Mean Reversion Model

[Pattym]

Hello Everyone,

I am a 2nd year B.A Economics, specialization in Quantitative and Mathematical Economics major at Carleton University in Ottawa, Canada.

Was hoping for more information on mean reversion models that would be toned down to the basic level. I searched the forum, couldn't find much, and google didn't help at all.

I am looking to test a few models in FX markets on a very short time scale (under 3 minutes) during times of high volatility (markets opening/closing and economic reports).

From what I understand basic system is that you create your mean value, and trade long/short as it goes to x standard deviations.

What I am thinking is what if in the long run, if you know which way the market is trending, you can use the mean reversion system to build a position that controls the size of the position by buying when it's 'undervalued' and selling on the 'overvalued' side of the system.

Am I off keel by thinking this is plausible? If yes, can someone point me in the right direction? If not, please point me in the right direction.

 

[Ken Abbott]

You're not "off keel", but you're several decades too late. People have been looking at things like this for a while - mean reversion, cointegration, all of it.

Start with basic ARIMA models to get a feeling for the notion of "memory" in a time series. I also have some basic material on mean reversion from my classes that I can send you.

 

[Lyosha]

(markets opening/closing and economic reports)

What's your investment thesis?

Am I off keel by thinking this is plausible?

Depends on your answer to the above.

 

[Pattym]

What's your investment thesis?


Depends on your answer to the above.

I'm looking to see how the basic model works first. Secondly, I think that high volatility events create discounted prices short term against the long term trend.

 

[Lyosha]

Well the "basic model" works by purchasing assets at a discount, and short selling at a premium. But that wasn't my question. My question is (and I don't mean to sound snobby, but I'm not sure how to better phrase this) - what makes you think it would work at all.

... and why do you think high volatility events create discounted prices? Is there is such a thing as a "long term trend"?

 

[Ken Abbott]

calculating a mean reversion on anything finance related is way complicated, if not downright impossible, to calculate.

I disagree. Estimating speeds of mean reversion is easy. Profiting from said estimates is hard.

 

[Lyosha]

So long as we're giving links to papers, might as well toss the elephant out into the room...

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1153505

 

[Pattym]

Well the "basic model" works by purchasing assets at a discount, and short selling at a premium. But that wasn't my question. My question is (and I don't mean to sound snobby, but I'm not sure how to better phrase this) - what makes you think it would work at all.

... and why do you think high volatility events create discounted prices? Is there is such a thing as a "long term trend"?

Markets have equilibrium values both in long run and short run. In the long run, this equilibrium value is created by fundamental data, but in short run it is made by the actions of individuals within the market. These individuals do not always effect the long term equilibrium, but can effect the short term. This causes mispricing in short term in terms of the long term equilibrium.

My thinking is that when economics reports come out, people act irrationally in terms of the long term equilibrium, and multiple mispricings are created on a very short period of time.

Is that a better answer?

 

[Lyosha]

Markets have equilibrium values both in long run and short run. In the long run, this equilibrium value is created by fundamental data, but in short run it is made by the actions of individuals within the market. These individuals do not always effect the long term equilibrium, but can effect the short term. This causes mispricing in short term in terms of the long term equilibrium.

I suppose it puts the rest of your ideas into context.

If you believe in theoretical economics perhaps your ideas make sense in your mind. But just like in the case of philosophy and biology, there is a rather massive disconnect between economics and the real world.

The reason why I asked you these questions is because your suggestion is commonly known among people who put these kinds of trades on as a great way to blow up. Do yourself a favor and before you invest any real money into anything take some time and purge yourself of any ideas about how you think markets should work in theory. In the real world everything is much more complicated than some cheesy economic models.

Take for example a slightly less cheesy economic model - I have yet to see you mention drift.

 

[Pattym]

Actually have not learned a single way to apply what I am learning yet, so came up with the idea on my own.

That's why I came here, to learn how to apply what I am learning. Anyone willing to point me in the right direction is welcome to help. The only problem is that real world problems use stochastic calculus, and I can't take that yet. Hence why I feel dumbed down models, back tested to prove validity is a good start to where I want to be.

 

[Lyosha]

google avellaneda and lee. Their 2008 paper. It's a good starting point if you want to learn about mean reversion.

 

[Raden Ananta]

Hello pattym,

A friend of mine, a trader, sometimes use mean reversion model in fx trading. But it only works at spesific situation, not every time. I am afraid if you do that in time frame under 3 minutes, it is very difficult to determine unbiased mean with 95% degree of confidence. Also, I think we have to fit the distribution to assure the mean is not changing (avoiding error type I,II. We could have wrong entry price and take profit for that error).

In my point of view, economic report released has no effect to market current situation because what being told from report is what was happening last month, not this month. Seems you are going to be a news trader :D


Cheers,

RA