ASX Stock Pairs Trade Journal

[skc]

Hi skc,

Looks like I found the problem.
Changing SD to +- 2.00 produces a bunch more signals (e.g., 4 STO pairs today).

Would you recommend using SD of +- 2.00 or sticking with the default setting of
+- 2.70?

And do you ONLY take signals that show divergence on or is this not a must have.

sleepy

Exactly. Setting the SD to 2.7 vs 2.0 makes a huge difference to the amount of signals generated.

I set my signal at 2.0, but that doesn't mean I take all trades/signals straight away. Usually checking the charts you can sort of "feel" that the stock hasn't yet finish its fall, or the recent spike (on the stock that the program is telling you to short) hasn't yet run out of momentum. I keep such pair on close watch, and wait for better entries (e.g. when the retracement is hitting recent support). At the end of the day, the entry SD doesn't matter - your P&L is purely governed by the entry price ratio.

But by doing this it also means that I miss trades that revert perfectly the next day... and I can't say I've recorded all the trades that I didn't take to see if my gut feel increaes / decrease my overall profitability.

So the short answer is there is no perfect SD...

FWIW I wouldn't long STO at the moment as the chance of cap raising is too high. I also wouldn't short LNC as further asset sale is supposed to be imminent. Sure STO may bounce or LNC may fall a few % tomorrow - but trading against those big events on the horizon is akin to picking pennies in front of a steam roller. The down side is easily 10-15%...

 

[sleepy]

Thanks skc,

So after you get SD signal and having looked at ratio charts, checked news etc in order to select which pairs you wish to trade, you then time your entry based on price chart action (e.g., suport/resistance etc). Is that correct?

And when back testing to select what pairs to add to watchlist do you use 3yr or 5yr, or 10yrs. Or does it not mattter which?

sleepy

 

[skc]

Thanks skc,

So after you get SD signal and having looked at ratio charts, checked news etc in order to select which pairs you wish to trade, you then time your entry based on price chart action (e.g., suport/resistance etc). Is that correct?

And when back testing to select what pairs to add to watchlist do you use 3yr or 5yr, or 10yrs. Or does it not mattter which?

sleepy

Pretty much. I put pairs on closer watch when SD criteria is triggered. Here's an example of how I am thinking.



Signal is to long PDN short ERA. The ratio chart looks ok, and trading in the overall larger time frame trade. The divergence (percent from mean) isn't that large (only 2%) and no where near historical extremes. Correlation has really broken down in the last 2-3 months. So on those alone I wouldn't place the trade now.

Look at the individual charts...



ERA sitting on a ledge very dangerously imo. $12.9 is holding things up but there is very little doubt that it will be retested, esp if the overall market was to retrace a bit (very likely given it's come a long way quite quickly).

If $12.9 is tested, that's a 3.5% fall from today's price... keep that in mind.



PDN is still well above last major low at start of July. What looked like renewed strength (after the bullish ABC correction in August) has probably broken down on today's bar. This suggests to me that the price may fall back down towards $3.4-$3.5. And that's a 5-8% fall from today's price.

So overall on the balance I decided not to take this trade. I've still got it on my close watchlist. The current ratio is 3.6, and I will re-assess if the ratio expands to 3.75 or so. Of course that's all speculating based on a mix of technical analysis and "gut feel" - for all I know, ERA could plunge and PDN will spike tomorrow...

With back test I only back test 1 year. The longer back test is only available in the recent software update so I've not bother going back and test again.

 

[Pairs Trader]

New trade

Long NAB @ 22.84
Short ANZ @ 22.33

Sold NAB @ 25.92 (13.48% gain)
Covered ANZ @ 24.17 (8.24% loss)

 

[SilverRanger]

Picked up my first pair of the month at market close today. Long DJS @ 5.06, short MYR @ 3.92.

Sold DJS @ 5.25 ( 3.75% gain)
Covered MYR @ 3.93 ( 0.26% loss)

Also noted a big divergence on NWS/NWSLV this morning, but only managed to get a half-arsed fill on NWSLV @ 14.54, closed out @ 14.8

AND should have gone back in at market close

 

[sleepy]

Hi skc,

Thanks for the detailed reply. Very informative.

Signal is to long PDN short ERA. The ratio chart looks ok, and trading in the overall larger time frame trade. The divergence (percent from mean) isn't that large (only 2%) and no where near historical extremes. Correlation has really broken down in the last 2-3 months. So on those alone I wouldn't place the trade now.

Do you not pay much attention to correlation when selecting your pairs for your watchlist?

As PDN/ERA would not be on my watchlist bc of the low correlation (<70%) and low profit (avg -$99.20 ... 1yr backtest). .. and hence no trigger would appear. However, I can see your logic in assessing each chart individually.

Also do you look the RSI chart or not bother?

sleepy

 

[skc]

Hi skc,

Thanks for the detailed reply. Very informative.



Do you not pay much attention to correlation when selecting your pairs for your watchlist?

As PDN/ERA would not be on my watchlist bc of the low correlation (<70%) and low profit (avg -$99.20 ... 1yr backtest). .. and hence no trigger would appear. However, I can see your logic in assessing each chart individually.

Also do you look the RSI chart or not bother?

sleepy

There is much talk about correlation, but it's co-integration that you want. Do a google search and you can find more detailed explanations. Correlation is a proxy to co-integration and the program pairs thing up for you that way.

PDN/ERA had good correlation a while back, and they are the only 2 uranium producers on the ASX. So I am happy to trade them when things look right.

RSI - talk yourself through what RSI shows and convince yourself that it means something...

 

[SilverRanger]

Sold DJS @ 5.25 ( 3.75% gain)
Covered MYR @ 3.93 ( 0.26% loss)

Also noted a big divergence on NWS/NWSLV this morning, but only managed to get a half-arsed fill on NWSLV @ 14.54, closed out @ 14.8

AND should have gone back in at market close

Managed to get back in this trade this morning with a slightly better entry point, currently have my brokerage covered for it

 

[SilverRanger]

There is much talk about correlation, but it's co-integration that you want. Do a google search and you can find more detailed explanations. Correlation is a proxy to co-integration and the program pairs thing up for you that way.

PDN/ERA had good correlation a while back, and they are the only 2 uranium producers on the ASX. So I am happy to trade them when things look right.

RSI - talk yourself through what RSI shows and convince yourself that it means something...

RSI isn't strictly on my set of trigger criteria, but I would be happier to have it confirming with my other criteria when I pull the trigger than otherwise.

In theory RSI is just a generic momentum oscillator, and seeing the ratio (the relative value of two stocks) as "overbought" or "oversold" still makes sense to me. In a way, the RSI actually is better aligned with your profit and loss than the +/-, as sometimes you may see the +/- "reverting" while you are losing money, simply because the standard deviation (the denominator) is bigger

 

[skc]

Pretty much. I put pairs on closer watch when SD criteria is triggered. Here's an example of how I am thinking.

Look at the individual charts...

ERA sitting on a ledge very dangerously imo. $12.9 is holding things up but there is very little doubt that it will be retested, esp if the overall market was to retrace a bit (very likely given it's come a long way quite quickly).

If $12.9 is tested, that's a 3.5% fall from today's price... keep that in mind.

PDN is still well above last major low at start of July. What looked like renewed strength (after the bullish ABC correction in August) has probably broken down on today's bar. This suggests to me that the price may fall back down towards $3.4-$3.5. And that's a 5-8% fall from today's price.

So overall on the balance I decided not to take this trade. I've still got it on my close watchlist. The current ratio is 3.6, and I will re-assess if the ratio expands to 3.75 or so. Of course that's all speculating based on a mix of technical analysis and "gut feel" - for all I know, ERA could plunge and PDN will spike tomorrow...

So much for that... signal to close the trade



Long PDN Bought $3.7 sold $3.75 (+1.35%)
Short ERA Sold $13.31 cover $13.17 (+1.05%)

And that's how you miss a trade!

 

[sleepy]

Well I still cant/didnt get a signal for ERA/PDN so I presume its because Im using the default settings and using 100 day lookback.



However, even when I play with settings I still cant get the trigger signal.
And why 150 days on charts rather than 365 days?.

sleepy

 

[SilverRanger]

Managed to get back in this trade this morning with a slightly better entry point, currently have my brokerage covered for it

Covered NWS (1.98% profit) and sold NWSLV (0.95% loss), so roughly 0.6% profit after brokerage.

Not the best you can get out of this pair, but reasonable for an overnight position with double the normal leg size

This pair should be written to all pair trading textbooks

 

[SilverRanger]

Just wondering how is everyone doing for the month of September?
For me I am sitting around 8% (vs 5.4% of ASX200) at the moment, although 5 out of my 7 pairs are still open (I was meant to close MCR/MRE today but didn't )

 

[FXanyone]

Just wondering how is everyone doing for the month of September?
For me I am sitting around 8% (vs 5.4% of ASX200) at the moment, although 5 out of my 7 pairs are still open (I was meant to close MCR/MRE today but didn't )

I have had one of my better months. Took some grief from NWS vs APN, BLY vs DOW and MGX v FMG. Had a great run from QAN v MAP.

 

[sleepy]

Are you guys using the default settings ... or have you modified them based on your trading experiences?

sleepy

 

[SilverRanger]

Are you guys using the default settings ... or have you modified them based on your trading experiences?

sleepy

I'm not using the default settings, choosing the parameters in itself is an optimisation exercise, as you are trying to find an optimal balance between the no. of trades and profit per trade, this would be different for every pair

Managed to get out of MCR/MRE this morning with a slightly better exit

 

[skc]

Just wondering how is everyone doing for the month of September?
For me I am sitting around 8% (vs 5.4% of ASX200) at the moment, although 5 out of my 7 pairs are still open (I was meant to close MCR/MRE today but didn't )

Awesome month for me until this week... looks like I will have to settle for 5.4% although it was as high as 8.5%. Absolute shocker this week.

I did get that MCR/MRE closed yesterday...

 

[sleepy]

When calculating % return per month ... are you including closed pairs only or also those currently open.

sleepy

 

[skc]

When calculating % return per month ... are you including closed pairs only or also those currently open.

sleepy

I do mark-to-market. I find that it's easier to take a loss if it's already been marked off so to speak.

 

[sleepy]

There is much talk about correlation, but it's co-integration that you want. Do a google search and you can find more detailed explanations. Correlation is a proxy to co-integration and the program pairs thing up for you that way.

skc,
Sorry must have missed this comment earlier.

Based on the definitions below I see what you mean about co-integration being more important.

Co-integration
- The likelihood that, when two stocks deviate in valuation, they will revert to the mean

Correlation
-The measure of how two securities move in relation to one another


But I cant see how Correlation is a proxy for co-integration ... surely they 2 different things. Wouldn't it be better to have Co-integration also included in pairtrade finder and have the ability to filter by that.

sleepy