Pair Trading Strategy Journal

tatankas · Jun 23, 2009

Don,

Imagine the following scenario:
Stock A beta is 1
Stock B beta is 2

You short 1k stocks for stock A and long the same amount for stock B.

The market goes down 15% in 10 days, most probably your long side will lose more than your short side, since the beta difference.

You are in fact losing the market neutrality, just hoping the market goes up.

There are 2 options, or you adopt a beta neutral strategy (not what I am doing), or you look for stocks with similar betas.

The key point here, is to use the appropriate beta, taking in consideration your time-frame. Most probably do not make sense to use a beta calculated using 3 year of data, when our time-frame is just a few days.

That's the way i see it.

Edit: I think it's possible to have highly correlated stocks, with different betas, so relying only in correlation should not grant us market-neutrality, the most important factor in a pairtrading strategy.

waltbx · Jun 23, 2009

Tatankas,

What you are saying is making some sense to this little head of mine. However, as I understand what you are saying, the online sites that present beta data are looking back 36 months. You suggest we look back a shorter period. Where/how is that shorter period beta data available?

Walt B

Quote from tatankas:

Don,

...The key point here, is to use the appropriate beta, taking in consideration your time-frame. Most probably do not make sense to use a beta calculated using 3 year of data, when our time-frame is just a few days.

That's the way i see it.

Edit: I think it's possible to have highly correlated stocks, with different betas, so relying only in correlation should not grant us market-neutrality, the most important factor in a pairtrading strategy.

cooper1308 · Jun 23, 2009

If the correlation is very high, chances are the beta's will be extremley close anyway.... By default correlation acts as a good Beta filter..... i.e Theoretically if correlation is at 100% beta's would be exactly the same......If the Beta's are out of whack, you are probably comparing correlation and Beta over a different timeframe...

I can't see the use of comparing either of these stocks to a basket of other unrelated stocks?We are trading the relationship between these 2 individual stocks..That's it.....Not the relationship of the stocks compared to the rest of the market.

For me this just further complicates a relativley uncomplicated system.

Each to their own.. if it adds further conviction to your trading go for it!...

Quote from tatankas:

Don,

Imagine the following scenario:
Stock A beta is 1
Stock B beta is 2

You short 1k stocks for stock A and long the same amount for stock B.

The market goes down 15% in 10 days, most probably your long side will lose more than your short side, since the beta difference.

You are in fact losing the market neutrality, just hoping the market goes up.

There are 2 options, or you adopt a beta neutral strategy (not what I am doing), or you look for stocks with similar betas.

The key point here, is to use the appropriate beta, taking in consideration your time-frame. Most probably do not make sense to use a beta calculated using 3 year of data, when our time-frame is just a few days.

That's the way i see it.

Edit: I think it's possible to have highly correlated stocks, with different betas, so relying only in correlation should not grant us market-neutrality, the most important factor in a pairtrading strategy.

Don87109 · Jun 23, 2009

Quote from tatankas:

Don,

Imagine the following scenario:
Stock A beta is 1
Stock B beta is 2

You short 1k stocks for stock A and long the same amount for stock B.

The market goes down 15% in 10 days, most probably your long side will lose more than your short side, since the beta difference.

You are in fact losing the market neutrality, just hoping the market goes up.

There are 2 options, or you adopt a beta neutral strategy (not what I am doing), or you look for stocks with similar betas.

The key point here, is to use the appropriate beta, taking in consideration your time-frame. Most probably do not make sense to use a beta calculated using 3 year of data, when our time-frame is just a few days.

That's the way i see it.

Edit: I think it's possible to have highly correlated stocks, with different betas, so relying only in correlation should not grant us market-neutrality, the most important factor in a pairtrading strategy.

Thanks, I see your point. Although I think that correlated pairs tend to dampen the beta differences, you are probably correct that it still has a some effect.

However, a few years back I remember trying to beta weight a portfolio and after a few months I decided that beta was terribly inaccurate. At the time, the sentiment from other traders I knew was that beta figures were not useful at-least for short term purposes. This probably re-affirms your statement that it: "does not make sense to use a beta calculated using 3 years of data".

I wonder if option Implied Volatility is a better gauge than beta to adjust market neutrality. True IV is not related to market movement, rather it's just a measurement of the stock price movement relative to itself. Still, my experience says that higher IV stocks will move further than lower IV stocks relative to the market (barring a news event). And IV might be a better predictor because it attempts to forecast future volatility, which is what we need -- as opposed to beta's historical perspective. Furthermore, IV figures are influenced by option betting which means money is at stake perhaps helping IV accuracy . I'll probably experiment with this idea going forward.

Don

yobo · Jun 23, 2009

Don't worry about beta. Gets things too complicated. I will be curious to see how this works out for you. I don't use PTF, but have my own spread sheets. Not sure what criteria you are using but I looked at the following:

200 day correlation: .67

Trading well within the 200 mean 2 stdev bands and also within the 13 day mean bands.

200 day average pair ratio average is .77 which will offer you some support, but the 200 day average is trending lower. Not good for going long.

The 13 day pair ratio average is .949 and also trending lower. The upper band is 1.08 and the lower band is .816. At today's close the pair ratio is still inside the 13 day bands closing at .851.

For me, ideally I would want to see the pair ratio deviate 2x from the 13 day mean before entering a trade to go long...somewhere between .77 and .816.

At this level, I would expect a snap back to the 13 day mean of .94ish. Thus the profit potential is around 16% on one side.

Hope it works out for you. I'll join you on the trade if my scenario pans out.

Quote from Don87109:

Thanks, I see your point. Although I think that correlated pairs tend to dampen the beta differences, you are probably correct that it still has a some effect.

However, a few years back I remember trying to beta weight a portfolio and after a few months I decided that beta was terribly inaccurate. At the time, the sentiment from other traders I knew was that beta figures were not useful at-least for short term purposes. This probably re-affirms your statement that it: "does not make sense to use a beta calculated using 3 years of data".

I wonder if option Implied Volatility is a better gauge than beta to adjust market neutrality. True IV is not related to market movement, rather it's just a measurement of the stock price movement relative to itself. Still, my experience says that higher IV stocks will move further than lower IV stocks relative to the market (barring a news event). And IV might be a better predictor because it attempts to forecast future volatility, which is what we need -- as opposed to beta's historical perspective. Furthermore, IV figures are influenced by option betting which means money is at stake perhaps helping IV accuracy . I'll probably experiment with this idea going forward.

Don

CBuster · Jun 24, 2009

Quote from cooper1308:

If the correlation is very high, chances are the beta's will be extremley close anyway.... By default correlation acts as a good Beta filter..... i.e Theoretically if correlation is at 100% beta's would be exactly the same......If the Beta's are out of whack, you are probably comparing correlation and Beta over a different timeframe...

the above statement has a few inaccuracies. stocks with similar and high betas do tend to be fairly correlated. however, if correlation is at 100%, it does not mean the betas would be exactly the same.

to illustrate, imagine stock A always moves the same amount as the market (eg SP500) and stock B always moves 2x the market in the same direction. hopefully we can all agree that stock A beta is 1.0 and stock B beta is 2.0. the correlation of these two stocks is 1.00 (i.e. perfect correlation)

Ms Varima-Garch · Jun 24, 2009

fyi, a good overview of the pairs trading process.
FREE !!!!

www.tradingwithmatlab.com/pairstrading.html

cooper1308 · Jun 24, 2009

Quote from CBuster:

the above statement has a few inaccuracies. stocks with similar and high betas do tend to be fairly correlated. however, if correlation is at 100%, it does not mean the betas would be exactly the same.

to illustrate, imagine stock A always moves the same amount as the market (eg SP500) and stock B always moves 2x the market in the same direction. hopefully we can all agree that stock A beta is 1.0 and stock B beta is 2.0. the correlation of these two stocks is 1.00 (i.e. perfect correlation)

Yeah that is true... I should've elaborated...

The above scenario could happen, however, In a pair trading world the ratio chart (probably the most important) would be trending...(Taboo for most pair traders)... If one stock is continually moving twice the other, I can't see this a good pair to trade as there is an obvious change in the underlying fundamentals.

Correct me if i'm wrong, but if a pair is oscillating around a mean, then a high correlation chart would have to result a close beta comparison.. A ratio can't can't stay the same if beta's are different?

CBuster · Jun 24, 2009

Quote from cooper1308:

Yeah that is true... I should've elaborated...

The above scenario could happen, however, In a pair trading world the ratio chart (probably the most important) would be trending...(Taboo for most pair traders)... If one stock is continually moving twice the other, I can't see this a good pair to trade as there is an obvious change in the underlying fundamentals.

Correct me if i'm wrong, but if a pair is oscillating around a mean, then a high correlation chart would have to result a close beta comparison.. A ratio can't can't stay the same if beta's are different?

mostly agree. however, the ratio chart will only be trending due to a beta differential if the market is also trending. if the market is chopping sideways the ratio could be too, even if the betas are very different.

besides all this, i only look at beta as a side issue. as a couple of you guys have pointed out, beta is not particularly robust - it is too prone to change. the 1 year will look differnt to the 3 month which will likely be different to the 1 month, etc. i will often check the beta to look for anything unusual, but that's all.

Ms Varima-Garch · Jun 24, 2009

Quote from Ms Varima-Garch:

paper trade, visualization-based. no make look at fundamental or news

detected spike in novartis on the finviz market map 'radar screen'

NVS -239@$42.002 (total position $10K) Novartis, Switzerland

EWL +542@$18.46 (total position $10K) Switzerland ETF

rationale:

4% spike in novartis (daily standard deviation about 1.3%)
high correlation ~87% between novartis and ewl
short major component stock against country etf

ok this pair trade has stabilized at about -$400, and the passage of time has not cured this. it's a paper trade, i'm closing it for a loss of about $400.

possible explanations:

1) i have NOT looked at the ratio of the pair (it's stupid, i know, i know, don't have the software yet). so i did not know where the ratio was relative to its historical levels, whether the pair is trending etc etc

this was ONLY finviz-based. i guess this means if you only look at one stock (instead of the pair in its entirety), you better trade that one stock.

a standalone short Novartis trade would've made some money, but the switzerland etf lost much more, maybe because of problems in the europe etc

2) various etf pairs strategies probably work better in the commodities complex, e.g. USO/XLE, GDX/EWC, ABX/EWC etc. Basically oil, gold, natural resources IGE, EWC, EWA all those various combinations. BHP

in any event, i am deeply disturbed by this and i think i need to bring myself back into a centered state of mind. i am going to do some breathing exercises and tai chi.