/ABG/ Arbitrage bots general

I'm getting almost done with my Binance arbitrage bot and I know there are a few fellow botters in here.What minimal percentage do you go for? Obviously on Binance we want more than 0.20% as that's the fee for 2 trades (without using BNB, else it's 0.10%), but trying for opportunities as low as 0.21% would result in fails where the orderbook moved before you sell and you end up with a losing trade.

Which bot is it? What other exchange are you using?

Only Binance, I'm doing arbitrage between BTC and ETH pairs. I coded it myself.


Source open? I haven’t even looked over at their api, have you actually seen a difference over the fees for that pair?

I always thought arbitrate was between exchanges. Have a git?

How much has it made? How long did it take? How much eth is it using to run across all the pairs for those tokens?

If it’s open source so I can make sure you aren’t robbing me and it works I’d send .5 eth.

>between exchanges
by definition, it is. now fucking clue what OP is talking about, he's likely misinformed and just has an algobot.

I think he is finding an arbitrage difference for eth coin pairs to btc coin pairs, which if any exist at all past the fees and he can just let his computer run then I want to know more.

he is doing triangle arb on one exchange

Does a difference even exist between trading pairs on binance? I am nearly done writing an algorithm bot for binance and may need to make adjustments.

Not really interested in sharing it, for now it made no money as it's not live, I'm still finetuning it, hence why I came to ask for advice.

I buy a token with ETH, sell the same token for BTC, then buy ETH with BTC and end up with more ETH than I started with, if well executed.

Binance already has a bot that does this, as with bittrex etc. You need to arbitrage between exchanges if you want to make any money.

I’ve wanted to do that, I need to pay more attention to when refs are open on bittrex etc. I’m only on binance.

trying to learn python from scratch to do exactly this but on a different exchange where a similar "triangle arbitrage" is noticably exploitable.

may many gains come your way lad

Wouldn’t the difference in price be created by the order book? If it was the exchange doing it to itself that would be strange.

Triangle arbitrage confirmed.

You could check this thread in a few hours, but it isn't very likely that any skilled bot user frequents biz. You'll get some mixed answers without details, but nothing worth reading.

Seems like you'll run into latency issues, unless you use predictive comparisons between BTC, ETH, and whatever other token. If you haven't read much on TA, it might be a good time to start. And yet, TA is mostly worthless.

You'll still be better off brainstorming ways to use predictive comparisons than you would be reading Veeky Forums replies, tho

I’m also guessing that if you have enough crypto to let it run into higher amount buy/sell orders I’m sure there would be difference from time to time

Wait the exchange itself has a bot that does this, or does someone already have one deployed? If it's the latter I can beat them to the punch if I have less ping than them.

Theoretically wouldn’t the exchange bot run local at the exchange?

If it’s the actual exchange it’s the first I’ve heard of that

No, large exhanges have their own bots they use for liquidity and arbitrage. You will never beat them as they don’t pay trading fees and will have the upper hand in speed.

What languages should I start with if I want to learn how to code for a trading bot like this? Python or something a good place to start?

Literally anything that can send a HTTP GET and a HTTP PUT. Mine is in C#

Mco is gonna moon lambo

Think ahead of your impulse, here

Is it all that likely that your network will restrict your performance, even if you cloudserve your bot local to Binance's servers?

Try another route. Don't compete on crowded fields, unless you're willing to sacrifice your motivation and confidence.

This is a simple case of "if you have to ask, you're fucked from the get-go". There's books and books and codecademy and etc. for normie-level-comprehension fellers like you. Pick a language, read a book, or spend some weeks digging the meat out of the open-source bots.

I'm willing to put in the work, I've always wanted to learn coding but never had a goal in mind so motivation was left a little wanting. Just not sure where to start but I'll look into financial programming and those open source bots once I learn some basics.

If you're gonna start your own fintech project, why bother with shitty arbitration? Why not learn TensorFlow and do some real AI prediction.

Do you only read the real time data and use that?

Do you by chance know how to get the prices, going as far back as possible?

You see what you are missing here, what you are doing will work too ofc, but this is a case where it's perfect to use machine learning. Simple regression will even do here, it will probably give you some very good results and that's not hard to set up.

>This is a simple case of "if you have to ask, you're fucked from the get-go"
This is the most retarded thing I've read all day. Nobody can ever be a beginner at anything, they're simply just fucked from the get-go!

And you were, what, born with proficiency in C#?


Why don't you try community college? Seriously, it isn't some "learn this, then you understand this" procedure to be able to compete in this sort of thing. It is no exaggeration to say it takes hundreds of hours of experimentation and exercise to reach novice levels of understanding. It's just the way comprehension of these things works.

Seriously, if you're asking Veeky Forums, go sign up for your local community college

Software Engineer here.

Made this exact thing in December, except my program scans more pairs than just BTC/ETH, and has a few other tools like viewing trade history for a particular currency, account balance, etc.

There are $1-$10 opportunities once in a while, but in general, it's a few cents at a time. This loops every second and finds the best buying opportunity.

You see to make a good AI OP, you need as much data as you possible can get. That is as simple as the following

price, volume, time

for pref all transactions ever done actually. Or at least a sample of the average in a small time interval. I've actually made a couple of AIs that matches this case very much earlier which gives very good result typically an error margin when training on only 2-3%, that's very low. No telling here though. Similar cases but different data I have dealt a lot with consumption of stuff not prices, but that doesn't matter you see

> (You)
>>This is a simple case of "if you have to ask, you're fucked from the get-go"
>This is the most retarded thing I've read all day. Nobody can ever be a beginner at anything, they're simply just fucked from the get-go!
>And you were, what, born with proficiency in C#?

If somebody says "I'd like to do X at professional level, what component Y should I self-study to do X?" you don't spoonfeed them bullshit, you tell them that it's hard as fuck and takes realistic amounts of time. There isn't any single language that will make that fucker better equipped, he doesn't even know the basics. So he should go learn the basics, regardless of the language.

What about events such as news? I know there are services that have APIs which can feed you this data also.

Adding on to this, the latency kills profit opportunities. Maybe if you bought an AWS server or something near the Binance servers, and ran the script from there, it could work.

Anyone interested in buying this?

He isn't fucked, he just has a long way to go.

For someone who allegedly knows something, you shouldn't equal the two.

Do anyone of them provide an api lol?
I guess not, you have been wiresharking and tampermonkeying to see the requests and responses haven't you?

I actually started on this a few months ago my self, but then I started to code on a different projects. It's not much code there yet, it doesn't really do much this far

All of them provide an API. Have you even looked?


Now news and events happens all the time, and this is what you train your ai for. Making one that actually understands th news and so, no that's a whole other big big big field. No you simply make it to predict the prices from exting data data. Use most of it for training and a smaller part for validation. It gets remarkable good at predicting.

Also smart to configure an ANN here, but most likely they will be very close. However one needs test to see. If you know about it about you could also use a SML derrivate, that is evolutionary computing it really has to run for ages and you really need cluster for that. The longer it runs the better it gets . When I cluster access I ran it for like I think it was 3 months or so, but it didn't get so good as ANN and regression

No offense, but hell no. You can't beat the house on latency, unless you're risking money on future market states.

Who said anything about me? I sure as fuck didn't. And yes, he's fucked for jumping straight into this shit, because he hasn't even started to try to understand how any of it (read: COMPUTARS) works. That's about as fucked as someone who can't swim trying to beat Micheal Phelps, or do you need a more basic example of fucked?

Kek obviously not, I have just googled for getting the prices and volumes a couple of times.

OFC they provide an api, because people make trading applications with them. Thanks user I'm going to have a look at this, I only hope it's possible to get all data. I was about to say that they would know what I was up to if I wrote them, given that it's not possible to get this old data. But again, they wouldn't really care because they make money from all sales.

>Hi girl can you open up your cloth
>Hi girl can you take you panty off
>Send bobs

H-h-hold on there, sir. I'm afraid you aren't licensed to entice

>binance flooded with bots
>op makes binance bot thinking he’ll profit

btw if you want to play a bit with this, check out cubist. It's free for the 1 core version, but you have to pay for the multi core one.

You also have r studio which is a ml powertool, this is free. It integrates with cubist, so you can just load data into it there, you can make your own plugins to read the data or.. You can simply parse your data and write it into a file cubist can read.

Then you just run it, again it takes a lot of time this shit. It doesn't need so much time as EC though. It isn't really so hard this stuff, you can read up on it your self. But ofc doing it the old way as you are doing here works too

>not using 4d arbitrage
>buy coin
>allow time to pass to the point until your coin is worth more

HOW do you secure your API access key and secret? I sure hope you're not keeping them in plaintext.

Wrote a triangle arbitrage bot a while back, made very little profit with it, there just simply wasn't many opportunities. Recently wrote a arbitrage bot between exchanges, easiest money I have ever made. Make sure you are using the websocket APIs for the exchanges, not the REST API.

>buy coin
>coin craters
>suddenly your arbitrage turns into major losses

So what do you use to transfer money between exchanges lighting fast to take opportunities?

Fuck you for asking, but Litecoin is the fastest