Would any of you be interested in trading software that notified you of potentially beneficial trades (including trades...

>Nigga, this is simple graph theory.

the OP pic is simple graph theory, obviously (weighted directed graph). I'm not retarded.

>trying to write a bot in 2017

You know how fucking late you are? Unless you are doing quant you aren't gonna make it, and arbitrage is a dead end, I'm telling you now. A good one is identifying ta patterns and buying the breakouts, backing that up with stats (likelihood of breakout and estimated growth)

You considering deltas between fiat currencies too?
Because of low activity, you could get 0.5 bitcoin for a ~5% discount if you bought in GBP on kraken earlier.

That's where the novelty in delta prediction comes in. I can't go into too much detail, but it's promising so far.

Good idea, but we've thought about that. Finding an optimized starting position in the graph is part of the winning strategy.

I’m interested - how to beta test?

>trying to write a bot in 2017
Why should I not try this?

Sameseizes

sounds very interesting I will pay schekels

looks comfy OP

If you could make money with it, why release it?

t. Making money with python no buttons a day

Because you'll be creating an a.i. algorithm which will be interacting with an innumerable quantity of other a.i., a spaghetti mess of competing interests acting without human rationality yet handling real finances.

It's all fine and dandy thinking "I'll test it thoroughly and will know what I'm getting into," but every exchange is already saturated with bots that compete against bots.

OP wants to exploit arbitrage, but that's a saturated field as well. Even a rigorous approach will encounter limitless setbacks.

It's actually the least comfortable, an idea which would require endless continuing work to implement.

make beta

>Too drunk to be posting about this at the moment
well with that level of integrity and ambition, where do I sign up to give you all of my funds?!

t. drunk dev millennial

Even just as a pure arbitrage engine, there is money to be made between exchanges.

As far as a beta test, I need to get it to a point that I'm comfortable with. But I'm seriously considering it. We're thinking a small % fee (0.1%?) to try to cover the cost of development.

Unfortunately, it isn't just time we need reimbursed for. We're using some academic hardware to train our NN, and that shit ain't free.

I'm very encouraged so far, though.

As of the dataset I'm currently using, just using pure arbitrage trading between exchanges, in simulations I'm getting a 5% return for literally no work other than moving currencies around.

Like RenoBot already does?

renoscoin.com/tutorial/how-to-setup-and-configure-renobot/

this is crypto not forex, you're vastly overestimating how competitive this market it.

and if the the major coins are too hard for you just scalp lower volume alts, the market makers for those are just a bunch of dumbasses that buy the rumor and sell the news.

Similar idea I suppose, but completely different methodologies.

Totally understand why you'd say that.

Thanks everyone for weighing in. I'm still optimistic, given the test results thus far. This has been the culmination of months of work.

Some exchanges round the numbers btw.

Cindicator's already doing it bigger and better.

Do you take into account the impact your orders have on the market in your simulation ? Or do you assume the market is perfectly liquid and can absorb your orders ? If so, then 5% is nothing

Because I feel like bragging, I also wrote a bot. And I swear the idea is so simple all of you would rage by knowing I'm making bank with it.

This is accounted for in the training data -- good catch, though.

I have no idea what the fuck any of this even is. Can't you just spend your time finding one trade a day to all in and make 10-30% a day like me? Your Idea seems like a waste of time...completely unnecessary.

Nice. If you want to chat, chanmail (at) protonmail.com

Oh yea I forgot..

Tfw you suck at math but btfo of mathematicians. Fucking dorks, I swear.

So you're making a minimum 1750% return a month? You should be a hedge fund manager.

wow congratulations you invented loopring

It's called an arbitrage machine, and people already have them
Making one public, however, is a different deal.

between liquidity and trading fees, I'd doubt you could make a lot.
Then you also have to know the order book is real. If its fake your engine will lose you money.
and finally publicly giving your engine out will negate its edge.

I'm doing something similar wanna talk?

Msc in math and neural networks
skype:bruisedcrow

brainlet here. what do we see in this picture ?

how are you any better than

github.com/butor/blackbird

?

Not related to anything Veeky Forums related, but neural networks are awesome, but I haven't found any good literature yet on 'unsupervised learning', do you have any resource recommendations?

OP,

how are your work any better than

github.com/butor/blackbird

?

Too much work. I didn't start trading shit coins to feel like I was working.