity, catastrophe theory, self similarity, scaling, power laws, wavelets, filters, self organized criticality, self organized maps, radial basis functions, neural networks, MSS series, and the list continuous almost endlessly.
What has the market of singular against any other field of study? Well, maybe that experiments are hardly repeatable, and lab conditions are far from controlled. But fortunately you still have statistics at hand, and you can build something like "control groups' using random number generators or operations as simple as shuffling to study *wanna be markets*.

I will no try to answer the second question by now, because it belongs a bit to another realm.

 

[nononsense]

Quote from alesanti:

A scientist related to market trading is essentially the same thing that a scientist related to any field of knowledge.
The scientist has to have an open minded and critical attitude and to use the artillery that more suitably fits the problem.
There exist a lot of possible candidates, say, fractals, nonlinear dynamics, complexity, catastrophe theory, self similarity, scaling, power laws, wavelets, filters, self organized criticality, self organized maps, radial basis functions, neural networks, MSS series, and the list continuous almost endlessly.
What has the market of singular against any other field of study? Well, maybe that experiments are hardly repeatable, and lab conditions are far from controlled. But fortunately you still have statistics at hand, and you can build something like "control groups' using random number generators or operations as simple as shuffling to study *wanna be markets*.

I will no try to answer the second question by now, because it belongs a bit to another realm.

In the above you are only expressing beliefs - very popular thing among loser-posters.

Better tackle that second question.