Learn to Trade Like a Math Geek
Thursday, March 15, 2012 at 12:53PMOn Wednesday we held a webinar "Learn to Trade Like a Math Geek". We had great turnout, and I want to thank everyone who joined us. I've received a lot of emails asking about the recording, which you can find at the bottom of this blog post. We discussed some of the key math terms involved in statistical arbitrage spread trading. Many of the questions were specifically about calculations and how they are done. It is important not to get too bogged down with the math. Cointegration tells you that the two stocks have a history of reverting to a mean level like a spring. When you design your statistical arbitrage trading framework, you want to build a portfolio of these spreads. You should see that most of them behave nicely, while a few continue on their trend away from the mean. You'll develop your own recipes for determining when spreads will behave well and when they will not. You may even develop some recipes for trading spreads that are very different, like the earnings strategy we discussed at the end. The beauty of statistical arbitrage spread trading is that you can design your own strategy however you find it works best. Here is the replay.
Written by Jennifer Galperin. Follow me on Twitter and StockTwits.
Reader Comments (5)
Hi Jennifer, thank you for this webinar.
I have questions regarding half-life and dollar value of z-score. Since you guys don't feed some data from some foreign markets yet, I decided to write an code to check all these concepts in R myself.
1. How do you calculate half-life?
If you take HL = -log(2)/theta , where theta is a solution of dSpread = theta(Spread - mean(Spread)) using LS method. Can you have negative half life?
2. How do you calculate dollar value of z-score? If you use this value to extrapolate entry/exit points for a spread trade, how stable is DV of z-score? Or you recalculate your spread/z-score with most current data once your extrapolation tells you to enter/exit a trade?
Hi fedor,
Jennifer is away for a wekk so I will answer with the help of Aris.
1. Negative half-life is possible and it means the spread is not mean reverting.
2. Simple calculation: (current spread level - mean spread level)/ current z-score. Nothing fancy, just a quick way to set symmetrical levels for me. Stop loss and entry
Trading like a maths geek must work! I read an article about David Harding who employs scientists and mathmaticians to program all the theory and he is really making a killing! His fund started with $10 million in 1997/98 and now stands at a staggering $28 billion!!!!
I use simple trend following for my trading... less stress and less commissions because im not getting in and out all day long!
Hey! Thank you for the great webinar.
I can't wrap my head around the spread mean. Isn't the whole basis for this kind of trading to work that the spread mean is 0? With the spread men being (stock) A-nB, and n being μ_A/μ_B. How do you calculate the spread mean?
Regards,
PCT
PCT,
Thank you for your question. When we do the least squares calculation, we regress:
Y = b * X + e, where b is beta and e is epsilon. We are solving for the b that minimizes the e.
We are not trying to set the mean to be zero. It might sometimes be zero but often diverges significantly from zero. I hope that helps.
Jen