Search results

I wonder how they manage to train a deep network using single CPU cores and no GPU... without taking months to finish the process.

The larger size you trade, the more that slippage becomes a problem. This is true on any market, those who are trading large size do so patiently accumulating or distributing their positions slowly, bit by bit. 20 lots will probably not move a market as liquid as the EuroUsd much, but if your...

Most exchanges work with Price/Time priority. Where price means that Bids and higher prices and Offers at lower prices get filled first. Then within each price level, the oldest order gets the first fill. But there are some that operate on pro-rata or Price/Size priority. In this case, within...

AWS comes in handy if you need to scale up in heartbeat and then scale back down in another heartbeat. For example, if you want to collect every tick and every order for every symbol like FINRA does...

and they're leasing for around $300... http://www.cmegroup.com/company/membership/individual/cme.html#pricing

I wonder how that worked, since T3 requires traders to be registered and licensed... and WTS' didn't

Some folks at MIT have designed RNN's to compose music. http://web.mit.edu/felixsun/www/neural-music.html This article has a good step by step on how it works http://www.hexahedria.com/2015/08/03/composing-music-with-recurrent-neural-networks/ A youtube video is worth 1000 articles... so...

Btw. This paper seems interesting & relevant to this topic. https://arxiv.org/ftp/arxiv/papers/1603/1603.00751.pdf

Im not sure if it is the right approach... if anything it'll allow to reuse techniques developed for machine vision (such as CNN's) and *maybe* avoid some of the pitfalls of working with timeseries data... Although RNN's do seem to be a more natural fit for the problem at hand. My initial...

Maybe, literally teach the CNN to read the chart with colors and all...

Hey guys, pretty nice discussion so far. How do you get adapt the CNN's design to work well with time series data? I was under the impression that these nets are better suited for machine vision tasks, and other nets such as RNN's (LSTM in particular) and Markov Nets are better suited for time...

But on the retail level you're paying way more than the $0.20 on 100 shares that you get from rebates... The rebate may help a bit with the cost, but it's not going to put you on the green in that league.

maker taker fees get more important as you get closer to institutional HFT... if you're holding positions for sub-second to a few seconds, then those fees are your bread and butter; if you are a retail investor holding positions for months, they're just a rather small fee on top of a much larger...

Thats a good point.

I think he means that he develops the technology, and can take a strategy given by the trader and put it into software... He states that he has a platform, fully certified with CME, not that he has a profitable trading strategy. It sounds like he's a developer. not a trader... and is looking...

This one's $49... from Yale... https://www.coursera.org/learn/financial-markets

most brokers offer some kind of smart router... but even if routed manually you do have some protection from the market's rules... If you're taking liquidity on a retail platform it should make little or no difference. If you're working on an institutional system or optimizing an execution...

you can plugin the yahoo api to python, that way it takes about a second per symbol to get a good chunk of daily data, make calculations on it and dump it to csv or have it loaded in memory and ready to work... this guy explains how to do it in a pretty straight fwd way...

They only execute a small % of daily volume across the different indices. But thanks to MiFid2 best execution rule the prices should follow each other very closely... even if it isn't as liquid as the main exchange. https://www.batstrading.co.uk/

yahoo... https://ca.finance.yahoo.com/q/hp?s=YHOO They give you the daily prices, put in excel and calculate the percent changes.