@Adam There might be some slight varieties in using pairs trading strategy. I just want to gather more information on their experience in trading. I think there are a lot of varieties in pairs trading and so there is hardly any standard way of doing it. According to your standard approach, we were supposed to construct beta neutral and dollar neutral portfolio. According to my reading, I would only agree with you that beta neutral seems standard but dollar neutral I barely seen it anywhere else. But I think each company has their own way of managing their money and so I think dollar neutral might be able to yield better hedging results. I even saw people using a simple approach to sum of squared deviations between normalized prices and yet the performance is outstanding. Like I said I think there are more approaches to dealing with pairs trading.

I think you are confused in your approach.

Performance does not indicate how returns are being generated and the risks assumed.In the first instance, if you are to quote performance, then you need to consider both 'gearing' and market risk. Simply having two positions -one long and one short means you have gearing, you may also be net long the market, and the positions may not be genuinely stationary (only appears to be in short windows whilst the trader keeps his job!).

Co integration is about reduction of integration, and more specifically in relation to stock prices a linear combination at levels of 2 (Engle Granger) or more (Johansen) become I[0]. The ECM is the glue that holds the series together, and provides the boundary for reversion which you then trade if you are trading this method.

There is no need if you are using co-integration and then play around with log return this, and log return that, and ADF this, and correlation this other thing. If you want simplicity, then simply ratio spread trade two related stocks, throw co-integration in the rubbish and be market neutral, that is $ Long Stock A, Beta Short Index, 1-Beta Long Cash, plus: $ Short Stock B, Beta Long Index, 1- Beta short Cash -> get the net positions and manage your hedge book accordingly.

I suggest you get another source of 'reading', because it is totally wrong in respect to market and $ neutrality.

I would suggest you don't have enough experience even to have an opinion at this stage, and you better off to go to University and study this subject properly as originally suggested.

Love to know the company that employs you - maybe they need a mentor, or I can at least trade against them :)

@Adam

One paper I refer to is this one: Pairs Trading: Performance

of a Relative-Value Arbitrage Rule

I don't have a lot of opinion on this subject. I am simply learning and I thought you were the one having big opinions about it..

Respectfully, you are trying to enter an area that is highly congested--and more importantly--one that has most people are analyzing the same way. That is not a way to be highly successful.

If you want to trade pairs, you need to stop thinking like the herd and doing what the herd does.

I can almost guarantee that Renaissance is not trading pairs the way people are discussing here--that includes traditional thinking about co-integration, correlation and even the most hallowed assumption about pairs.

@cheng @Adam,

Hi Cheng, I agree with Adam's opinion. The purpose of finding a co integrated pair of stocks is to find stationary process so that you can use mean reverting strategies to trade it. It does not guarantee market risk neutral or dollar neutral(in fact you won't get it in many cases). The reason you need adf test is because regressing non stationary time series(like most of the price time series) in many cases give you spurious result. Co integration can be thought as an exception. The book below discussed in depth how co integration works(however it is not in a trading context). It not a cheap book but will give you a good understanding of the concept.

Co-Integration, Error Correction, and the Econometric Analysis of Non-Stationary Data (Advanced Texts in Econometrics)