Algorithmic Trading

I'm preparing to launch a monstrous trading algorithm all over the Bitcoin market and steal everyone's money. Does anybody have experience with AI trading?

Other urls found in this thread:

quora.com/How-do-trading-algorithms-work
forextradersdaily.com/algorithmic-trading-professionals-reaping-record-profits/
traderhq.com/introduction-algorithmic-trading/
blog.quantopian.com/common-types-of-trading-algorithms/
youtube.com/watch?v=-fny99f8amM
twitter.com/AnonBabble

if (price < x)
buy;
if (price > x + y)
sell;

I'm using something a little more complicated, like neural networks and regression. What methods did you use to train your parameters?

if (price < x)
buy;
if (price > x + y)
sell;
if (price == x + y)
hodl

I wanna start. Got any resources?

no

>neural networks and regression
>Bitcoin
Hope you've got fire insurance.
Unless you think you can run the algorithm from a remote server, which might just be possible in other contexts but is simply foolish when applied to predicting the Bitcoin market.

Stanford Machine Learning on Coursera. Do the assignments

I'm going to use either AWS or my personal GPU. I'll try to keep a fire extinguisher handy

Thanks but how does machine learning play into it? I thought you found the algorithm by trial and testing but by suggesting machine learning do you mean to imply that the code should modify itself?

Also what previous knowledge do I need?

Regression in machine learning is just statistics. All of the algos fit themselves automatically to the data. It does take trial and testing and probably requires innovation for success. Some people take years to get the concepts.

If you want to really understand it I would suggest Linear Algebra and Multivariable Calculus which you can learn on MIT OCW. Good luck :)

(At least in financial markets)

The most important things to learn are:
- Machine Learning.
- String/Lexical Analysis (reading news articles automatically for trading info)
- Mathematical Statistics
- Quantum Mechanics and Relativity (Really useful in terms of its thinking, using unknown inputs to gain definite outputs)
- Multivariable calculus
- Differential Equations
- Linear Algebra + Optimization/Operations Research.

Woah thanks. I'm doing Engineering so I'll be covering those anyway

I do machine learning in my dayjob, and I have a finance degree, but I've never tried Algorithmic trading.

Is it trying to predict small pricespikes and make money there?
Or is it trying to predict the drift direction and make money that way.
Or does it use some wacky option arbitrage?
Or does it use triangular arbitrage?
Something else?

Any good resources on trading algorithms?

Youre an idiot troll who doesnt know how volatile crypto is. Enjoy your net loss.

Obviously I'm in the business so I can't tell you all the secrets, but the basic idea is that you use various established ml techniques to project a confidence interval of the price as far as necessary into the future and arbitrate the difference.

Thanks bro have a good day

>misplaced semicolon
>lose all of your money overnight

>arbitrate
why the fuck do you niggers use such awful vocabulary to obscure meaning and then claim that "you gotta keep secrets". you can talk about your shit in plain sight and nobody will understand

>using semicolons in Python

Some people understand more than others. Statisticians know what confidence intervals are and anyone who does machine learning knows what ml is.

what single word would you use instead of arbitrate?

user, what language(s) are you using to implement your strategy? Also, what exchanges are you going to run this on? I'm assuming you are looking to exploit arbitrage between two markets--correct me if I'm wrong.

I've been trying to learn more about this too. Have had bad experiences whenever I try to quality post on biz I get shit on. But this is sci so hopefully will go better here.
So like I said Ive been trying to learn about this, and one of the first questions I had was what is algo trading and what are the different types of algo trading. Here's some stuff I've ran across while researching this question.

quora.com/How-do-trading-algorithms-work
>1. Trade execution algorithms
>2. Behavior exploitative algorithms
>3. Scalping algorithms
>4. Predictive algorithms
>4.1 Mean reversion
>4.2 Trend following algorithms
>4.3 Chart pattern recognition algorithms
>4.4 Fundamental analysis algorithms
>4.5 Portfolio rebalancing algorithms
>There are another whole class of sentiment analysis algorithms that I am just not educated enough about to make a guess. I also didn't cover news based algorithms. also didn't include trading algorithms based on completely unsupervised learning algorithms

forextradersdaily.com/algorithmic-trading-professionals-reaping-record-profits/
>There are many different types of algorithmic trading strategies that are being effectively employed throughout the world, including:
>Trading Ahead Of Index Fund Rebalancing
>Trend Following
>Pairs Trading
>Delta-Neutral
>Arbitrage
>Mean Reversion
>Scalping

traderhq.com/introduction-algorithmic-trading/
>Index Fund Rebalancing
>Arbitrage Opportunities
>Low-Latency
>Market Making

blog.quantopian.com/common-types-of-trading-algorithms/
>short overview of common types of quantitative finance algorithms
>Momentum
>Mean Reversion
>Valuation
>Seasonality
>Sentiment
>Fundamental
>Technical

Veeky Forums is a shithole. It's just people shilling for pennystocks and memecoins. And they talk with such confidence despite knowing so little.

>- Quantum Mechanics and Relativity (Really useful in terms of its thinking, using unknown inputs to gain definite outputs)
any example of this?
>- String/Lexical Analysis (reading news articles automatically for trading info)

isnt this NLP

CS major watched a youtube video about neural nets, and now he thinks he can "steal everyone's money"

Good luck with that bud, naive time series prediction doesn't work with neural nets.

Veeky Forums had some decent threads when it was first made but ever since the shitcoin boom 95% of the board is shilling (in the most literal sense of the word) for worthless pump and dump coin #7000

AI trading has already been used for a few decades. Watch youtube.com/watch?v=-fny99f8amM

>using Python at all
Please stop.

Sure it does. If you can use classifiers for time series you can use neural nets. Pick your classes as buy and sell for example.

I use Python for its power and speed. I may use it on all the major markets but I have Bitstamp in mind.

I mainly plan to use arbitrage on the future value of the same exchange but arbitrating other exchanges may be a good idea.

what is the difference between machine learning and artifical intelligence?

AI is a lot more general, it's a research on how to make machines behave "intelligent", defining this term "intelligent" more closely for different tasks will lead to different goals and approach using different tools. Machine learning is one branch of AI when the goal is "learning", it is almost the same as statistics in terms of both the subgoals and techniques used.

The differences of ML and Statistics are more subtle: A doctor see the data of patients, comes up with the hypothesis, the statistician views samples of patients and check the validity of that hypothesis (i.e. hypothesis testing). ML find some useful patterns from the patients data (that may have been missed by other two). So now ML is used either to augment diagnosis, or automatize some difficult and tedious process. (after AlphaGo they are aiming for diagnosis done completely by computer lol). So the main difference with statistics is heavier dependence of ML on algorithms (complexity, storage, efficiency).
On the theoretical side, statistics is often about asymptotic behavior (like the convergence of sample based statistical estimates as the sample sizes grow to infinity), ML only focuses on finite sample bounds (given the size of available samples, ML aims to figure out the degree of accuracy that a learner can expect on the basis of such samples).
Another difference, in statistics it is common to work under the assumption of certain well developed data models (e.g. assuming the normality of data generating distributions, or the linearity of functional dependencies), in ML the emphasis is on working under a distribution free setting: the learner assumes as little as possible about the nature of the data distribution and allows the learning algorithm to figure out which models best approximate the data generating process.

I have no comment on the danger of AI or ML.

I know what a confidence interval is. I've taken basic statistics. I just don't know what "arbitrate the difference" implies. Are you doing your best to make future trades based off of a reasonable amount of risk that your target stock will behave favorably, a risk that you calculated from previous behavior? I apologize for the meme shitpost earlier, I just don't know what "arbitrate the difference" is supposed to mean and looking at a dictionary didn't help.

Think of the stock price an hour from now and compare it to the stock price right now. They have different prices. You can think of arbitration in terms of making a profit on the difference of those two values. Similar to how you profit on the difference in inter-exchange price. No worries.

>I use python for the power and speed

OP I'm gonna do you a big favor right now, and I'm not being mean.

You have no fucking clue what you are doing and will throw away all your money.

You're a fucking retard who doesn't know shit about equities. The more volatility the more profit.

You really couldn't get any more retarded you know it all fuckup.

Thanks for your opinion, but I disagree and I will prove my results without risk with backtesting before marketing my system. I think I will be successful.

Python is fast to program with and has fast C libraries like NumPy and CUDA modules.

Am not him but there is also an extreme to having to much volatility, more volatility means more emotional trading from the people involved, means less structured moves more chaos.

If you actually had a trading algorithm that is anything more than the usual a+b=c crap, I highly doubt you would be stupid enough to make this information known or present the existence of such an algorithm. That's your edge faggot, guard it with your life. You write some good code, spend a few years deleting and rewriting the same shit, becomes an obsession, but one day it will work. That code will show you the real meaning of freedom

Don't go after the bitcoin market though, it will be a huge bottleneck to your success you'll experience many problems not with your system but with what you're trying to trade over a more matured asset class with a better infrastructure for algorithmic trading (cTrader, MT5, Broker Api)

So arbitrage basically means "imma buy low and sell high nigguhhh here's the risk analysis that shows that this will have a 60% chance of being worthwhile"

Arbitrage is simultaneously buying something for $1.00 somewhere, while selling it at $1.01 somewhere else. For an almost risk free profit.

Basically, but depending on your data and algorithms your confidence could vary and if the price differences are high enough they could move the confidence to 99% or higher if the difference is so high it's outside of the confidence interval.

What's wrong with Python?

OP Look into this thing called IID, and then reconsider what you are doing

>protip: stock prices are not IID, yet all of the statistical methods you are using rely on IID data most likely.

This so much

Do you have any equities experience at all? Emotional trading is the most easily predictable... more structure comes from fear and optimism... if what you think you see is chaos, you just haven't been around it long enough.

Oh, I got you. Yeah, OLS has the assumption that data is your standard 'football' shape. There are techniques such as GLS that attempt to change the data to be homoskedastic but they are complicated. All that matters is the quality of the predictions you make. There are some really successful algorithms out there already.

>I'm preparing to launch a monstrous trading algorithm all over the Bitcoin market and steal everyone's money. Does anybody have experience with AI trading?

SO you are about to do something you have no experience or fucking clue about?

That's the best kind of thing to do.

But no, I actually have education in machine learning.

>cTrader, Broker Api
Too bad due to the regulators these are blocked in the land of the free and it's a crime for Americans to try to get around it.

I think this is overcomplicating an easily solved problem.
I would advise against trading forex, or any currency trading in general.
The most efficient way of making money that I've found is just trading penny stocks through the free think or swim platform.
Rather than algorithmic trading just do it by hand every day around 3pm when the markets are about to close.
I've been testing out a trading system for about a week now that gives me at least 6% per trade per day.
Basically the rules are simple...
1.Make sure that your trade meets every criteria listed below.
2.Only trade Nasdaq stocks, they are more regulated than OTC.
3.Find stocks that are priced between 0.50 and 2 dollars.
4.Narrow it down even further to stocks that have increased in price since yesterday.
5. Use the ADX indicator, look for a value over 25 that is increasing over time.
6. Find stocks that have a volumeOSC value that has just started rising above zero.
7.Finally use the stochastic oscillator to determine if the price is overbought or oversold.
Buy the stock when it's oversold under 20 and rising...

My strategy makes a lot of assumptions about how the market works, but so far it's made around 6% per trade if executed properly.

Just a follow up to this.
Technical analysis has gotten a lot of flack for being psuedoscience, and it is, but there are trends in the market that are consistent.
The stochastic oscillator assumes that the price for any stock is normally distributed.
This means that once a stock is at its low it will go up only if bad news about the company hasn't scared off investors.
It is very important to use common sense and research these companies before you invest in them.

Thanks for the advice man, I appreciate it. It sounds like you use a lot of data metrics to determine which companies to invest in. Despite what you said, that actually sounds like the perfect application for machine learning. Maybe I can develop a machine learning system that can superpower the returns you get from your strategy. It would just take the data metrics and run it through a price prediction algorithm or a neural network. All you would have to do is give each stock a confidence score of how well you think it might do. Maybe we can work together. If you're interested, email me at [email protected]

Lol, are you for real?

Yeah totally, hit me up. I've already created prototypes of time series forecasting algorithms. And bitcoin might have lower margins than penny stocks. I've looked into penny stocks before and they're excellent for returns. I completely forgot about them.