ASX Stock Pairs Trade Journal

[skc]

The market is iffy, the volatility is high. As market behaves less rationally, divergences are often and wide and this usually means a good time for pairs trading.

I am going to post my trades for the next couple of months as an attempt to revive this thread and get some discussion going.... the trades and prices are real, but the volumes are based on a account size of $100k.

 

[skc]

Trade #1 opened - Long CPA @ $0.985. Short DXS @ $0.945. Size = 15% of account for each leg.
Trade #2 opened - Long ALZ @ $2.54. Short WDC @ $9.26. Size = 15% of account for each leg.

Will post some charts later.

 

[skc]

Trade #3 opened - Long STO @ $11.39, Short ORG @ $12.78. Size = 15% of account for each leg.

 

[notting]

WPL and OSH are nicely spread against STO if you want to triple against one, for a bit more fun!
Long RIO Short BHP!?
Long PDN short ERA.
Long JBH short HVN.
Long MST short WOW.

 

[notting]

WPL and OSH are nicely spread against STO if you want to triple against one, for a bit more fun!
Long RIO Short BHP!?
Long PDN short ERA.
Long JBH short HVN.
Long MST short WOW.

Sorry should read MTS
You can add Short SXL Long SWM

 

[skc]

WPL and OSH are nicely spread against STO if you want to triple against one, for a bit more fun!
Long RIO Short BHP!?
Long PDN short ERA.
Long JBH short HVN.
Long MST short WOW.

PDN/ERA are off my watchlist these days as they are too small and volatile. I wouldn't long JBH against HVN because HVN is trading at NTA while JBH is trading like 8x NTA. MTS and SWM both had recent downgrades so no long trades on those for me.

RIO / BHP is the only decent candidate, but the market needs to turn up (probably mid this week) for it to be a winner imo.

 

[notting]

I bet they will all pay simply because it is so chaotic that sectors will move with macro sentiment details, will be neglected till things settle down and company sentiment dominates again.
I'm in the RIO BHP having shorted BHP a few days ago and slowly filling up the RIO long today.
Have small position in SWM and SXL which I will increase if they diverge further.
Have fairly full position on the WOW MTS one(MTS's relative dividend strength should give it support regardless of downgrade).
Waiting for JBH to stop being so extreme relative to HVN before entering that, but whatching.
Have also been short BPT short AWE which is still diverging, so adding to that as BPT tends to correct hard and you need to be there to get it.
Took you up on ORG against STO at the end of the day. Not a pair I would have put together with only 50% syncronicity and ORG utility element would give it relative strength!? But was feeling a bit happy!

 

[sinner]

No offense to you guys (and I don't wanna harp on this), but I can't believe people are still chasing barely cointegrated spreads which look like this:


when it only takes a little bit of effort to find (or build using regression) spreads that look like this:


and using signals like Boll(14) when you could be using kickarse signals like:
http://cssanalytics.wordpress.com/2009/07/29/differential-dv2-calculation/

These links have good ideas and setups:
http://www.tradingtheodds.com/2010/08/34826/
http://www.tradingtheodds.com/2010/08/pairs-trading-part-ii-spy-vs-rth/

 

[notting]

I probably would be doing what ever you are talking about if i understood it!
Regardless, I have a lot of fun with pairs.
It seems to hedg my psycology and seems to assist me often take profits on one then on the other a few days later, which is pretty cool.
Trading nakedly, unless it's system induced, seems to destort my psycology too much unless I just love the stock like COH.

 

[sinner]

I probably would be doing what ever you are talking about if i understood it!
Regardless, I have a lot of fun with pairs.

If you want to have an edge in statistical arb, it helps to have a good understanding of the statistics behind it. I didn't care about it while studying at uni, so had to actually teach myself from scratch.

Autocorrelation (can the time-series significantly predict itself?), cointegration (test for stationarity), multicorrelation (do any other time-series predict this time-series), etc.

It seems to hedg my psycology and seems to assist me often take profits on one then on the other a few days later, which is pretty cool.
Trading nakedly, unless it's system induced, seems to destort my psycology too much unless I just love the stock like COH.

Yah, market neutral is psychologically good. But don't let it fool you, make sure you are market neutral (hint, none of the spreads you chose are even close). I never see anyone in this thread even talking about adjusting for beta. skc is the exception of course, as he applies his regular bottom up and top down analysis to the pairs, so the regular rules don't apply. For the rest of us who don't have alpha coming out of our ears, we have to try and keep within the bounds of science.

That means looking for (or constructing yourself - absolutely nothing wrong with this and how they do it in the big boys world) a spread or pair which is at minimum weakly stationary, where the components exhibit low multicorrelation and the spread exhibits statistically significant short-term (1 moment lag) positive or negative autocorrelation.

 

[skc]

No offense to you guys (and I don't wanna harp on this), but I can't believe people are still chasing barely cointegrated spreads which look like this:

when it only takes a little bit of effort to find (or build using regression) spreads that look like this:

and using signals like Boll(14) when you could be using kickarse signals like:
http://cssanalytics.wordpress.com/2009/07/29/differential-dv2-calculation/

Thanks Sinner. What's the units for the vertical scale on the 2nd chart? Std Dev or something else?

Interesting signal constructed by DV. It is trading divergence over much longer timeframes (with 250-day lookback) than what I look at. You would be able to ride some larger mean reversals but it also entails holding longer and probably much subsceptible to copmany fundamentals. I will probably give it a play with some quicker signals.

The Boll(14) signal has actually not worked for many of my pairs for some time. Indeed, taking the signals as they come would have yielded -25% since Nov last year. I've been using a lot of discretion with my entries and exits - often with entries only after much divergence and much quicker exits than full mean reversals. It's a market for quick hits whether one is trading directional or pairs.

P.S. If one was to take offence to your post they are not interested in improving their trading...

 

[sinner]

Thanks Sinner. What's the units for the vertical scale? Std Dev or something else?

Units of ratio.

Interesting signal constructed by DV. It is trading divergence over much longer timeframes (with 250-day lookback) than what I look at. You would be able to ride some larger mean reversals but it also entails holding longer and probably much subsceptible to copmany fundamentals. I will probably give it a play with some quicker signals.

Actually it's the opposite dude. This indicator measures 2 day range/strength. It will spit out signals with a hold time <5 days. 250-day %rank is to normalise the data over a recent range (in this case normalise the data over 1Y), an insanely useful technique for normalisation. Almost all of David Varadis work uses %rank for normalisation.

The Boll(14) signal has actually not worked for many of my pairs for some time. Indeed, taking the signals as they come would have yielded -25% since Nov last year. I've been using a lot of discretion with my entries and exits - often with entries only after much divergence and much quicker exits than full mean reversals. It's a market for quick hits whether one is trading directional or pairs.

Like I said "skc is the exception of course"

 

[skc]

Units of ratio.

Sorry I am still unclear. What is the unit of the vertical scale on the 2nd chart? You mean that is the actual ratio? The image you've posted had the scale partially cut off.

Actually it's the opposite dude. This indicator measures 2 day range/strength. It will spit out signals with a hold time <5 days. 250-day %rank is to normalise the data over a recent range (in this case normalise the data over 1Y), an insanely useful technique for normalisation. Almost all of David Varadis work uses %rank for normalisation.

I am not sure that is correct. It's taking a calculated average of the C, H, L over the last 2 days, but the Percentrank function will rank it against the 250 day range. So for instance, ABC 2-day average is $6, while 250 day range is $5 to $15, it will have a DV2~10%. XYZ with the same range but 2-day average of $14 will have DV2 of ~90%.

It's basically calculating how the share has done in the last 2 days (on average) compared to how it's been travelling for the last year.

Do tell if I am have completely mis-interpreted it.

Like I said "skc is the exception of course"

Thanks. You managed to make me feel slightly better after being down 3% today.

BTW - I agree that with ETF you can (and probably have to) treat it entirely as a statistical exercise.

 

[sinner]

Sorry I am still unclear. What is the unit of the vertical scale on the 2nd chart? You mean that is the actual ratio? The image you've posted had the scale partially cut off.

That's the actual ratio. I cut off the scale because I don't want anyone identifying my spread based on its price.

I am not sure that is correct. It's taking a calculated average of the C, H, L over the last 2 days, but the Percentrank function will rank it against the 250 day range. So for instance, ABC 2-day average is $6, while 250 day range is $5 to $15, it will have a DV2~10%. XYZ with the same range but 2-day average of $14 will have DV2 of ~90%.

Err...sort of yes and no? ABC 2-day avg is $6, the 250day %rank is where that avg sits if it was ranked 0-100 against all the other 2 day avg in the set. It's not measuring 2 day against the the 250 day "range".

Not sure why you think that would generate "divergence over much longer timeframes". Plug the raw and %rank normalised into a spreadsheet, you will see how it works. The %rank function just provides a good "robust" adaptivity mechanism, to adjust based on current market conditions.

e.g.

"When the 2 day close/strength is within the top 30% of the last 250 days close/strength values then short." is the sort of logic it tries to express, instead of the usual "when the 2 day close/strength is above some arbitrary threshold X then short"....in one year the threshold of X might be good but next year the indicator may never reach X!

Thanks. You managed to make me feel slightly better after being down 3% today.

BTW - I agree that with ETF you can (and probably have to) treat it entirely as a statistical exercise.

Sweet, you owe me one, I'll hit you up tomorrow to make me feel slightly better when I do my monthly rotation in my super account (which has been 25% invested in XJO for the last month). :swear:

 

[skc]

That's the actual ratio. I cut off the scale because I don't want anyone identifying my spread based on its price.

Ok. Never seen ratio chart shown like that before. Looks like an awesome pair to trade.

Err...sort of yes and no? ABC 2-day avg is $6, the 250day %rank is where that avg sits if it was ranked 0-100 against all the other 2 day avg in the set. It's not measuring 2 day against the the 250 day "range".

Ok. Got it now. Didn't pay enough attention to the fact that it was working off strength. I thought it was just calculating some weighted function of the actual C, H & L prices.

Using this you will get very quick signals. Probably short anything that has two consecutive days closing on its highs...

Sweet, you owe me one, I'll hit you up tomorrow to make me feel slightly better when I do my monthly rotation in my super account (which has been 25% invested in XJO for the last month). :swear:

It's alright. Share market always goes up in the long run

 

[sinner]

Ok. Never seen ratio chart shown like that before. Looks like an awesome pair to trade.

Hint: ETF space is still littered with spreads that should have a theoretical correlation of 1, basically you can compete until a HFT adds it to their arb list. Here's a free one, which for example HFT arbs probably won't touch (if you can guess why then you will have no problem finding more): XRT:XLY (NYSE).

If the spread narrows more than it should relative to market vol, someone with big pockets has started to arb it, time to move on.

Using this you will get very quick signals. Probably short anything that has two consecutive days closing on its highs...

Which is very profitable in markets/spreads showing statistically significant negative 1 lag autocorrelation...something that pops up a lot in indices in high vol

I am sure you can see how it would work on the example spread I gave above, especially if you can trade intraday. You can increase efficiency by working the order hard instead of just splurging at market.

It's alright. Share market always goes up in the long run

Doing ok overall, just miffed 'cos the rules say I gotta rotate out right when the tea leaves say market is ripe for a rally. Still in SLF and IOO, at least for June.

 

[notting]

Great now SKC, since this is your thread, can explain that to me. Just pretend I'm 5 years old and very gullible. Your in jail, and I have control of all your money!
I'm sure others would benefit.

 

[skc]

Hint: ETF space is still littered with spreads that should have a theoretical correlation of 1, basically you can compete until a HFT adds it to their arb list. Here's a free one, which for example HFT arbs probably won't touch (if you can guess why then you will have no problem finding more): XRT:XLY (NYSE).

If the spread narrows more than it should relative to market vol, someone with big pockets has started to arb it, time to move on.

Which is very profitable in markets/spreads showing statistically significant negative 1 lag autocorrelation...something that pops up a lot in indices in high vol

I am sure you can see how it would work on the example spread I gave above, especially if you can trade intraday. You can increase efficiency by working the order hard instead of just splurging at market.

The correlation is great and it looks simply enough to take small bites out of it with some auto scripts. Unfortunately, I am a complete idiot when it comes to ETFs, the US market and writing scripts. It's also not possible for me to trade US intraday... OK it's possible but I don't want to live like a bat. I will switch over if my ASX pairs ever stop working for me (and if ETF industry hasn't blown up by then ).

Great now SKC, since this is your thread, can explain that to me. Just pretend I'm 5 years old and very gullible. Your in jail, and I have control of all your money!
I'm sure others would benefit.

What is it that you want explained?

Not really my thread - I just waste a lot of time on it...

 

[notting]

Short ABC Long BKN.
It's only 21% correlation but has crossed 4 times in 1 year - fairly safe on this large spread you'd think.
All my others are kicking butt so why not? (truely dangerous thought)

PS Don't worry about explanations SKC I'll bumble along.

 

[skc]

Short ABC Long BKN.
It's only 21% correlation but has crossed 4 times in 1 year - fairly safe on this large spread you'd think.
All my others are kicking butt so why not? (truely dangerous thought)

PS Don't worry about explanations SKC I'll bumble along.

Not a pair I have as BKN is bundled with mining services while ABC is more building materials. BKN is down a lot because it A). Is in mining services and B). Had a big profit downgrade of late. Usually when big fundamental news moved the share price a lot, I put the stock on ignore for some time - unless I think the market has it wrong.

Anyway.... there are always reasons to not take a certain trade I suppose.

P.S. Why did ALZ trade down to $2.19 today? And why was I not looking?