Hacker News Reader

Comment by djb_hackernews | Hacker News Reader

>This bot watches Donald Trump's tweets and waits for him to mention any publicly traded companies.

Ok, simple enough, go on.

> When he does, it uses sentiment analysis to determine whether his opinions are positive or negative toward those companies.

Eh, sentiment analysis isn't perfect but it has come a long way, plus Trump does typically say what he means in simplistic language.

> The bot then automatically executes trades on the relevant stocks according to the expected market reaction.

Ah market psychology. Stop. Do not pass Go. Do not collect $200. This is the tricky bit that a toy project will never get right and turns your slick algorithmic trading project into a monkey and a dart board.

Though this is a sweet example of an API mashup.

View on HN

It's perfectly workable. The bot just needs to weigh its trades properly and provided the sentiment signal has any alpha, it can be combined with other analysis to run a profitable strategy.

I've seen this working and working very well (day job), although not specifically with Trump tweets.

If it works, the last thing you want to do is post it on HN, because it'll stop working once everyone starts using it. Everybody else running the bot will bid up the prices of affected stocks before you can get to it, and so by the time your bot gets around to executing its trade, you'll be buying at inflated prices and lose money as everyone else takes their profits.

Not necessarily. If you can get a bunch of unsophisticated traders on relatively high-latency connections (AKA the readers here) to execute a known strategy, you can actually make a lot of money by simply frontrunning them. :)

It's possible to profit from increased volatility without knowing the direction of the move.

http://www.investopedia.com/articles/active-trading/040515/h...

Especially because the effect of Trump's tweets has rapidly shifted. His Boeing tweet hurt their stock price considerably. But just a few weeks later, his Nordstrom tweet helped quite a bit.

I guess you could weigh it against his favorability polling, but you're right, this is all just guesswork masquerading as science.

Those are two very different companies, though. It's not like the average consumer can make any sort of buying decision that will affect an aerospace company, but they can certainly patronize or boycott a department store. Plus Boeing likely does a lot of business with the federal government which Trump is now head of, while retail outlets are highly unlikely to be directly affected by anything a President does.

That's true, which means any algorithm would also have to take the nature of the company being tweeted about into consideration.

Can confirm. I've been running this bot for 8 minutes and currently have a $7.8 million short position against the fake NYT. SEE YOU IN COURT, @maxbbraun!

So, we should just invert the bots request. If Trump reacts negatively, the bot interprets that as a buy. And a positive Trump reaction should be a sell.

Now your holding lots of Nordstrom and NYT, and riding the wave up. Congrats, your going to be a millionaire!

Instead of sentiment analysis directly maybe use a model to predict whether the stock is going to rise or fall.

Invert references to media-related companies. And double down on references to defence contractors.

Reply on news.ycombinator.com