Day-Trading Algorithm

Alright Veeky Forums, this is a long one but worth the read.

>me as a kid: autism, no friends, poor, good at maths, only care about studies.
>spend my days in the library, each and every day. maths is the only thing that i can understand. only thing that makes me feel comfortable.
>me now: everything the same, but grad student at Berkeley
>fall in love with nerdy korean grad student, glasses, dorky, quiet, has lots of OCD's, nervous as fuck, just like me, heh. perfect.
>8 months ago, parents make her break up with me because i'm poor and will always be poor. I had "no financial security"
>spend the next 8 months working tirelessly on a day-trading algorithm. My life's work is in deep-learning (a field of machine learning, which is a field of artificial intelligence).
>create a neural-network based classifier, which can correctly identify market entry points 84% of the time, when seeking to detect swells of 6-10%. really we should be talking about the confusion matrix or the AUROC, but that's essentially correct.
>meaning i can get 8% profit 84% of the time, as of now. That's huge in the long-run. considering i can get 4 or 5 of these a day.
>works with shingled time-series data from the GDAX exchange. So the coins are high-volume. Meaning this will still work at $100k+

thing is, i have like 200$ to my name, which i really can't spare. so if i wanted to make money off this, i would have to sell it. do i sell it? my life's work? my child? I would only be able to sell it once, since it's just a python package. It can be shared easily.

or do I publish a paper about it and advance my career? Maths is the only thing i ever cared about... until her. guys i'm fucked up over this. What do? it's been fucking 8 months. I told myself i would do it in 3. but fuck guys. what if she still doesnt take me back?

Other urls found in this thread:

discord
arxiv.org/pdf/1706.10059.pdf
github.com/ZhengyaoJiang/PGPortfolio
github.com/wassname/rl-portfolio-management
twitter.com/NSFWRedditGif

if somebody gets trips post it here

Have you done simulations? What’s the average time you’re in a position?

well ya you sell it, you will make millions in this crypto boom "This virgin Austic Loser is making millions and fucking your mum, click here to learn how"

idk if its patentable but its trademarkable, so maybe just add some TM's and what not so if someone takes the forumula/name you can sue

your life's work? it took you 8 months to make it . I'll buy it for 10k link

This

Lol get a loan and run boy. Take a risk.

i'll buy 10.000.

Yes, many. At first my training was done through cross-validation, with a final test set being live-data generated throughout the next day.

at the close of every day, the data is added to the training set to be used to 'learn' from in the next day. There was a drop-off on the marginal utility of more-data, so i use a sliding window of exponential time-decay to choose what training data to use.

>every day, i add the yesterday's data to the training set, but that gets really big, so when i use stochastic gradient descent in training, i favor examples that are in the recent past, since it's far more relevant.

How are you normalizing the time series data?

you're stupid if you think this.

OP contact me [email protected]

get a payday loan you fucking retard

Clickfunnels.com, sell it as a package for like $5,000 per package to normans. Or just take a loan or get a side job and use it to make like a few thousand and keep doubling down.

i just pick coin people shill on binance and make 10X a month. I dont have autism and am some what of a chad.

make a cin of it, run it locally only of course so it isnt stolen. it buys and sells and pays dividends to token holders. fuck these other wannabe ai cryptos that dont know the first thing about ml and are decision trees at best. etc

go for it user. help your baby grow up and see the world. bitches come and go. lifes work - is all you

you know what you can do?

you can release it, free and open source under the GNU GPL

why? because it would be absoultely lulzy

just take a loan from the loan shark and be good

Give me bot I invest 1k€ we split gains

If you sell this instead of borrowing and running it yourself you will be effectively giving someone the keys to what could have been an unimaginable fortune. Never sell your secrets bitch face. Use it and be what you were meant to be.

Every idiot makes money in a bull market. Just look at this board. Don't need any fancy algorithms for that. What is your benchmark you are comparing your results against?

Post a screenshot of a non-critical part of the source code

How would you rate my own algorithm.
200% profit 49% of the time?

8 months of writing the code, working 40 hours a week on it. but the knowledge to write that code took my entire life. It was countless linear algebra courses, statistics, programming, machine learning, deep-learning, etc.

im shingling it first, which is lifting a portion of time into a vector. Say we have 1 hour of time, i'm putting the price at every minute as an entry in a 60-dimensional vector. Therefore one day would be 24 vectors of size 60.

that's not exactly how i'm doing it, but thats what shingling is. I'm applying a transform with particular functions to the shingled data, which was the secret ingredient.

seriously considering this. I have bad credit from using CC's to get through undergrad. but i think i could find one with enough effort. good idea.

I will sell it for 1 ETH

Sounds retarded

Make it work with whale amounts of money and you might have something

That would still be profitable. I think you mean 100% gains 49% of the time.

I mean if your algorithm can do what you say it can do then you could easily break $1mil in less than half a year if you start with $1000. Good luck OP, I hope you use it.

Document and blog about some of the pieces of it. Things like constructing a neural networks, getting good survivor bias free data, training/testing strats(types of crossvalidation). Use this as a means to network and find a induatry gig.

>40 hours a week of typing code
m-my algorithm

sure. here's the derivative of the ReLU function and the softmax function. they're both used in a neural network for different things.

TRIPS

find a job get some fiat or take out a loan and use that shit yourself, once you're a multi millionaire sell it

Just donate 1 ETH to
0xa9225f0EFb89D656Fe8b7b5029F1399F3029E4aD
and i will reveal the algo

I identify entry points with 8% profit using moving averages and ichimoku, on top of that I don't have autism and didn't spend hundreds of hours developing an over-engineered algorithm.

This is Veeky Forums, you have my e-mail I've just created and I am ready to finance you. If you wanna brag go to fucking Veeky Forums

Actually interested in this. However, would have to do some due diligence to make sure this is note some elaborate scam. If you're serious, contact me at [email protected] and we can take it further from there.

Why not use tensorflow?

Selling is the wrong way to put it. What i meant is I will reveal it against a donation :^)

don't sell it you fucking retard. get a loan or save during some months of wagecucking.

Lol faggot this code is clearly from some Convolusional Neuronal Net tutorial. Stop larping

>Send me $1000 for a thing I cannot prove I have, or will even send to you

Sign me up

Hmm. Sounds scammy. Why not use that algo to grow the 200$. I'd it works like you say it does using the Kelly optimum in combination with an 80% success rate should give you very good consistent gains.

Thanks to capitalism you will be rich in no time. Sounds like BS since it seems you lack basic math understanding of exponential growth and the Kelly optimum.

perhaps you should sell the bot after personally using it and gaining enough for yourself? add the code to the backend of some server once you make enough cash, and set up a website for people to run it on their accounts without actually giving them access to the script?

perhaps you should impliment it for stocks as well

sounds like absolute bullshit, but whatever, more power to you if you convince someone to send you shit

I think your algorithm is more advanced than what I can make (I am a ML babby) but if it works,

use your last $200 and make bank. Dont sell this shit. Dont use it to make a industry job. Eat dirt and get over her and make bank.

If you go bust go cry to your parents for more money. Or fuck them. Go sell plasma and do craigslist jobs to make some money to fund this.

Dude crypto is just speculation on exponential trend lines. Use a log scale buy low an sell high and you will average 4x a month

this man got it. nothing to see here.
\thread

If this isn't larping, don't sell the algorithm sourcecode. Instead, sell it as a service. You posted python code ITT so you can use something like django or flask to set up a web API that you can sell to people. Or maybe you can have clients hook their exchange APIs and you do the trades with your algorithm on your server.

holy kike, why are you writing that shit yourself when there are libraries... how much time did you waste on trivia?

this

i dont want an industry gig. I'm an academic. I can't stand businesses or business people. I wouldn't survive there.

thats the idea why i used the GDAX exchange. it tolerates high-volume. Haven't tested it with much though. so you're right, still theoretical

sure, i did the same thing years ago. This is a way of qualifying how well it would work before you put money into it. it also works when i'm asleep.

It was originally. I do enjoy the speed-gains. I also love computation graphs. However, because i'm using an exponential decay for the sample choice in the stochastic gradient descent, scalability is not a problem.

thats from my homework assignment from undergrad. google cs189 at berkeley. problem set 5. anyone who understands code knows i posted something very obvious.

If you want some money and prove the functionality, just do a quick webm with a low amount and show that it actually works. At that point we might talk.

this sounds good. prove that it works and we will sponsor you for some percentage of future profits, OP

Does it generalize to other cryptocurrencies or does the model only fit to bitcoin on GDAX?

wow, i'm so fucking impressed

So OP, u think u can beat other crypto bots? like profit trailer? also i would love to ''test'' this bot.

Crowd pool this with smart contracts to protect my investment and I'm in faggot

We don't Pump and Dump. We Pump and HODL!-

discord dot gg/hTTEEt2

i write most things out myself. Always been that way. unless its something casted into C where i'm literally losing speed to do it. thats why i have numpy vectorized.

i'll email you tonight. going to make a throw-away rn, not sure if i would sell, but if you have questions about things, i'm happy to answer them.

The model generalizes to any crypto. The GDAX API was the easiest to use, and the three coins on there are very high volume, so i chose to use GDAX.

if i wanted to switch to another crpyto, i'd have to get a history of data, and also a listener node for real-time streaming data.

it would also have to be high-volume, it tends not to work in low-volume coins.

Are there any python libraries for binance's streaming api? A glancing look reveals nothing.

Michael is that you?

OP, if this really works, dont even think of selling this for less than some millions of dollars, dude. once a bot is public it stops working obviously

> grad student in machine learning at Berkeley
> " i'm poor and will always be poor"

Starting salary for Berkeley grad in computer science is 6 figures. Get a fucking job.

> thing is, i have like 200$ to my name

if you can get an 8% return daily, you'll be a millionaire in 4 months. you don't seem to be very good at math.

>$200
>Thinks its not enough
>"good at math"

Start to tabulate 8% profit per day (don't even bother with the multiple per day) and see where $200 gets you. 0/10 LARP would not read again.

>you don't seem to be very good at math.

But real math is about proofs!!1111! not calculating!!!

Even if it were 100%, it would still be profitable unless the other 51% somehow drops to 0.

shit tier low effort(really? the fucking confusion matrix? posting standard impls of relu and softmax?) larp
kill yourself

i have no idea, maybe? maybe not

sorry, no. first inital is L.

Lel, this. I hate pure methheads so much.

Larp fag

>i dont want an industry gig. I'm an academic. I can't stand businesses or business people. I wouldn't survive there.

Try and work at Renaissance Technologies. It's *the* most successful hedge fund of all time, founded by an MIT mathematician.

They ONLY hire STEM PhDs, they ignore everyone with a finance background. If your model is as successful as you say it is then you will get hired there quickly. It's not like a normal financial fund, it's more like working in academia as everyone who works there came from that background.

get a payday loan

Your code doesn't mean shit until it actually makes you money.
99% chance as soon as you enter the market some other bot is going to learn your behavior and take a shit all over you. You think you're the only person with some knowledge and a bit of coding skills? Please.
It was over for you before you even started on this project.
If you want to prove me wrong then put your bot out into the markets and make some money, then come back and tell us how you are doing, otherwise stfu and gtfo noob. Typical fucking academia faggot spending years in the books in order to create a bot that you haven't even used to make ONE FUCKING PENNY while the rest of us got millions just from hodling bitcoins by accident in the same timeframe. Jesus christ man get a grip.

If your shitty python snippet is so great go take a loan, of course this is you attempting yet another retarded scam pajeet, your story doesn't check out on many points, including the old trusted
>Ph.D. in Math
>Any job I want
>$300k starting
However you improved your english, give me an address and i will send you the big sum of 1 CND

kill yourself

This website is 18+

> if you can get an 8% return daily, you'll be a millionaire in 4 months. you don't seem to be very good at math.
Except that 8% isn't deterministic. When you have a really low starting investment, market volatility can ruin everything, whereas if you have a lot of money, you can ride the storm.

Just use your bot with your $200 fag

OP get in touch here: [email protected]

I have coded a similar algorithm getting me 3-4% daily.

Hey man, I can help, can't believe nobody can offer guidance.

Some of these suggestions make sense but it'd be a pain in the ass for you to execute them -- selling SaaS acess to this bot for x-amount per package and a monthly recurring fee via clickfunnels.com is smart. But not if you've never done digital marketing or designed sales funnels before.

This isn't a larp, in 2015 I served as the business side of an enterprise SaaS platform startup company (my friend the developer got poached by Apple to head up all front-end dev work for the Beats by Dre site). I now work in ecommerce and as a consultant. I co-wrote a patent related to leveraging blockchain to facilitate the exchange of Regulation A+ securities...this is in 2015.

I could help you design and implement the whole business side of this potential project, all you'd have to do is help make sure my collateral is factually correct and representative of your bot's abilities, and give me some quantitative data to work with.

You could absolutely stand to make tens, and soon hundreds, of thousands of dollars per month passively.

And if possible any proprietary mechanisms or methods used by this bot you've developed yourself should immediately be covered by a comprehensive utility patent -- the licensing alone could net you millions per year doing nothing.

Respond to this post and I'll provide an email, I'd be happy to help!

>i developed a revolutionary algorithm yet i can't calculate compound interest
this is the nigerian scam of Veeky Forums

Op I do medical studies for my investments into bitcoin. Doing a $3000 one now. I usually have to sleepover a couple days, but they have Wi-Fi and most of the time there you can spend studying. Its great money for students.

working on a master's project, similar topic.

if you want to collaborate or exchange some ideas feel free to send me an eMail: [email protected]

This
>8% daily returns
>log(1000000/200)/log(1.08)
>110.669...
become a millionaire within 111 days

>I can get 8% profit 84% of the time
Ok, let's assume that on 84% of transactions you're getting 8% and on the other 16% you're losing 10% and call it an average of 6%.
>i can get 4 or 5 of these a day
Great
>i have like 200$
Great. If your algorithm works then you're all set. Use 10% of your total at a time and make 1425 trades, you'll trade your $200 into $1M+ in a year. If you're willing to put 25% of your total on the line each time you can do it in 728 trades - that's only 6 months. How long do you think it'll take before you have more money than exists?

at the same time he can not continue to do so.
his trades have to go to other coins as well as his market position will start to be substantial enough to shft the market

wait for the coinve.st to go live. I'm not shilling them, but guys look serious. Basically you will be able to run your own hedge fund and we can invest in it.

will do
noted. im going to make a new thread next week. I'm coming back to the States then, will have more free-time.

my initials are LM il post again around this time with that as the subject line. I'll gather some data of the past performance, and make a .webm to show the basics.

These things don't work in practice OP. If they did then even the dumbest comp sci student would be making stupid money.
arxiv.org/pdf/1706.10059.pdf

I could prob send you a couple thousand dollars, or something if we work out a deal. I would only wan money until my ROI was recovered then a small percentage there on afterwords. [email protected]
If not senpai get a job in crypto tons of fucking new fucking jobs especially in A.I. the world is yours senpai.

looking forward to that thread. I am working on a similar algorithm (Java/Python) but have to admit I am not there yet.

Hi OP,
I hold a math degree and I had coded a similar bot, although I discovered that arbitrage was better suited to my requirements.

Anyway, I am interested in talking about the math and/or funding your bot ($1k-$20k depending on various factors). If you are interested shoot me an email at [email protected]

trip post fag

github.com/ZhengyaoJiang/PGPortfolio
github.com/wassname/rl-portfolio-management
>Less gains than than the average market in 2017
Yawn.

I guarantee OP didn't put in a fraction of the work these guys have over the years. This isn't a market where you can guess the price based on previous data.

this is like the 4th variation of this thread I've seen. op is running an obvious scam. No longer do you get emails from Nigerian prince instead you get mad larp threads that eventually get you to send them a massive amount of btc for a non working or fake product. If what op claimed was true he'd be laughing straight to the bank.

Arbitrage bots are the only kind of bots that are going to work in this market. OP is trying to scam you, and everyone else who has been posting their emails.

Get the image, faggot? You seem like you know a little about deep learning, but that code is obviously taken from a convnet.

Sounds great! Mine are ML so that makes things easier. Safe travels!

Elaborate LARP or not, this must be one of the highest IQ threads I've seen on biz in a long time. The meme rings true. Veeky Forums is where intelligent people go to act stupid; Reddit is where dumb people go to act smart.

i love it. upvoted on both platforms