Who wants to create a partially automated trading system with me?

Who wants to create a partially automated trading system with me?

Other urls found in this thread:

github.com/sciencefreak500/StockPuller
veekyforums.com/thread/1802042/business/who-wants-to-create-a-partially-automated-trading.html
twitter.com/SFWRedditVideos

Could I make it with Python?

I'll pay you 5xmr for a good polo loan algorithm or machine learning or whatever works as close to perfect as possible.

I've already got a basic Python lending bot. Just needs to be smarter.

What language? Am interested

>Could I make it with Python?
yes
> polo loan algorithm
sounds interesting I would have to look that up
>What language? Am interested
I'm more familiar with doing statistics in R, but Python will work.
I use R and Python together. Some of it might have to be in Python tho.

I'm new to Python. But my plan is no automated trading system.
I see no value or money in it because you can't compete with the autists who studied it, are located as near as possible to the stock market and are financed by big money from investment banks. It is a waste of valueable time in my opinion. Choose your battles buddy.

I'm more interested in automation of tasks/decisions regarding risks/chances. What if my software can warn a company of a sudden decline in orders because one customer goes bankrupt? Or if a company is going to order more next week. If our logistics partner is going to increase prices due to higher costs for him. How much are companies willing to pay for my software that warns them? Small incidents could trigger something bigger.

Like a cockpit that signals to a pilot something is wrong. But my goal is to warn them before it happens so you can react accordingly.

I was actually thinking of something similar because that cost of being able to do automated trades is too high. I was thinking about creating a system that follows a particular stock and text messages you to execute the order, so you will lose some and win some, but in the long run, you will come out ahead.
>I'm more interested in automation of tasks/decisions regarding risks/chances

Maybe you could follow options and futures not to invest in them but as an early signal to real life risk.

My problem with the stock market is that it is unpredictable because of emotions.
You can run the numbers with Tesla, forecast their cash flows, play around with growth rates and still it is overvalued as fuck. So why are Tesla shares +50 Dollars since Dezember? How can you measure the hype of a stock? Do you think you could have predicted the downfall of Nokia at that time? How could have Nokia screwed up that big?

I'm sure it is possible to make money with an algorithm and the stock market, even for small guys like us.
But is it worth the time? I wanna focus on things I have way more control over, things I understand and can measure.

Like I said above, in my opinion it is a waste of valueable time to compete with people way more qualified than I am in a market I do not fully understand.

any long-term forecast would not work. this would only work for very short term.

Instead of spending time reading fake news and charts you would spend time finding irregularities in the system.

>I wanna focus on things I have way more control over, things I understand and can measure.

this comes with the sacrifice of having too much bias. if you are creating models for people to understand, there is going to be biased in the models built in. i'm more interested in just the prediction without interpretation.

>my opinion it is a waste of valueable time to compete with people way more qualified than I am in a market I do not fully understand.

I actually don't beleave this to be true. their tallent is in talking.

>any long-term forecast would not work. this would only work for very short term.
Not sure how you are going to invest short term when the price is depended on long-term DCFs?

>Instead of spending time reading fake news and charts you would spend time finding irregularities in the system.
Please explain. What irregularities?

>this comes with the sacrifice of having too much bias. if you are creating models for people to understand, there is going to be biased in the models built in. i'm more interested in just the prediction without interpretation.
Okay. But this is what I'm talking about, the price itself is not unbiased. There are factors already influencing the price you can't measure. Like Tesla I mentioned above. I would have lost a lot of money shorting it, all my calculations said it is massively overvalued.

>I actually don't beleave this to be true. their tallent is in talking.
I'm not only talking about the guys in suits. I'm also talking about the guys with a PhD in math trying to find the next algorithm. The guys renting out offices as near as possible to the stockmarket to be the first with new information. The guys investing big money in computers and glass fiber cables.

I am interested. I am a python dev. A few days ago I was wondering if I could write a script to make a graph similar to the "fastgraphs" ones. What I quickly discovered is that there are no good free sources of data. Would you want to pay for data? I think a lot of data could probably be parsed out of Yahoo Finance, but it does not seem to be exposed through a public API.

>Anonymous (ID: vd0RWMLv) 02/16/17(Thu)09:28:55 No.1802

>Not sure how you are going to invest short term when the price is depended on long-term DCFs?
I was thinking very short term trades, like trade greater than every 3 days, since it takes 3 days to clear.

>Please explain. What irregularities?
I actually think finance is rigged with fake news.
I came across something along the lines of anomaly detection. There are a lot of things to experiment with and implement.

>I'm also talking about the guys with a PhD in math trying to find the next algorithm.
If you can read and implement to literature you're good to go.

>The guys investing big money in computers and glass fiber cables.
We can't do what they do, but that might be a good thing. Our models would have to be smarter since we can't depend on speed. We would be out of their hair.

You have to stay low key as possible.

yeah, i'm having a data problem also. i think this is a way of reducing access on purpose. i think we could just scrap it firs. i heard yahoo had an api that had a 15 minute delay. i scraped it from google finance but i could only send a request every minute.

I have made a automated trading system with C# on NinjaTrade. AMA

Works best with a 5 min chart on CL.

What models do you use?

I use a series of confirming indicators that go long/short and exit. I manually traded CL for years before that so I know the market really well. I'm still trying to filter out some shitty trades though.

So these are pretty well-established indicators?

I did a hybrid time series model that seemed to perform pretty well in forecasting, but only works for near term.

You're conflating "algorithmic trading" with HFT. You won't be able to compete in HFT, but that doesn't mean you can't devise a system that operates in the extended short term (hours/days/weeks).

have you done any algo trading? i mainly don't dwell in finance that much but am interested in creating something, mostly for the sake of the challenge.

HFT is a another good example why it is so difficult to predict or calculate what a fair price would be. It shows the emotional side of the market.
Some HFTs 'test' the market by buying and selling shares just to watch how the market reacts and make their move accordingly. The market reacts to something based on 'trap'. You trade in hours/days/weeks. HFT trades in seconds and could therefor mess up your trade in seconds.

I wish you all the best if you would like to try it. I'm sure it is possible to make money.

is there a gauge to measure market sensitivity?

I've tested fairly basic strategies but I have neither the econ background nor the capital to begin trading on anything other than paper. I'm just a CompE student interested in some projects with data science and machine learning.

That's going to have an effect on everybody though, it's not exclusive to medium term algorithms. You're still at a little bit of a disadvantage since you would potentially be less diversified, but it's not huge. You just have to make sure you don't panic sell and decide how long you're going to hold the bag if a bet goes south (which are scenarios that need to be considered anyway). Unless you're holding an option about to expire, you have time to figure out what to do with that asset. The HFT effect won't permanently sink the stock.

Like I said I wish you all the best.
There are to many factors that need to be considered and some of them are out of my control. I just wanted to mention that there are other interesting and profitable areas, not only finance.

go on ...

The simplest script you can write is :

x = 50 (the price you bought the stock)
if current_price < x + 0.05:
SELL

How do you guys make orders, is it a manual set up or an api from your broker?

I wrote hundreds of scripts, indicators and expert advisors in mt4 and it's very easy way of implementing, testing and optimizing strategies if your broker uses metatrader as a client trading platform.
If you want to dive in check out jimdandy channel on youtube.

Ended up writing java gui app for stock forcasts that's using scrapped data with 15 delay.

Sure. Hit me up with more info

There's a hidden api look it up. You just construct the correct url. Look at the url for the csv files

Careful though, yahoo api doesn't work well with etfs, particularly with dividend data

I already did this for the $ES but without a comprehensive theory of price change you wouldn't understand why it works, and would start tinkering with it and fuck it up, instead of just watching it and marveling at the money stacking up.

It's also platform independent. All you need is:

The ability to render volume candle charts, not time; the ability to customize the volume count of each candle; the ability to draw a continuous 12-candle trailing Least Square Moving Average line; a derivative indicator of the sin of the LSMV; an anchorable simple moving average curve (MIDAS curve, because on a volume chart, all moving averages are volume weighted moving averages), and a simple Effective Volume strength indicator.

For trade volumes below 11 contracts at a time it cannot be detected, and cannot be beaten.

You can't actually get historical dividend data from it though... you it ends around 2014 IIRC.

What is "near term"? Hours; days; months?

Check my github, i pull the whole Market using multi-threading and Yahoo finance. It pulls daily chart data, span two years of time.

github.com/sciencefreak500/StockPuller

This is a start if you are planning on using node.js

npmjs.com/package/simpletradingapp

it would probably be hours/days
if i had the money i would do it by the min

I have one written in Julia for google finance, but i think google made it more difficult to harvest

>How do you guys make orders
I would probably have to do them manually because to qualify to make automated orders takes like $10,000.

That's fine for building your own data source starting today, but I'd kind of like a historical data source going back 10+ years....

If you have a somewhat usable product and a way to turn a profit getting 10k from an investor like me is nothing...There's probably a few rich people on this sub.

Try stooq.com the have nice csv files but limit number of connections per ip.

If you're only looking for older data so that you can add newer later then that may be a good way to go.

What do you mean? Other areas where you could earn money with your coding skills? The possibilities are endless.

>data mining/big data/BI/data science
>machine learning
>web development
>scripts (to automate something)

Look at the economies of scale in software. One simple program that solves a problem for a small niche market could make you a millionaire, maybe even a billionaire (extreme example). Small improvements in your software could expand your potential market, your costs to do it are low as fuck.

trusting a random guy on Veeky Forums?

Is you a robot?
veekyforums.com/thread/1802042/business/who-wants-to-create-a-partially-automated-trading.html

What the fuck.
Why would somebody do this? Ads?

>Why would somebody do this? Ads?

pretty much.

>veekyforums.com/thread/1802042/business/who-wants-to-create-a-partially-automated-trading.html

do i pass the Turing test?

Yeah, yahoo is better for this in my opinion, I also multi-threaded that fucker, with a thousand threads I was able to download the whole Market in a minute.

Info works nice for importing into excel

You can actually adjust the parameters for that. Yahoo finance create CSV files based off of conditional statements in their URL, you can just set the end date to be 10 years instead of two years like I have it.

what are good books for a guy to learn how to forecast in R?

automated trading is hard. To much info. People have literally spent billions developing systems that barely work. I suggest just learning to trade using price action. Works much better. You can write an indicator that sends you alerts or shows you what stocks fit your possible trade category. Example: low rsi with increase in volume may pull up stocks that are due for a breakout/in the process of breaking out.

I suspect this is bullshit and only those having good intuition manage to make money using standard indicators/techniques.

I tested hundreds of strategies, used every even slightly popular indicator and most common combinations of them and different entry/exit strategies with different money management scenarios.

Nothing like that makes consistent profits when put into code.

I have a shit load of books on R.
What are you going to apply it to?