Veeky Forums programming

Dumb question: When writing a bot that will be putting buy and sell orders on an exchange, how do you test that it works correctly without losing money because of coding errors?

You play around with just a little bit of money so you don't lose all of your $500

Start with the minimum trade size, on Bittrex it's like $2. Unless you're testing during a crash or dip you're not going to lose much.

Perhaps play on a shitcoin with two accounts? That might get you banned/be considered laundering though.

Include a backtesting functionality, give it some imaginary bitcoins and send it on its way. You need to draw down historical data from exchanges and use them to test your strategies.

Never trade using WIFI, always use a direct connection to your router. Your problems are always latency, lack of liquidity, bad coding, no error management, etc. It's not a simple task.

rent a vps closest to the exchange's servers

Set an order of minimal size and set a price that won't get filled.

Thank you.

Actually, writing bots is too sophisticated, so it's a waste of time.

You'd need to have a supercomputer in order to be accurate and fast enough, plus transactions fees would take all your gains anyway.

Also, the exchanges limit the number of orders you can submit, how often you can call the apis. So once they see that you are pinging them all the time they'll ban your account and report you to the FBI.

Right now there are thousands of chinks running bots that were designed by finance and statistics Ph.D's, there is no way you can beat them.

lol nice just botted 100k!

This, although I imagine it would be fine to scrape the order data in real time. Create a class that represents your order being added to the usual buy/sell list (only locally), and also your change in values subsequent to the order. Have a terminal/console print out a ledger of your orders, as well as a formatted list containing the buy/sell data.

wtf?
Just back test it using historical data.
Slowly feed it into the bot and see when it would/wouldn't make trades.

Top kek, my 700 lines of Python have taken .05 btc to 25.5 since June 13th.

Write unit tests.

Curious what strategy you use? I'm a software engineer, relatively new to crypto trading, and getting really fucking sick of dealing with the UI on exchanges. Thinking of writing up a quick CLI just to be able to cancel my old order and place a new one with one command, but writing a trading bot sounds interesting too. I've heard you can make money just using a market maker strategy (buy a bit below the bid, sell a bit above the ask, profit off the normal volatility).

Use a Mexican instead.

I'm not divulging anything but scalping is a good strategy, you know it works because you'll see faggots on Veeky Forums wailing about it. Just make sure you let the other guy win sometimes, otherwise he'll stop playing.

What if it just keeps going down after your buy gets filled?

see if they have a sandbox you can play with

What's scalping?

are you retardeD? stop fudding kys, i run a bot on a 2 ram 1 cpu virtualbox instance, like what the fuck are you talking about, bots using api keys, to perform high statistics you would only need a lot of ram and that would be trillions of data which you cant get from api, please stfu

basically buying below ema and selling after it crossess ema or goes to a certain % of earnings you want it to,

You can google Bollinger Bands and use those to trade with a bot, you can also try using indicators like RSI to buy and sell.

Im thinking about using deep learning to make bot. Bad or good idea?
Prolly gonna use tensorflow

Oh, Ive only tried it with a simple moving average which wasn't profitable, I'll try it with that

doesnt work, there are no patterns to learn.

So.basically scalping and those traditional bot would be the best? Ill try to make one for project ajd see how it does i guess

Good idea, didn't work for me but it's bound to work for someone.

How do you botters choose what coins to trade? Assuming you have a profitable scalping algorithm would you just go for the most volatile coins?

>how do you choose which coins your bot will trade
The bot decides what to trade and when, that's the point of writing a bot.

Wrong.

if there wasn't a pattern to learn, then no daytrading bot (or human) would be successful at all.

do you have proof of this? if it's true I'm leaving my job rn

By backtesting performance in specific market conditions on specific currency pairs

Having it choose coins over complicates it, it's easier to just pick pairs for it that work well.

Proof or LARP

Zoom out

Exactly.

Before anyone asks... no, I have no idea what it did the other night to make 15 btc.

you don't have a record of the exchange it's trading on? come on user..

I assume there's a danger of "overtraining" by relying on backtesting. Do you just keep that in the back of your mind or how do you deal with that, if it's an issue?

Either something mooned or something pumped and my bot got in on it, I hadn't made any changes recently so I haven't bothered to look.

Not asking for specifics but in general terms do you have any resources or recommendations for how to approach making a bot? Things to emphasize or mistakes to avoid?

so, u made 10 btc from 0.5. that jump is definitely an error, otherwise your bot made 150% in one trade. also, find the account in whatever exchange and view your history

So far I've made around 28 btc from .05, some of those dips are withdraws. I only log the balance if the bot hasn't traded for a minute, that definitely wasn't a single trade - it doesn't throw the entire balance at a single trade.

Don't mistake a btc rise for an alt dip, don't put in more than you can lose, if your bot works don't expect faggots to believe you.

Did you consider buying a bot and customizing it or did you make the whole thing from scratch?

I wrote it from scratch, I've got a CS background and figured it would trade better than I did. It does.

stuff like eth is so violatile that it must be possible to just buy and sell and scalp. are bollinger bands semi reliable?

Congrats, that's a ton of money you've made so far. I have a lot of downtime at my computer for my job so I'm going to try and make one. I'm a little familiar with python, but am most comfortable using R. I know they combine well so maybe I'll use both

Thanks. I only used Python because I started with the intention of using keras. I'd written a Robinhood dividend capture bot last year in C#, which I should probably revisit now that I've got the 25k to daytrade... I think it's up by $0.35, so yeah, crypto is a nice change.

I was always skeptical of 'quant' technicals based trading because I had an econ professor who convinced me that stock market movements are random walk in the short term and so -EV after fees. But with crypto fees being low enough, I've seen enough good results that it merits an attempt I figure

how do you implement python into html?
i'm tempted to write a bot too

You're too stupid, I'm sorry.

so your bot is using machine learning?

How DO you implement python INTO html? Clearly not understanding that question indicates I'm using machine learning, yes.

Are there any free bots where you can just place buy and sell orders?

Like you put an order to sell at ath and know it's going to go down after that and set a buy in at a lower price?

Ever thought of selling it?

I would be interested, for research reasons only, not for the lambos on the moon

i'd spare some time in google without reading shit article if you give me some hints

it'd be a good idea for a very smart bot, but it's better to make it fast so he can buy on few signals (rsi, macd) and sell on other signals
it'd lose some money, but it'll probably won more

I'm not going to sell the bot, no. I thought about building a ponzi scheme around it - what's the coin that does that already - BitConnect? I dunno, I don't really need investment, I'm doing just fine.

props to you, pretty impressive, what exchange are you botting?

I've got bots running on 3 exchanges now! ;)

what do you think about public bots (blackbird, gekko, etc) do you reckon it would be useful to build something around them?

i'm quite a noob at coding, would that be too difficult to approach for a newfag in your opinion?

>scalping is a good strategy
Veeky Forums 2017

Many books suggest otherwise, but if you make money you're good I guess.

I wouldn't sell my bot because I wouldn't want your instance competing with mine to get orders filled... so why would anyone release a working bot for free? Maybe you get enough faggots running your code, pushing the markets in the same direction, and then you've also written a better version that predicts what your first bot is going to do a minute ahead of time, so essentially you've got a P&D scam with a bunch of unwilling participants.

I didn't look at any other bots, I can't tell you what they're actually doing, but I wouldn't trust them.

Do you do arbitrage?

You using progression strategies? Your drawdowns look Martingale-like to me.

You have got ~70% drawdowns in a period of just three months. If you already got such large drawdowns in3 months it is likely that there will be even larger drawdowns in a year long or two year period.
I once did this for a living.

>it is likely that there will be even larger drawdowns in a year long or two year period.
What I meant by this is that the account would likely lose 100%.

what exchange has the best historical data availability?

been thinking about making a bot for myself for a few weeks. is there ample history available for testing?

I know minimal R and zero Python, is there any point in trying to learn enough to write a bot?

Do you find regression analysis useful or do you think it's sufficient to just utilize price history patterns?

Not yet.

My graph is just shitty, those 4 lines straight down aren't losses - they're errors.

When I was looking at keras I was getting historical data from Poloniex for training.

Regression analysis, I'm not looking for batman or triple-fringed teacup handles.

All the inputs will be various pricing averages and things like that. Or do I need to come up with some crazy insight about the correlation of the crypto markets with world market indicators?

You ever hear those guys who claim the Bible contains secret messages?

If secret messages in the Bible told me how to make a 1% a day ROI I'd enroll in seminary tomorrow

Did you study algorithmic trading on your own? Where would you recommend I start - time series or multivariate analysis? What do you think about sentiment analysis?

>The bot decides what to trade and when, that's the point of writing a bot.
So your bot keeps checking every altcoin and tries to catch P&D or does it take advantage of the volatility of high liquidity coins (eth, ltc etc.)?
Do you think it makes sense to take the orderbook into account or just the trade history of a coin and the current price?

python is really easy to learn, if you have any programming experience (or want to get some) just go for it.

I study procedural generation, this is a hobby - I haven't looked at much, honestly. Get coding.

Look at the orderbook but don't believe it. Look at all the orderbooks. Anything can become volatile.

since youve said you used python to use im guessing that u used somekind of nn. then why did DL did not work for you? curious. what was the pittfall? are you using somekind of cloud computing with vps? or just regular cloud server?

i mea keras

Tell you what: You show me a trade history with 560x returns, I'll quit my 6 figure job at Google right now and work for you.

...why do I want to hire you? What can you do for me?

little tip would be appreciated user. im guessing you used cNN? what was the most difficult part? why did DL not work for you? any goot MLT resources?

also like how many layers and stuff? thx

Pajeet, is that you?

Huh I guess you weren't larping. I've written a few bots. The latest one identifies local mins and maxes on a STOCH graph but it's not profitable after trading fees. I'm good at maths but not University level. How complex is your bot?

this is me. i was phoneposting.

nice
started working on a bot a few weeks ago, initially wanted to use dnn's so started to crawl data for every coin i could get from 2 exchanges, currently 2 weeks and running, gathered 40GB of data, didnt have the time to get into the network design yet so just started with testing a scalping like strategy, works meh

do you have a fixed profit at which you sell, like 5%? because atm i figured this doesnt scale because the coins have to much difference in price, 5% increase for ETH isnt that easy as for DGB

Compared to other things I've created it isn't complex at all - that's really nice, actually, since I'm always tweaking it. Keep iterating, you'll make it.

>fixed %
No, that strikes me as insanity - if you're profiting after fees then you're in good shape even at .5%.

You've given me new hope in actually making a profit with a bot

i guess youre right, its a very naive approach at the moment

leaving the point of sell open, what do you base the sell decision on without missing a pump?

you should check out the documentation for the sites API first. Usually the function to make an order returns something like an order number. In the case of my bot it returns an order number stored in the variable "orderData" but i currently don't make use of it.

you can use mine as a reference OP. github /skilo83

You can do it with TA I guess. High RSI, stoch, Bollinger bands top, all of the above, whatever

Jesus christ user, it's called unit testing. Write a mock-up of the API you'll be interfacing with and write tests for every conceivable scenario.

>implement python into html
HTML the language for making websites? You don't, you use the exchange's API.

Did you train your bot on TA indicators or just price data?

GDAX has a sandbox. I would imagine other exchanges do to.

Sup my dudes, I'm a software engineer with some background in math and economics and also an amateur crypto trader. I've been interested in crypto bots for a while now, but never really got around to doing anything serious about it. Is there some discord or something similar, where we could discuss these kinds of projects? Or at least some decent resources to get me started?

Thanks.