Australian (ASX) Stock Market Forum

Statistical Arbitrage Market Neutral Strategy

Jump to new
P

pairstrader

Joined
22 November 2011
Posts
8
Reactions
0
Hi All,

Some of you may have heard about Statistical Arbitrage Pairs Trading, I have just recently filled out a blog about the strategy and also opened a twitter account for everyone to follow.

In a nut shell, 80% correlated stocks on ASX30 that have diverge from each other of about 2.7 standard deviations we enter a long and short position then when they converge back to their long term average 1 standard deviation we exit the trade. It's algorithm signal based and also trade on the US Market.

Basically, I'm running a proprietary software thats been getting excellent results. Visit the blog to read up about the strategy and contact form is on the right of the blog if you want to see more results.

http://twitter.com/#!/au_pairstrader
http://aupairstrader.blogspot.com/

Best thing about the strategy, is that its an income strategy that doesn't care where the market moves. It simply waits for two stocks to converge back together. Before running live, we did 6 years of backtesting using all sorts of parametres.

Cheers
 
Yep, continued divergence does happen, however you must remember we're trading 80% correlated stocks i.e. NAB/CBA BHP/RIO OSH/CWN.. so divergence doesn't happen often and continued divergence is even rarer. In any case, all positions as part of risk management have stop losses.

So if the market falls 5% (BHP falls 4% RIO falls 5%)

We have back tested over 6 years and identified the 2.7 standard deviation as the tipping point for an entry signal from our algorithm for convergence.

If you want to know me and see signal results I have a contact page on the blog.
 
Nice blog. Good to see another pairs trader... they are a broker's best friend.

There's a pretty big thread on pairs trading here
https://www.aussiestockforums.com/forums/showthread.php?t=14508

I note that you don't restrict yourself to trading pairs that are fundamentally related...

What software are you using to monitor the prices, calculate deviations and generate signals? Excel?
 
Hello Skc,

Its an effective income strategy especially during volatile periods.

I trade professionally for clients - automated for them as I execute the pairs signals.

Ideally you want to be sector neutral, however because the top 30 shares make such a large composition (75%?) of the All Ords, the correlation is strong enough for you to trade.

The model is a proprietary model that we've developed with an institutional client, its a common strategy but we give clients who are time poor, the strategy exposure and we manage risk and slippage that affects the way the pairs work.

Currently have three open pairs, one flat and two are up, good result given the uncertainty of what's ahead in the markets.

Cheers Mate
 
Hi pairstrader,

Nice to see another PT journal going, will definitely be following with interest.

Some quick questions if you don't mind:

1. Do you trade dollar/beta neutral?
2. Are your signals filtered by company news (e.g. NWS plunged heavily due to xxx-gate)
3. How automated is your trading system? Is it doing the opening and closing of trades?

SR
 
Hi SR

1. Yes we are dollar neutral - however not Beta neutral, Beta is close but we do want outperformance aswell as convergence.
2. Our algorithm is statistical and data based - not fundamentals. However, our traders apply fundamental filters themselves and do not enter pairs if there is a fundamental reason not to.. also if ex-div etc.
3. Automated in the point of view for the clients. So yes, entry and exit signals. Some people do not realise the difficulty in execution and how critical "slippage" is (posted on my blog) hence, if your not in real-time mode and think being a few cents off either way doesn't matter, then you're mistaken. US market we are building full automation, for trader as well :)

Send me a PM if your interested from a retail client perspective as that forms my day-to-day job.

It's a great way to talk about an interesting strategy.

Cheers
 
Good Morning All,

I just added a new blog post, its on November's Pairs Trading results, it's what we trade on a daily basis. Have a look:

http://aupairstrader.blogspot.com/2011/11/results-from-our-proprietary-trading.html

So just to update, at the moment for my clients we have three open pairs:

BHP/AMP - flat
IPL/WDC - in the money
ORI/NAB - in the money

I will give all you folks the results when exit signals come (hopefully soon!!)

But you can also follow my twitter account for regular updates, also we can always discuss on this forum!

Have a good day
 
bhpamp.jpg

See above chart, they fit the criteria.

divergence.jpg

A close up view shows a divergence over the 2.7 std deviations from its average.

BHP/AMP is a while away from receiving a exit, so we sit tight.

Cheers
 
pairs, what happens when all the fund managers start doing the same with their bots? Might there develop a situation where almost every stock moves in unison like a big flock of birds with no divergence to profit from?
 
Gringotts Bank said:
pairs, what happens when all the fund managers start doing the same with their bots? Might there develop a situation where almost every stock moves in unison like a big flock of birds with no divergence to profit from?
Click to expand...

Implied and realised correlation vary across market conditions.

Pairs trades made while implied correlation is high and rising usually have a strong chance for convergence. You will get drawdowns on pairs trades when implied correlation is high and reversing to low.
 
sinner what's the difference between 'implied' and 'realized' correlation in pair trading?

Anyone have an opinion on pairtradefinder software? They make some huge claims about performance.
 
Hi Gringotts Bank,

Pairs Trading is not a new strategy - fund managers, hedge funds etc all around the world have used it.

Even with their existence, share prices will always diverge for whatever reason and converge back together. For the US market, we are still testing the pairs trading model and so far the results have been outstanding, actually even better than Aussie market.

Important to note, that there are more pairs in the US and higher volatility (higher loss and profit) but the same model and rules used here works very well in the US. Now, it's operating where 60% of US market volume is generated by robots and algorithm trading - so regardless of their input - correlated share prices will still perform the way it has over the last 10, 20, 50 years.

You can buy software off the shelf/internet like pairstradefinder - but you must know that refined models, re-weighted models and continuous analysis on performance and pairs ranking is required to ensure it works not in the short term but in the long run. Market dynamics change from time to time, therefore stringent analysis is required.

Thats what we do, we have a team of smart research and programmers, that spend a lot of hours refining the model including constantly testing the US market before we go live for clients. Another thing is, buying the software is one thing, but you have to execute it and minimise slippage. Liquidity on the Aussie market is a massive issue and does affect the way the pairs are traded.

Most of my clients do not have the capabilities to trade themselves, real time, any time. Receiving signals and executing them like what a robot would do is critical.

Cheers
 
alrighty. Can you post some extended backtests, here or on your blog please? Also some broker verified trades would be nice if you have that. :)
 
Gringotts, you can PM your email and I will send it from work.

The blogs are a bit restrictive on how much spreadsheets/pictures I load.. it could blow up :)
 
Gringotts Bank said:
sinner what's the difference between 'implied' and 'realized' correlation in pair trading?

Anyone have an opinion on pairtradefinder software? They make some huge claims about performance.
Click to expand...

Implied correlation is what you get if you calculate the correlation implied by options, variance swaps, other derivatives.

Realised correlation is what you actually end up with on close or expiry.

I used to trade pairs, but since value at risk is theoretically infinite and the edge after brokerage is so tiny unless you are using huge chunks of capital (or leverage) which just exacerbates the risk problem these days I just trade implied correlation.

Pairs trading is often touted as being market neutral but in effect you are just laying off your odds elsewhere, by betting that both

1. Correlation of your pair will remain high, i.e. you are betting that the pairs current deviation from its mean is caused by a temporary fluctuation in correlation which will go away soon.
2. Cointegration of your pair will remain high. Usually traders cope with this psychologically difficult portion of the bet by keeping their pairs within a specific industry or by using hard timed exits. This is the part of the bet that often turns a theoretically infinite loss to a real one. What do you do when cointegration collapses from 1 to 0?

To me it seems easier, safer, cheaper to just bet on implied correlation going up when it's down and going down when it's up. You can for example wait until implied correlation is very very high and then buy variance or an options straddle on a stock which has correlated well with the index over the short term but over the long term has little index correlation and at the same time sell variance or an options straddle on the stocks index.
 
To give you a better idea of what I mean,

pairstrader (the OP) trades huge huge chunks of money trying to get good fills on pair ratios which look like this:

Screen Shot 2011-11-23 at 12.35.15 PM.png
The bet is always the same, once the pair ratio price exceeds a bollinger band boundary (or ATR realtime boundary) then structure a long/short trade such that you profit if the ratio returns to the mean.

Personally for correlation I trade in models that look like this:

Equity
Screen Shot 2011-11-23 at 12.33.06 PM.png

Futures
Screen Shot 2011-11-23 at 12.32.38 PM.png

and find myself much more willing to place the "same bet", with better defined risk parameters and 0 chance of cointegration loss.
 
and find myself much more willing to place the "same bet", with better defined risk parameters and 0 chance of cointegration loss.
Click to expand...

Hi Sinner,

Your equity and future chart looks quite interesting. Can you explain what you mean by your last line? Where on the chart will you go long/short the correlation? Actually, is there a correlation if you are just going long/short straddle on a single stock?

SR
 
SilverRanger said:
Hi Sinner,

Your equity and future chart looks quite interesting. Can you explain what you mean by your last line? Where on the chart will you go long/short the correlation? Actually, is there a correlation if you are just going long/short straddle on a single stock?

SR
Click to expand...

I'd like to know too. Thanks.
 

Similar threads

bigdog
Dump Trump
9 10 11
Replies
214
Views
19K
wayneL
wayneL
R
The 7 Moats of Quality Companies
Replies
7
Views
7K
divs4ever
divs4ever
RichKid
Coin Tossing - Risk & Money Management
Replies
0
Views
5K
RichKid
RichKid
R
The Financial Endgame
Replies
18
Views
9K
qldfrog
qldfrog
Top