/biz/ - Business & Finance

Software engineering and stock trading have absolutely no correlation whatsoever, so I'm not sure what you're asking. If you want to learn how to trade stocks take a course or read up on the different strategies at investopedia.com

>>871631
>Software engineering and stock trading have absolutely no correlation whatsoever

Yes... which is why I made this post and am trying to learn.

>>871631

idiot, how do you think bots are made? Out of FUCKING MAGIC? IDIOT!

op just look at technical indicators, spend some time on the market watching how these indicators work. If you can predict movements then you can program your robot to do the same.

so to re-iterate.

read up on a shitload of different technical indicators. (Moving average, stochastics, Bollinger whatever).

See how these indicators work in real time.

?????????????

go back to sucking dicks for $1 a pop

>>871654
Bots are made by software engineers who know how to trade stocks you incredible faggot

Your programming and machine learning experience isn't really worth shit if you don't have a solid grasp of the fundamentals of the finance side of things.

You might know HOW to program an algo but you need to know WHY you're programming it to do whatever it is you want it to do and for that, you need to be better educated about things than just copying strategies from elsewhere. Start from the very basics and learn trading fundamentals as if you'd never even used a computer; the imporant ratios, trends, all that shit. Then you can tart thinking about strategies (even if you are copying them from elsewhere) becasue at least then you'll know what exactly they're trying to accomplish and why they theoretically work.

>>871660

MUH

>>871616

PCA, SVM, NN all are things you can look into. You should have gotten Chan's other book (Quantitative Trading) first.

>build own algos but statistics based
>yes, more than buy and hold

>>871660
Yeah no shit, he knows that. Did you not read his post? He knows the software aspect and now wants to learn about trading.

Also if you think they have "absolutely no correlation whatsoever" you either are completely unfamiliar with algorithmic trading, in which case you shouldn't be posting, or don't know what the word correlation means.

>>871616
Hey OP, look here:

www.reddit.com/r/algotrading
https://www.reddit.com/r/algotrading/comments/3det1n/algotrading_passive_income_worthy/

I think generally speaking you need some domain knowledge before you jump straight into the programming side, i.e. you need to learn how to trade before trading w/ algos. As a source of extra income the return on your invested time might actually be better on something like app development than algo trading.

That being said, I'm sort of in the same position you are, I have some experience in ML/data science and I'm looking to apply it to trading stocks. I mostly use R and Python; maybe we can work together?

>>872634
back to b scum.

>>871616
You need to know how to trade before making a program that trades, so don't dive for the most complicated stuff first. Just get an idea of beginner level investing.

>>871616

You are me. Five years ago.

I did it. I built an algo trading platform atop the IB API.

It works, but as you've imagined it all boils down to strategy.

I built a hundred strategies before I had six I could lean on.

I made nearly a half million in a month before those strategies stopped working and I had to get back to work.

You're thinking "I'm a programmer, so naturally the best way to learn to trade is building a trading platform."

In retrospect, had I spent three months paper trading I would have saved myself a year of R&D in the beginning.

It's like a chess game. Learn your order types. Know your pieces and how you can move them.

In my experience, it all boils down to "Trending" or "Mean Reversion" behaviors.

Refer to the white paper "MR Swing" written by Dave Abrams and Scott Walker.
Refer to the book "The Evaluation and Optimization of Trading Strategies" by Robert Prado.

Ask anything. I'll answer. Excluding strategy specifics.

>>873004
Thank you friend. Good to know this isn't just a pipe dream.

>>873004
Does IB give you real time market data? Complete noob. Same boat as OP.

>>871616

Do you even neural networks?

>>873044
Yes, and the of over fitting on financial data is very high.

I'm sure there are guys doing it, but you would have to really put in some work for your feature engineering.

>>873004
What was your starting capital and do you designate the same amount of trading money to each strategy or do you put more on 1 strategy because it is the most reliable or has the higher return or whatever?

>>873047

Hi. Thanks for your response.

I am a MSEE (recent grad) and my master's project was on Neural Networks. I am not a programmer to be honest, but I would like to learn. I used the neural network toolbox in MATLAB. I started investing a year ago and I want to write an algorithm too but don't know where to start.

gl.

>>873004
So what do you do now? Still go to work, trading or both?

>>873004
Why do you suppose the strategies stopped working?

>>873036
Yes, and Six Months of historical. Down to 1sec Resolution, and Tick-by-Tick for real time but be warned that they summarize those ticks so you'll want to consider tick-data supplementary and not primary.

>>873044
Tried it. Not suitable. Genetic Algorithms too.

>>873052
30k. Minimum for equity day trading w/ Reg-T Margin is 25k. ForEx has different rules. Look into that if you've got less funding. I built an algorithm to brute-force-optimize my war chest of strategies. In essence, they're all competing for placement in "tomorrow's" portfolio. Selection of the fittest.

>>873085
I'm still trading. Still cooking strategies. Still refining my platform. Still making money, but significantly less since my first gen strategies stopped paying out.

>>873091
End of July 2014. QE stopped. Uncertainty took over. Liquidity dried up over night. The market is moving with fewer, yet proportionally larger trades. The over-all trading volume stayed approximately the same, but at the micro-level it's a desert punctuated by violent moves. My latest generation of strategies is keeping up, but is less than 10% as profitable as the old ones in their prime. I'm working on that, and I've made great progress but I haven't properly cracked it yet.

That's how algo trading works. Fill niches until you can trade profitably every day. If you're very good, your platform/engine/AI/whatever is smart enough not to trade when it can't win.

>>873161
Thanks!
Do you have any tips/tricks, any TA that you recomend from avoiding?

Just anything that will save me time and money, obviously other than giving away your secrets?

>>873179

MACD doesn't work. That's a gimmie, but it's indicative of the problem with TA.

I refer to TA as "metrics" like a yard stick. It won't tell you where to cut, it just measures length. Use it as context for your strategies, not as the rule book.

Three bits of advice that'll save your dick:

1) There is no singular perfect strategy. Fill niches. A great strategy works for about 3 weeks to 9 months consistently. Usually followed by weeks or months of shit performance until it comes back "into regime." Refer to Markov Chains for an understanding of this, but don't use that math literally because it's convoluted and mostly useless in this context.

2) Read that book by Pardo. I spelled his name wrong in my earlier post, but for the love of fucking and all that is holy, read that book. If I had read that book during year 1 of R&D, I'd be a much richer man today.

3) Sometimes the only winning move is not to play. Your algorithms should know this.

>>873206
Are forex and shares the same shit, just different strategies when it comes to algos?

>>873206
Do you trade with your entire war chest, or do you limit yourself?

>>873297

Disclosure: I do not trade ForEx.

From a programming point of view, it's the same.
From a risk point of view, ForEx lets you get crazy with the margins and that scares me away from it.

It shakes differently, but sometimes the same strategies will work just fine. Not often, but sometimes.

Some strategies are simple tricks. Some are far, far more sophisticated.

>>873343
My "war chest" is my best few strategies.
I don't include every one I've ever coded because most of them are shit. You wouldn't put a broken gun in your armory, right?
Refer to this pic related to better understand the process of strategy dev and deployment. >>873004

>>873004
Thanks for the reading suggestions, Anon. Not OP here, but similar situation. n00b to algorithmic trading, but have a basic grasp of programming, and I have domain knowledge from a few years trading more or less successfully on my own. Have sufficient capital for a few years' runway. Have sufficient time to develop. ("So why am I hanging out on /biz?"... yeah, I ask myself that a few times a week myself, but then I see a thread like this one...)

How did you get started? I've toyed with the idea of building something like a simple MACD system as a homework exercise - with the real value of the exercise being "do the backtesting that proves something that trivial won't work anymore."

Is a "homework assignment" like that a viable way to kick off the cycle of R&D/test in real life/see how well the execution compares to the model/etc? I have no illusions that hacking together the algo equivalent of "hello, world" will make money, but will it give me enough of a critical mass of knowledge that I can move on to building interesting systems, one of which might actually be profitable?

ITT

I'm not too well read on mathematics but I was wondering how I would go about learning how to use mathematical modeling and algorithms for trading purposes. My engineering course doesn't cover them and I'm generally interested.

>>873349
What computer language do you use to make your algorithms? Curious, because I might be headed your way in a couple of years.

As someone who dabbles in trading and is a programmer, but isn't an algorithmic trader, I'm going to guess what the right answers are, for fun.

>>873718
Mathematical models are ways of describing something in a more concrete matter. Maybe they use them to figure out which way a sector of the market is going in general?

An algorithm is just a formula or a way of doing things. If you put in some values, say stock price and beta values, or dividends, the formula is supposed to tell you whether it goes up or down. If it's right, you can make something that uses the formula on it's own to make decisions on stock trading.

Now the problem becomes finding ones that work. From what I can tell they're pretty decent at probability/risk because statistics is just statistics, but getting money out of them isn't easy.

>>873749
It probably doesn't matter, unless he's using a specific scripting language or has a library he really likes. Python would do just fine.

>>873926

i'm learning python right now! in school, my engineering program was matlab-based but havent touched it since graduation.

started learning C a few months ago but got stuck at pointers. i switched to python now.