Our First Mini- Hack-a-thon and Results From Past Projects

Last night, we did our first mini-hackathon.  

Fun stuff.  Seriously.  It was like going back to college and staying up late goofing off and chilling with laptops in the dorm rooms.  This is going to become a monthly tradition if not more frequent than that.

Anyways... our newest project is something more ambitious than the projects I've posted.  And I can't get into all the details until we have our prototype / beta up and running.

But the basic premise is that we're using certain events to predict dips in stock prices.

What are the implications?  Well, there's a lot already done in the algorithmic trading space and even news mining, but we've found a niche that doesn't appear to have been focused on much, yet.  This means that we could potentially short shares of the company or execute more sophisticated trades using option contracts... for profit.

How much profit?  Some of our initial research showed dips of 0.5-2.0% on a given day.  That's HUGE!  If we can successfully predict any stock movement a high percentage of the time and also be confident that the dip will be with a specific range, then we can make money in the markets.

Disclaimer - this doesn't constitute as trading advice.  I don't want any SEC issues sharing our strategy for trading using our system.

So, you might be thinking 0.5%, 2% those aren't big numbers... and by themselves you're absolutely right.  But I'd suggest you check 'Flash Boys' sometime and see how those programmatic traders make money off of lightning fast trades that exploit 1-2 penny opportunities.

While we don't have millions to trade like these guys, it brings us to the first way to make money on small percentages - volume.  Basically, if your trade volume is large enough, you'll make good money.

Say you trade $1,000 worth of Stock A @ $50 (that's 20 shares) and it moves up by 0.5%.  That's $50 x 1.005 = $50.25 per share of Stock A x 20 shares => $1,005.

You've made $5.  No big deal.

Now if you had traded $10,000, you would have made $50 bucks.  Or $100,000 = $500.. etc.  You get the point.  The more money you throw into the trade, the more money you can make (or lose..).

But as I said before, I don't have millions of dollars to throw at these trades to make them worthwhile.  Instead, we'll use the second principal of leverage.  In the trading world, one of the ways to get leverage is to use options contracts.  There's already been plenty written on options and you can find one of my favorite reference guides here.

The trade strategy? Essentially, we're able to exploit the small movements in stock prices using a 2-2-3 strategy of buy 2 puts ATM, sell 2 calls ATM, buy 3 calls OTM.  This is partially a credit spread to offset the cost of the puts, the extra call is optional and done expecting the price drop to be temporary.

Since we're expecting the price to drop this will be a directional play.  Once when there's a dip, we can cover our calls and sell the puts for a profit.  Then the calls remain open normally as free (paid for from profit) contracts that can maximize the upside.

In practice the ratio works better at 8-8-9 to 10-10-11 as the extra contract cost is offset by the other contracts.

--

Anyways...

 I digressed into discussions on trading options (which was a previous life).  Our little project will automate the tracking of such events and also do the calculation of the various trades available and the profit potential.  All trades should be positive Expected Value (EV), meaning if the payout is 100:1, we only need to be right 1/99 times to be profitable.

It's a big project, but we've broken it down into manageable chunks.

1. Event Tracking - Creates real-time signals for us to monitor
2. Computation Engine - Performs the calculations on options contracts to determine the EV of various trades.
3. Learning Engine - Tracks performance of each trade computed overtime and calculates variance of EV vs. actual value if the trade were executed as expected.  Provides recommendations for ideal trade.  Uses both Recurrent NN and Reinforcement Learning.
4. Trading Algorithm - Performs the trades based on EV and variance.

Our mini-hackathon left us at the half-way through the Event Tracking phase.  A few more scripts should have our monitoring system up and running. =)

Maybe, two weeks and we'll have a vetted MVP.

--

Projects from Days Gone Past

So, during the initial creation of this site, I set out to build a few mini-automation projects with IFTTT and other services.  Looks like 1-in-3 is running as expected.  The Apple Store Freebie is definitely working as I see more posts being created daily.  I've not seen any posts or successful runs from the Google Play Freebie or the Amazon Kindle Free Top eBooks.

I'm going to go back and explore the Google Play Freebie Project and see if one of the other recipes will work.  Otherwise, I'll look at other options for automating that.  Not a big deal either way.

For the Amazon Project, I'll leave this current project running to see if anything will turn up over the next few days.  I planned on revisiting the project anyways to build a new RSS feed using a different tool.  We'll see how that pans out.

All in all, not a terrible result.  We're here to build, test and learn.  If I've learned one thing in programming, its that code rarely runs correctly the first time.  It's a trial and error process of debugging and testing and validating.

Cheers!
Alex