Chabah on Automated Trading Redux

[TraderZero]

Quote from maxima120:

You took my question a bit too serious.... The point was - you either BlackSwan fan or TA fan. You cant do both. The whole idea of TA based on normal distribution. And the whole point of Black Swan is that normal distribution do not exist.

A bit of kerfuffle isnt it.

This statement isn't precisely accurate. Taleb's assertion, and the one he trades on is not that TA/etc doesn't work, it's that the distribution used is incorrect (which, as you note would invalidate some TA, but not all). Incorrect however, doesn't necessarily mean, not normally distributed at times.

His assertion is that there is significant negative skew and very high kurtosis in the distribution resulting in signficantly fatter tails than is generally theorized. He also generally asserts that distributions change over time, they aren't constant (just as almost all things in the capital markets change, so do distributions).

As such, he believes certain options should be systematically mispriced in the marketplace. His strategy takes advantage of this theory.

The hardest part of implementing Taleb's strategy effectively is the necessity of finding the "cheapest" volatility in the marketplace in securities that have a beta near or greater than 1. Requires a tremendous amount of computing power, and if you had ever seen Empirica's trading floor, it largely consisted of servers that crunched data continuously in an attempt to identify the cheapest volatility in the marketplace.

The problem with Taleb's strategy is that as he's gotten more "famous", his strategy has gradually lost it's ability to post wildly outsized gains (as occurs with most stategies) due to the shape of the volatility smile in the marketplace. What's occurred is more and more people starting to use his strategy, which results in volatility getting more and more expensive due to people bidding up the price (increasing the volatility) on the OTM options that he favors. Also doesn't help that quite a few high yield fixed income funds have realized that there is a VERY high and shockingly stable negative correlation between changes in spread in that market and the level of volatility in the equity markets (meaning Taleb's strategy works as almost a perfect hedge for a high yield fixed income portfolio).

PS - Terrific thread and i apologize for the off topic post but given the topic of Taleb came up, I thought a bit more info might prove useful. Am very much enjoying following the development of your system. Thanks.

 

[Chabah]

I appreciate the post - good to get back to a little theory after I have been crunching numbers - without the benefit of a server! I have been busy reinventing the wheel, trying to figure out which options fit my strategies the best. I've cooked up a little spreadsheet, but this can't last so I am not sharing - very labor intensive. The worst part is that I can't retrieve option strikes and prices yet. But also I am trying to figure out what the questions should be.

I need to figure out the system first though, then I can try to automate within TD. I definitely like the limited amount of capital put at stake with options, but I am sure there is a lot I don't understand yet which is why I am tinkering with small positions and just getting a feel for the market. Look, now I am off topic on my own thread! But we'll get back to it.

Thanks,

Damien

 

[Chabah]

Here's an update on the last 7 weeks or so.

I have kept the same IN/OUT criteria based on a simple daily moving average. I don't stress if something moves into close position for just a day or so - I really only evaluate trades once a week, unless something pops or drops. So far I haven't had to make any intraday trades since I sold BP the day after the well blew. The strength here is path independence (see Principles). This criteria I consider automated, even though I am monitoring it manually with Google Spreadsheets.

So my ownership is through options. Ah, options. The swings are just magnificent when you are 98% in the market. I can afford that since my account is just $20k - if I were trading say $200k or more I would be more protective of my capital. At this point, I'm willing to risk more for the opportunity to radically grow my account.

The huge spreads on most options make jumping in and out very unpalatable and difficult to automate. For example, right now OPEN Jan 2011 $40 calls are $6.8/$7.5 BID/ASK - more than 10% of the bid. In order to open or close a position you really need to work the middle of the spread and update the order every few minutes. That is not trivial to automate.

My evaluation of each option as to whether it is a good value is consistent - I go through some charts and automated calculations to forecast future values and likely returns. I "like" returns that could be 400% or more. However, I am the first to recognize that my price "forecast" is happy horse&^%# and pretty much pulled out of my butt. Nonetheless, it's the best I have. I buy the best looking options and table the rest.

When it comes to buying, my $20k account pretty much eliminates position sizing. When you have a $60 AAPL option and a $2 SBUX, good luck with position sizing. So my position sizing is pretty much non-existent, and I just bet on my favorites. My account is more than 50% in AAPL options, so there you go. With a larger account ($100k+) I think you could have parameterized position sizing, but it's tricky. Not nearly as controllable as with normal stocks. Being able to pull the trigger is a must though - I lost $1k on MA but I very nearly lost a lot more - following my close criteria after a couple days saved a bundle.

Here's the big question, though - if I can manage this system of 30 symbols monitoring about 10 minutes per day (3-5 minutes 3 times: Open, Noon, Close) and spending less than an hour making trades once a week, how much do I really need to automate the system? Monitoring and tweaking a fully automated system would almost certainly require more time - especially if amortizing the development time over 5 years or so. So for now I plan to keep running my "system" as efficiently as possible, which for the past two months is only about 2-3 hours per week.

Thoughts?

Damien