Please Scroll Down to See Forums Below
Stock Market
- Thread starter MarthaStewart
- Start date
MarthaStewart
New member
I originally thought of the code as 2% over 21 days (a trading month). I have had that going for months and it is scarily good.
Frequently they will go up 2% within the first day and almost always in the first week.
I set it all up to be sonservative since it is designed to be an in and out everyday.
Ideally you would put money into several stocks. So it told you 5 stocks triggered, you would put money into all of them.
At some point you will run into issues with how much money you put in - for instance if you put in a million dollars, you are going to show up pretty significantly in the volume for that day.
That then kicks in the Heisenberg Uncertainty Principle - you are no longer in the observational realm and instead having a direct effect on future behavior of the stock.
But from what I have seen, $200K seems to be high enough for good returns but low enough that it won't break anything.
I could go on and on talking about this - it gets really complicated.
I have written some code that runs on a parallel cluster of computers, but unfortunately that is expensive to maintain, so I am currently running some "scaled back" code until I can get more money for this.
Frequently they will go up 2% within the first day and almost always in the first week.
I set it all up to be sonservative since it is designed to be an in and out everyday.
Ideally you would put money into several stocks. So it told you 5 stocks triggered, you would put money into all of them.
At some point you will run into issues with how much money you put in - for instance if you put in a million dollars, you are going to show up pretty significantly in the volume for that day.
That then kicks in the Heisenberg Uncertainty Principle - you are no longer in the observational realm and instead having a direct effect on future behavior of the stock.
But from what I have seen, $200K seems to be high enough for good returns but low enough that it won't break anything.
I could go on and on talking about this - it gets really complicated.
I have written some code that runs on a parallel cluster of computers, but unfortunately that is expensive to maintain, so I am currently running some "scaled back" code until I can get more money for this.
Last edited:
What factors do you measure or is this something you don't care to share? Fine either way.
What I'm really wondering is how do you account for policy making?
What's todays five. You can take this to investors and say...see dillholes, give me your money.
What I'm really wondering is how do you account for policy making?
What's todays five. You can take this to investors and say...see dillholes, give me your money.
MarthaStewart
New member
I have a few investors. Half a million dollars so far - but I am trying to do it all on my own so that I don't have to report to anyone (they are being really flaky right now anyway). If I find I really can't do it on my own, then I will have to ask for help from them.
Plus, most of them just want me to only let them use it... I don't blame them.
I can't really say what factors it measures - not because I want to keep it private, but because of the nature of neural nets and the genetic algorithms. I feed it the data and it does the rest - and it doesn't always do it the same each time.
Generally speaking though, it isn't intraday data and just time series historical data of end of day data.
It is completely oblivious to news - that is why it doesn't do as well during earnings time.
The code that runs on the parallel cluster system is much more powerful and can do more - that one has the ability to do better against news - but I can't run that right now - this currently is on a single system.
There is no "today's five" - it is just whatever triggers the algorithms for the day - I will look now and see what it triggered - but I don't know how many it will be.
I'll post them up in a few, need to login to the server.
Plus, most of them just want me to only let them use it... I don't blame them.
I can't really say what factors it measures - not because I want to keep it private, but because of the nature of neural nets and the genetic algorithms. I feed it the data and it does the rest - and it doesn't always do it the same each time.
Generally speaking though, it isn't intraday data and just time series historical data of end of day data.
It is completely oblivious to news - that is why it doesn't do as well during earnings time.
The code that runs on the parallel cluster system is much more powerful and can do more - that one has the ability to do better against news - but I can't run that right now - this currently is on a single system.
There is no "today's five" - it is just whatever triggers the algorithms for the day - I will look now and see what it triggered - but I don't know how many it will be.
I'll post them up in a few, need to login to the server.
MarthaStewart
New member
FatRat - I can set up a bot user that just posts the code to the chat board everyday if you want. I can do it until I get my company up and going.
If there is demand for it and George doesn't mind, then I will do it if I have time.
I don't want to manually have to do it though since I might be too busy.
Also, it doesn't loop at small or large cap (or rather, it looks at everything) - it just looks at what I give it - which is anything that has at least 1500 or so trading days of history.
It currently looks at about 1800 stocks over the 3 US markets.
I would like to get a data feed that has intraday data, options data, and also other markets - Reuters has a good one, but I can't afford it right now.
Here is a list of what it gave me for today - keep in mind that it is "safer" to only trade on the ones that have average volumes over 300K a day.
The ones less than that still usually pop - but they are spastic and your buy/sell is more likely to have a larger impact on its price.
Also, as we saw before, even though this code is trained to look for a rise of .5% - it looks like it is safer/better to go with .4%.
(I have gone in and adjusted the code so that it should later be better for .5%, but I haven't updated it yet - it has to run for some time and I have to get to work soon).
ALT
ARTNA
BER
BAT
CNBC
CLMF
ECL
HITK
IGT
IMH
INV
RNR
TLK
EWST
AP
BDY
BMT
So of course, if I had to say what I would reall look at, I would focus on:
ALT
BER
ECL
HITK
IGT
IMH
RNR
BDY
I would give RNR special focus over the next month.
They are an excellent company and the ass beating they took yesterday was pretty much unfounded.
If there is demand for it and George doesn't mind, then I will do it if I have time.
I don't want to manually have to do it though since I might be too busy.
Also, it doesn't loop at small or large cap (or rather, it looks at everything) - it just looks at what I give it - which is anything that has at least 1500 or so trading days of history.
It currently looks at about 1800 stocks over the 3 US markets.
I would like to get a data feed that has intraday data, options data, and also other markets - Reuters has a good one, but I can't afford it right now.
Here is a list of what it gave me for today - keep in mind that it is "safer" to only trade on the ones that have average volumes over 300K a day.
The ones less than that still usually pop - but they are spastic and your buy/sell is more likely to have a larger impact on its price.
Also, as we saw before, even though this code is trained to look for a rise of .5% - it looks like it is safer/better to go with .4%.
(I have gone in and adjusted the code so that it should later be better for .5%, but I haven't updated it yet - it has to run for some time and I have to get to work soon).
ALT
ARTNA
BER
BAT
CNBC
CLMF
ECL
HITK
IGT
IMH
INV
RNR
TLK
EWST
AP
BDY
BMT
So of course, if I had to say what I would reall look at, I would focus on:
ALT
BER
ECL
HITK
IGT
IMH
RNR
BDY
I would give RNR special focus over the next month.
They are an excellent company and the ass beating they took yesterday was pretty much unfounded.
MarthaStewart
New member
There are a few ways to get it to handle news - the hard part is getting access to a news feed for stock data that goes back for 10 years.
I do have a few ways that I can do it though if I had the news.
There are some other ways too that you can at least take into account earnings time.
the parallel code does way more stuff, so that is where the code that handles the news better resides - but I don't have those machines right now.
It also really helps out with pairs trading.
I do have a few ways that I can do it though if I had the news.
There are some other ways too that you can at least take into account earnings time.
the parallel code does way more stuff, so that is where the code that handles the news better resides - but I don't have those machines right now.
It also really helps out with pairs trading.
MarthaStewart
New member
I gotta catch the ferry into work - I'll be on again in about an hour.
Similar threads
