ASX Stock Pairs Trade Journal

[SilverRanger]

Ib has moc order but they must be submitted 15 mins before close for the US market and withdrawal is not allowed right? !5 mins implies more significant price changes.

There probably is, but I have never used it myself (not that I need to). What is more handy from IB is the Pair Combo, simply buy or sell the pair at limit or market (or LMT + MKT), use that to lock in your profit with trailing stop.

Have not study intraday signals but do chart them. Some diverge more than few trading sessions while others revert during the trading session. SilverRanger how many trading session do you hold the position given that the price diverge and did not revert on the trading session?

Same treatment as any other EOD signals, if it diverges further enough for absolutely no reason, I might consider layering/averaging in

 

[skc]

I don't normally trade indices, but I think this warrants a trade. An 8% gap pretty much explained by the $A rise. My exit would be a 4% closing up of the gap.



IG markets provides a A$ denominated contract for the US indices...so that takes care of any currency issues.

 

[fingl]

Here are some equity curves with comparison to with and without correlation for trade entries.

With correlation rules adding on, the equity curve become more losing and max dd become more. 1) No correlation 2) Correlation 30 > 0.6 3) Correlation 90 > 0.6 4) Correlation 180 > 0.6

Why the equity curve is performing lacking with more correlation adding on? Isn't correlation important? Anyone has done a mass data backtesting statistics on whether correlation matters?

 

[skc]

Here are some equity curves with comparison to with and without correlation for trade entries.

With correlation rules adding on, the equity curve become more losing and max dd become more. 1) No correlation 2) Correlation 30 > 0.6 3) Correlation 90 > 0.6 4) Correlation 180 > 0.6

Why the equity curve is performing lacking with more correlation adding on? Isn't correlation important? Anyone has done a mass data backtesting statistics on whether correlation matters?

Interesting. Here's my take on correlation...

Correlation matters because of the underlying premise of pairs trading - i.e. two shares that are influenced by same factors diverge for no apparent reason. However, correlation needs to breakdown during the divergence / convergence process. If share A goes up 1% each day and share B goes up 0.8%, their correlation will be high (0.8), but this pair certainly is not profitable, even though a signal may eventually be generated.

That might explain why by taking trades with long periods of good correlation (180 > 0.6), you end up trading more A & B type set up which results in poor profitability.

Here's a screen shot of a great pair...

Notice the peaks and troughs (green verticle lines) on the ratio chart (i.e. when they are most diverged) correspond to the lows on the correlation chart. Notice also the peaks in the correlation chart (red lines) correspond with not a lot of divergence from the mean. So putting on a trade when correlation is low can yield better results, provided that correlation is normally good.



BTW the new software is coming out with co-integration calculated which is going to be fantastic...

 

[fingl]

So putting on a trade when correlation is low can yield better results, provided that correlation is normally good.

Why are most of the traders and Jared taking trades with correlation at least 0.7 or better with taking trades on low correlation yielding better results? Most of Jared trades have correlation more than 0.8. skc do you have correlation rule for your entries?

There are some equity curves improved with correlation rules and big loses are not taken while others have more big loses with correlation rules similar to the jpeg posted. Should pairs be curve fitted to have correlation rules on entries depending on whether the rules improve the equity curve?

 

[skc]

Why are most of the traders and Jared taking trades with correlation at least 0.7 or better with taking trades on low correlation yielding better results? Most of Jared trades have correlation more than 0.8. skc do you have correlation rule for your entries?

There are some equity curves improved with correlation rules and big loses are not taken while others have more big loses with correlation rules similar to the jpeg posted. Should pairs be curve fitted to have correlation rules on entries depending on whether the rules improve the equity curve?

I don't have a fixed correlation rule. I like to think that I address that by only pairing stocks that are fundamentally similar. With an example like above, you can definitely see how you don't need correlation to trade that ratio chart for mean reversion.

Play around with the signal timeframe and the mean. Look for a ratio chart that is range bounded and zigzags a lot, and find a mean line that plots close to the middle - and you have found a decent reliable pair. Correlation doesn't actually come into it.

 

[fingl]

skc, you use a lot of discretionary in your pairs trading do you? Finding the right ratio chart and picking the right signal. The pair you post is the perfect pair to trade. Most of the ratio charts are either a little upward or downward oscillating. They are in a trading range mostly for a while before shooting to new high or low in a furious manner. How do you protect against taking trades in trading range but is going to take out the high or low of the trading range?

Do ratios from different sectors tend to move together? Which sector is most suit for pairs trading and what are the reasons?

 

[skc]

skc, you use a lot of discretionary in your pairs trading do you? Finding the right ratio chart and picking the right signal. The pair you post is the perfect pair to trade. Most of the ratio charts are either a little upward or downward oscillating. They are in a trading range mostly for a while before shooting to new high or low in a furious manner. How do you protect against taking trades in trading range but is going to take out the high or low of the trading range?

Do ratios from different sectors tend to move together? Which sector is most suit for pairs trading and what are the reasons?

Yes I use a lot of discretion and I don't always trade at the signal price - I tend to wait and back my market reading skill to hopefully get an extra tick here or there for better entry and exit. Doesn't always work out but I see it as good practice...

You will get ratios that blow out when you are in a trade. If that happens with news then I exit immediately taking the loss. If it happens without news then my hard stops of 2% or time-based stop will be hit eventually.

Here's a good example. A great pair to trade until EQN was the subject of a takeover. Fortunately there was no signal before the gap up but I could easily have been in a trade as well.



Different sectors offer more or less opportunities over different times, but that's mostly due to spates of fundamental news and/or corporate activities. E.g. I traded a lot of media stocks last year but I probably haven't had one on for 6 months - too many downgrades going on.

It's best to just trade the sector you are familiar with and avoid sectors that throws out big swings... e.g. trading mid-cap miners will always expose you to an exploration risk so I tend to keep my position size smaller. REITs on the other hand rarely have single piece of company specific news that move the ratio out of wack too much. So they are better stocks to trade. Similarly with infrastructure and utilities.

 

[luckyforteja]

Hi All,

Gone through the whole journal, very informative. I have couple of questions
1) What is the best instrument for Pair trading, mainly for shorting?
2) What would be minimum amount that need to be investing for each leg of the pair?

I have downloaded the trail version of pair trader and tried CFD(both long and short), with a leg size of 2000 and realised that brokerage was killing me.

Cheers,
Sri

 

[SilverRanger]

Hi All,

Gone through the whole journal, very informative. I have couple of questions
1) What is the best instrument for Pair trading, mainly for shorting?
2) What would be minimum amount that need to be investing for each leg of the pair?

I have downloaded the trail version of pair trader and tried CFD(both long and short), with a leg size of 2000 and realised that brokerage was killing me.

Cheers,
Sri

1) For ASX it would be CFD for the availability and leverage, although IB is starting to offer share shorting (but only limited to 20 or so stocks), so you can save a bit of brokerage there
2) Depends on the liquidity of the stock and your provider, this can range from 5% to up to 100% of your leg size, but on average about 15% - 30% (50% for IB)

With a leg size of 2000 you are effectively paying 0.5% per trade, so that's 2% in total!! I would be happy enough if I can break even with that amount of brokerage costs
But if that's the amount you are willing to risk, I think you are better off starting in the US market, where brokerage cost isn't generally an issue like the ASX. The flip side - sleep 1 am at night or wake up 6 am in the morning !!

 

[luckyforteja]

1) For ASX it would be CFD for the availability and leverage, although IB is starting to offer share shorting (but only limited to 20 or so stocks), so you can save a bit of brokerage there
2) Depends on the liquidity of the stock and your provider, this can range from 5% to up to 100% of your leg size, but on average about 15% - 30% (50% for IB)

With a leg size of 2000 you are effectively paying 0.5% per trade, so that's 2% in total!! I would be happy enough if I can break even with that amount of brokerage costs
But if that's the amount you are willing to risk, I think you are better off starting in the US market, where brokerage cost isn't generally an issue like the ASX. The flip side - sleep 1 am at night or wake up 6 am in the morning !!

Thanks for the reply mate.

Would you recommend any broker for US/UK trades ?

Cheers
Sri

 

[fingl]

A trader should establish some rules to determine whether a pair performance is good for trading regardless of how good the historical performance. Exceeding historical max draw down is one way but this may delete some excellent pairs. Traders should expect max draw down to be exceeded in real time trading. What is the optimal rules to determine whether a pair is good for trading?

 

[fingl]

What is the optimal rules to determine whether a pair is good for trading?

Should those pairs exceeding historical max draw down in real time trading be deleted from the portfolio because their performance is no good currently compared to historical performance?

 

[SilverRanger]

Thanks for the reply mate.

Would you recommend any broker for US/UK trades ?

Cheers
Sri

Interactive Brokers, the cheapest everywhere!! Haven't done much trading in the UK, but from memory it's better to use CFD over shares to avoid the stamp duty, even then the brokerage is still not as cheap as US

 

[SilverRanger]

Should those pairs exceeding historical max draw down in real time trading be deleted from the portfolio because their performance is no good currently compared to historical performance?

There's no right or wrong answer for this, personally I do remove any bad performing pairs from my watchlist, but I would refresh the list every quarter.

 

[fingl]

There's no right or wrong answer for this, personally I do remove any bad performing pairs from my watchlist, but I would refresh the list every quarter.

How do you define bad performing pairs? Is quarterly refresh the optimal adjustment? You would carry on trade a bad performing pair to the next quarterly refresh?

 

[luckyforteja]

Interactive Brokers, the cheapest everywhere!! Haven't done much trading in the UK, but from memory it's better to use CFD over shares to avoid the stamp duty, even then the brokerage is still not as cheap as US

Thanks for the reply mate.

 

[skc]

Hi All,

Gone through the whole journal, very informative. I have couple of questions
1) What is the best instrument for Pair trading, mainly for shorting?
2) What would be minimum amount that need to be investing for each leg of the pair?

I have downloaded the trail version of pair trader and tried CFD(both long and short), with a leg size of 2000 and realised that brokerage was killing me.

Cheers,
Sri

I started with ~$5-6K per leg and was able to generate reasonable return after brokerage. $2K is pushing it a bit imo for ASX shares.

If you use CFDs make sure you are using DMA, and not paying the spread each time.

Here's a screen shot of a great pair...

Notice the peaks and troughs (green verticle lines) on the ratio chart (i.e. when they are most diverged) correspond to the lows on the correlation chart. Notice also the peaks in the correlation chart (red lines) correspond with not a lot of divergence from the mean. So putting on a trade when correlation is low can yield better results, provided that correlation is normally good.

View attachment 42959

BTW the new software is coming out with co-integration calculated which is going to be fantastic...

So much for my great pair... coming into dividend season for these trusts and the yield chasers are pushing MAP from >7% to 6.5% yield. AIX is 5.4%.

Moral of the story? Know your dividend schedule well in advance.

 

[fingl]

$2K is pushing it a bit imo for ASX shares.

What about 2k for the us market?

 

[luckyforteja]

I hope that people are still using this strategy successfully.

Can we have some updates ?