I've implemented a NLP program for this kinda thing, actually a full prototype (parses news from a feed, then trades and manages positions), but not sure how to tweak it to be profitable. Used to do high frequency market making. PM me your email.
unfortunately, you're about 6 years too late. flyonthewall was good for a time. there are various machine readable feeds now
u will always be at least 50-100ms slower than the big guys, depending on where u r. In fact they will be selling to u. If u think about being in the "middle" of the 2min "trend", then how much are you risking in this algo ? Few hundred thousand $ or just to buy/sell 100-200 shares ?
The fastest "news reader" I know can generate alpha signals within 2 microseconds of the news arriving at the *first* hop in the delivery chain. By the time the news gets to where normal traders can even start to process it (you are not the first hop ), the really fast players already have their orders queued at the exchange. That doesn't mean there isn't a middle ground that can be useful, but you should be aware of just how fast "fast" really is. It's really not that hard to build a system like this, and the parsing capability can be quite complex in a very short period of computation time.
you're missing the point. unless you're simply trying to capture pennies (and retail traders shouldn't try to capture pennies because they will never be able to compete with market makers or funds who have a commission edge and the ability to pay for the best and latest technology), you don't need to enter a trade within milliseconds.
Who is "you"? There are market participants for whom this makes sense or might make sense. For most, it doesn't. Everybody needs to decide for themselves where they fall on the spectrum. The conversation wasn't limited to "retail".
So u wan to spend a few hundred thousands on a system designed to beat the retails, who usually loses 95% of the time ? Go ahead. We, programmers, CPU servers shop, news feeder, Data centre, is ready to earn your money
I have traded "news" for over 20 years, just the way I trade it changes. It wasn't that long ago you could trade the biggest number of the day, the US trade numbers, reasonably well by calling up a bank/broker after the announcement and getting a price. For the past 6 years or so I have traded scheduled economic figures via a co-loed algo. Now I am quite aware my algo/set up won't be the quickest and that someone could claim I provide free money to the smart people etc but it works, at least for me. Things have changed over these 6 years and some aspects don't work like they did but the set up still isn't redundant. Now back to the point of this thread. I also trade news headlines, of the unscheduled variety. I merely point and click trade when reacting to this news and it works. It is patently obvious that there are algos out there who can react faster than me but at the moment it does not matter. There is enough liquidity, and with my knowledge/experience hopefully, to provide me with enough of an opportunity to make money out of this. However just like I had to use technology to change the way I trade numbers I am aware I must adapt or my method will eventually be redundant. Hence the desire to somehow automate the trading of news. I wont need to be the quickest, or have to strip the efficiency down to the nearest micro second, but i would like to evolve..