Talk:Pairs trade

Should have, would have, could have

You should experience a gain on the short position and a negating loss on the long position, leaving your profit close to zero in spite of the large move but if pair eventually converges then you would end up with zero profit even with the whole stock market going up. Or am I missing something here? --Anon. Wikipedia User

19th or turn of the century invention?

I remember reading in a Financial Times pull-out I think, about some trader who used this method long ago, probably when the ticker-tape or telegraph had just come into use. The trader made and lost a few fortunes, and eventually comitted suicide. So its wrong to write that it was invented in the 1980s. 92.29.113.54 (talk) 20:43, 11 December 2009 (UTC)

Model-based pairs trading

While it is commonly agreed that individual stock prices are difficult to forecast, there is evidence suggesting that it may be possible to forecast the price - the spread series - of certain stock portfolios. A common way to attempt this is by constructing the portfolio such that the spread series is a stationary process. To achieve spread stationarity in the context of pairs trading, where the portfolios only consist of two stocks, one can attempt to find a cointegration relationship between the two stock price series . This would then allow for combining them into a portfolio with a stationary spread series . Regardless of how the portfolio is constructed, if the spread series is a stationary processes, then it can be modeled, and subsequently forecasted, using techniques of time series analysis. Among those suitable for pairs trading are Ornstein-Uhlenbeck models , autoregressive moving average models and (vector) error correction models . Forecastability of the portfolio spread series is useful for traders because:

1. The spread can be directly traded by buying and selling the stocks in the portfolio, and
2. The forecast and its error bounds (given by the model) yield an estimate of the return and risk associated with the trade.

According to a well-known paper of B. Do, R. Faff and K. Hamza, the success of pairs trading depends heavily on the modeling and forecasting of the spread time series.

[OLD ENTRY FOR THIS SECTION BEGINS HERE]

While it is commonly agreed that individual stock prices are very difficult to predict, there is considerable evidence suggesting that linear combinations of them may possess a degree of predictability. Now a pairs trade is equivalent to buying or selling the spread, which is a synthetic asset constructed as a trivial linear combination of the two trade stocks, say X and Y, involved in the trade: Long (+) X and short (-) Y. Consequently, the spread time series can be subjected to statistical mean-reversion models and forecasting. Further, these spread forecasts, if accurate, are very useful, because the spread itself can be traded.

In one modeling approach, the spread series is first detrended. The (presupposedly mean-reverting) detrended spread is then fitted a low-order autoregressive moving average (ARMA) model, which has optimal predictive capabilities on the training set . Since a pairs trade is equivalent to buying or selling the spread, the model thus yields the expected trade outcome as well as the best-case and worst-case scenarios for each time instant in the forecast horizon. This information can be used to manage the risk associated to a pairs trade.

Comment: The above section describes and illustrate a scientifically sound approach to pairs trading. I wonder if it could be included in the main article on pairs trading? Of course, the link to the image should be replaced by the actual image, and references should be listed in full.

Pairstrader (talk) 16:06, 26 November 2010 (UTC)

First, we don't link to external images or advertise "Courtesy of" some commercial site. You will not be able to include any links to iQfront. It doesn't qualify as a reliable source and also doesn't meet our external linking policy. I've removed the link from this page as well.

Second, I see you have cited the first sentence of each paragraph. Where does the rest of the material in each paragraph come from? If it is not also explicitly in the cited source, then what you've written appears to be original research or synthesis. Yworo (talk) 19:39, 26 November 2010 (UTC)

Yworo (talk) 19:39, 26 November 2010 (UTC)

Thanks for your feedback, Yworo. As I indicated above already, it was not my intention to link to iQfront, but to cite the original source of the image. I am not able to upload images directly yet, but I do think that the image I provided would illustrate the rather theoretical discussion (second paragraph) on spread models and forecasts.

Regarding my citations: That prices of linear combinations (e.g. long X and short Y = pair spread) of stocks possess predictability is fairly common knowledge among traders; the reference I provided is one recent resource in this respect. The rest of the first paragraph elaborates on the beneficaility of the forecastability of the spread for pairs traders. The second paragraph describes one modeling approach which is based on a scientifically sound method (the second reference), and explains why modeling the spread is interesting for pairs traders (i.e. because the spread, which is being modeled and forecasted, can be traded directly by buying and selling the stocks that make up the spread).

Pairstrader (talk) 08:05, 27 November 2010 (UTC)

Well, first, if that image belongs to iQfront, it can't be uploaded as they own the copyright. Unless they have released it into the public domain or licensed it under a free-content license such as GFDL or Creative Commons.

Second, I looked at the one article that was accessible, and nowhere in it did I see any statement to the effect that "there is considerable evidence suggesting that linear combinations of them may possess a degree of predictability". We can't interpret articles: if we make a statement in Wikipedia, that statement must be in the source in some form. I suspect the same deficiencies in the second citation. Interpreting the sources rather than reporting what they say is original research. Putting two sources together to support ideas not explicitly in either of them is synthesis. Do you have any sources that directly state what you are trying to convey? Yworo (talk) 22:15, 27 November 2010 (UTC)

To my knowledge, the image is released into the public domain - do I need to submit a written permission? Regarding the references I must emphasize this: That stock prices cannot be (profitably) forecasted is just the Efficient Market Hypothesis. That assets can be combined into portfolios (with long and short positions) which may possess predictability is encapsulated in this statement of the first paper: "These tools help to discover new synthetic instruments (expressed as spreads of existing instruments) that have more predictable time series with a typical mean-reverting feature and not directly observed by the majority of other market participants." Maybe I should refrain from using the term "linear combination" as it seems to be the source of confusion.

Pairstrader (talk) 12:52, 29 November 2010 (UTC)

iQfront has a clear copyright notice at the bottom of the main page. Not sure how you navigated to that image, but it would be covered by the site copyright. I see no reason to assume its been released into the public domain. There is in fact no need for a copyright notice, all works are automatically copyrighted at the time of creation. Only an explicit statement can release them into the public domain. Since you say you think it is so released, simply provide the link to that statement. If you don't know of such a statement, it's not in the public domain. Getting permission would not be sufficient: you would have to get them to post a statement on their website either releasing the image into the public domain or under a free-documentation license as mentioned above.

The source of the confusion is not "linear combination", it's "considerable evidence". What you've got is a single research paper, so it should be presented as "so-and-so" say, describe, have theorized, or some similar language. I don't think we've got a "proof" here and I'm sure it doesn't always work out as theorized.... Yworo (talk) 07:53, 1 December 2010 (UTC)

Yworo, I did an extensive literature review on the topic and rewrote the whole section (see Model-based pairs trading above until the text [OLD ENTRY FOR THIS SECTION BEGINS HERE]). The references are now more to the point. As concerns a picture illustrating an ARMA spread forecast, I have been in contact with iQfront, but have not yet received any confirmation from their side. When may I be able to upload any picture, by the way? Pairstrader (talk) 19:05, 11 December 2010 (UTC)

If you were to create your own image, you could upload it at any time. For other images, iQfront would have to publish the image on a page giving the license terms; otherwise, you need to follow the instructions on this page. Yworo (talk) 14:55, 12 December 2010 (UTC)

Oh, the rewrite looks very good to me. Much superior to what's in the article now. Go ahead and implement it. Be sure to use full citations for the refs you just have as "research paper". Yworo (talk) 14:58, 12 December 2010 (UTC)

Ok, I will do it soon. Thanks for your constructive feedback. I also received a confirmation email from iQfront that they are about to publish some screenshots "for educational purposes" on their site, with public domain terms. Pairstrader (talk) 15:42, 12 December 2010 (UTC)

Great, that will solve the problem. Yworo (talk) 16:16, 12 December 2010 (UTC)