Re: [CPT-FGC] Re: Hi

  • From: Ilitirit Sama <ilitirit@xxxxxxxxx>
  • To: cpt-fgc@xxxxxxxxxxxxx
  • Date: Mon, 3 Feb 2014 17:05:00 +0200

http://www.cs.utexas.edu/~pstone/Papers/bib2html/b2hd-GECCO06-trading.html

Forgot to mention this interesting bit of research:
http://www.nature.com/srep/2013/130425/srep01684/full/srep01684.html

tl;dr:  The higher the number of searches for specific financial terms via
Google Trends, the greater the prediction of a stock market crash.



On Mon, Feb 3, 2014 at 3:52 PM, G B <sigma.g19@xxxxxxxxx> wrote:

> Never heard of Peter Stone. Can you send me the title of the article or
> the pdf? I'd love to read it.
>
> The reason why its frequently mentioned/used within the Hedge Fund
> industry is due to how rapidly HFs change, as can be seen in the quant
> meltdown of 2007 (In three days, you could've made R1 turn into R500, if
> you were at the right place at the right time). Testing for adaption here
> is extremely interesting, and saves you years of work. HFs is also the
> canary in the coal mine in world of finance.
>
> You are spot on with the over-fitting issue. :-) I corrected this when I
> implemented mine. When using fundamental analysis, they'd implement like,
> 15 different fundamental metrics, of which maybe 4 or 5 have any statical
> significance. In fact, certain researchers went to the extent of running a
> GA to select the variables for "best-fit" first... which is akin to
> data-mining in the sense that they are finding the best variables that fit
> with the given data BEFORE running the algorithm.
>
> And yes, the evolutionary algorithm is trained using a data-base first,
> before it goes live. Typically, the GA selects the variables and the ANN is
> what does all the decision making. Having the computer decide what it
> should do for both is unwise, or at least, I think so.
>
> Like I said, you put garbage in, you going to get garbage out, that is how
> it works with these black box machines, so yeah, people have used GA's
> before, but like we both mentioned, how you program it and I would is
> totally different, which means it would learn differently, just as we do.
>
> You also mentioned "Buy-Low/Sell-High" when there are economic shocks,
> which is a great strategy, sure, but stock prices will adjust accordingly
> anyway, so while they may be cheaper than usual, they won't necessarily be
> mispriced, which is what I'm interested in. I can't always wait for some
> kind of external economic movement to make money.
>
>   As you've said, I don't want to match the market at all, else I'd go
> passive. Stocks are priced at the value that the market values it at and
> those are often subjective prices, more often than not.If I had a basic
> program to initiate buy/sell commands based on whether a stock falls
> below/above a certain price and expect it to earn money, I'd be a
> billionaire by now.
>
>
> On Mon, Feb 3, 2014 at 2:29 PM, Ilitirit Sama <ilitirit@xxxxxxxxx> wrote:
>
>> @GB
>>
>> I did some reading up on using GA against the stock market.  It seems
>> some Hedge Funds use them already, but most of these funds are already
>> defunct (probably not because of the use of GAs though).  Was that paper
>> you mentioned by Peter Stone?  It answers a lot of the questions I had.  It
>> seems the evolution of the ANN happens way before the data is run on live
>> data.  Stock Market GAs also tend to suffer from "over-fitting", which is
>> something I also mentioned.  They try to adjust to past performance in
>> order to generate an expected result.  However, most GAs don't take into
>> account random fluctuations in the result, so they fail miserably when they
>> try to derive patterns from the data.  The best they can do is to try to
>> match the curve, which of course is very bad.
>>
>>
>>
>> On Mon, Feb 3, 2014 at 2:19 PM, Ilitirit Sama <ilitirit@xxxxxxxxx> wrote:
>>
>>> He needs to confirm the subscription.  The link should be at the bottom
>>> of any of the admins' messages.  Give me his email address.
>>>
>>>
>>> On Mon, Feb 3, 2014 at 2:03 PM, lindsey kiviets 
>>> <lindseyak@xxxxxxxxxxx>wrote:
>>>
>>>> lol,
>>>>
>>>> @ilitrit , gimme th fgc link again please.i added Hilton myself , and
>>>> it seems he is not added to the mails.
>>>>
>>>> ------------------------------
>>>> Date: Mon, 3 Feb 2014 13:59:14 +0200
>>>>
>>>> Subject: Re: [CPT-FGC] Re: Hi
>>>> From: ryan820509@xxxxxxxxx
>>>> To: cpt-fgc@xxxxxxxxxxxxx
>>>>
>>>>
>>>> *lol* Guys will just have to put their cellphones where I can see em,
>>>> or else :P Scumbaggery at it's finest :P
>>>>
>>>> I'm busy editing the vids. Will start uploading soon :)
>>>>
>>>>
>>>> On Mon, Feb 3, 2014 at 1:55 PM, Ashraf Barendse <
>>>> ashraf.barendse@xxxxxxxxx> wrote:
>>>>
>>>> The first rule of Random FG club is...
>>>>
>>>> Gohan is a charge character. I know this because LB knows this.
>>>>
>>>> =========================================================================
>>>> You are subscribed to the Cape Town Fighting Game Community mailing
>>>> list.
>>>>
>>>> //www.freelists.org/list/cpt-fgc
>>>>
>>>>
>>>>
>>>
>>
>

Other related posts: