Intraday Pair Trading/Stat Arb

Quote from Papa Smurf:

Hi RK. I have nothing to sell. It's just an invitation to have an exchange of ideas that goes beyond what can be handled in a chat room format.
More...

I think you have a fair point about getting some mutual reciprocity from serious parties who have something tangible to share. IMO automated pairs trading and arbitrage is a very compelling strategy with broad participation but with piss-poor discussion. Not a sharing, loving feeling perhaps.

Maybe choose and post an example that you feel is compelling and worthy of consideration but preserves for you the core IP of your strategy. You might just be pleasantly surprised where the discussion leads.
 
Bone, thanks for your post.

The offer still stands for discussion offline. We've got a long list of statistical methods some of which we've tried already and others we haven't.

Best regards
 
Hi:

There is an older article from TASC about pairs trading. In the article they define the spread as

Spread = (LastA/CloseA) – (LastB/CloseB)

They then define the volatility band as
VB = (HV1 + HV2)*(1/252)*(1-R)

where R = correlation between the two stocks
HV1 = historical volatility of stock 1
HV2 = historical volatiltiy of stock 2

They then make the usual bands

Spread +/- 2 VB (or so)

I have searched google but can't find where the formula for VB comes from-does anyone know?

Thanks!
 
I though this was posted but didn't see it up yet........sorry if this is a repeat.

I have a math question from the following TASC paper on pairs trading:

http://store.traders.com/v20526daysto.html

They define the spread as
Spread = (LastA/CloseA) – (LastB/CloseB)

Then define the volatility band as

VB = (HV1 + HV2) * sqrt(1/252)*(1-R)

where
HV1 = historical volatility of stock 1
HV2 = historical volatility of stock 2
R = correlation of the two stocks

I've searched google with no luck-can anyone explain where the VB equation comes from?

Thanks!
 
Quote from bennfine:

I though this was posted but didn't see it up yet........sorry if this is a repeat.

I have a math question from the following TASC paper on pairs trading:

http://store.traders.com/v20526daysto.html

They define the spread as
Spread = (LastA/CloseA) – (LastB/CloseB)

Then define the volatility band as

VB = (HV1 + HV2) * sqrt(1/252)*(1-R)

where
HV1 = historical volatility of stock 1
HV2 = historical volatility of stock 2
R = correlation of the two stocks

I've searched google with no luck-can anyone explain where the VB equation comes from?

Thanks!
More...

He's just taking the summed volatility of the two stocks, adjusting from annual to daily volatility (252 trading days a year), and reducing it ( factor of 1-R) for highly correlated stocks to get a volatility band.
 
Quote from comintel:

He's just taking the summed volatility of the two stocks, adjusting from annual to daily volatility (252 trading days a year), and reducing it ( factor of 1-R) for highly correlated stocks to get a volatility band.
More...

Good description. FWIW I backtested many variations of that pair trading "system" outlined by Cowan and couldn't get it to be profitable in any of the pairs that I tried it on. I tried the more obvious pairs, picked pairs based on fundamental factors, stocks with common fundamental inputs but different industries etc etc.

IMO you need something a bit more robust than the system you are looking at!
 
Quote from comintel:

He's just taking the summed volatility of the two stocks, adjusting from annual to daily volatility (252 trading days a year), and reducing it ( factor of 1-R) for highly correlated stocks to get a volatility band.
More...

Thanks....is there any theory behind it?

Var(X-Y) = Var(X) + Var(Y) - 2Cov(X,Y) and all that, but their VB formula seems ad hoc; or at least I can't find it justified anywhere.

Do people have a preference for working with X-Y or X/Y when pairs trading?

Thanks!
 
Quote from bennfine:

Thanks....is there any theory behind it?

Var(X-Y) = Var(X) + Var(Y) - 2Cov(X,Y) and all that, but their VB formula seems ad hoc; or at least I can't find it justified anywhere.

Do people have a preference for working with X-Y or X/Y when pairs trading?

Thanks!
More...

Good point - I do not know. Maybe he tried various things and went with what seemed best. Maybe you can email him.

Ratio should definitely be better.

A lot of people say cointegration is a better screen than correlation for pairs.
 
Hi everyone,

I am looking for recommendations for good pair trading execution software. (Just to be clear not looking for a pair trade finder.) We are using simple software right now. I am looking for specifically something has order slicing functionality built in and various types of fills for the second leg so that we get good execution without paying through the nose. We have found we are not getting best execution right now. We don't want to write our own code if we don't have to.

Likewise if someone knows a coder who might be able to help and has experience coding trading systems please get in contact.

Any suggestions are appreciated.

Best regards
 
You must log in or register to reply here.
Back
Top