Market Neutral Corner | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Zeusjoes Market-Neutral Hedge Strategy 805 replies

Spectre's Corner 273 replies

FX Market Neutral Systems 5 replies

Statistical mean of the market [quant corner] 191 replies

Complete Trader's Corner, by Sam Shenker 0 replies

Attachments: Market Neutral Corner
Exit Attachments

Market Neutral Corner

  • #21
  • Jan 29, 2016 3:41pm Jan 29, 2016 3:41pm
Quoting Tardigrade
Since I am not a big fan of standard deviation/normal distribution, plus more free parameters=more overfitting, my own focus is on Correlation Coefficient only with supplementary method to calculate distance Share your thought and we may have some discussions going.
It really depends upon how you calculate your correlation coefficient. Some coefficients have many of the same assumptions (normality, equal variances, etc.) as regressions and other parametric methods.
  • #22
  • Jan 29, 2016 4:10pm Jan 29, 2016 4:10pm
Quoting Sciurus
{quote} It really depends upon how you calculate your correlation coefficient. Some coefficients have many of the same assumptions (normality, equal variances, etc.) as regressions and other parametric methods.
You are right. To avoid being caught in the Gaussian, one would ditch Pearson and replace it with Spearman; finding a substitute for SD's bit tough though.

Thanks.
No Highs? No Lows? Must be Bose!
  • #23
  • Edited 6:03am Jan 30, 2016 4:49am | Edited 6:03am
Quoting PipMeUp
This being said I'm very interested in your way to manage the positions without using standard deviation or at least quantiles of some distribution. I'm all ears.
By managing positions are you referring to rebalancing?

We may come from different schools in that regards---mine's static hedge so the rebalancing won't happen as often as your dynamic hedge.
No Highs? No Lows? Must be Bose!
  • #24
  • Jan 30, 2016 5:46am Jan 30, 2016 5:46am
@merquise
I think it is important to understand the difference between correlation and cointegration. Whereas everybody has a good grasp on what correlation is, cointegration is not a word commonly used. They look like they are the same thing. If you search the term in google or wikipedia you are immediately thrown I(0) and I(1) series at your face... Let's make it simple:
- Two series are correlated when they have a tendency to move in the same direction at the same time (or the opposite direction if the correlation is negative).
- Two series are cointegrated when the distance between them has a mean value. But they can move in any direction at any time.

In the picture I posted, the first two series (two straight lines actually) are going away from each other. They always move in the same direction. Any step the blue goes up, the red goes up as well 100% of the time (correlation=1). The distance between them increases perpetually. It has no long term mean.

The second picture shows two series that will grow infinitely but the average distance that separate them is constant (zero here). They "dance" around each other but they are free to go up or down independantly of each other's move (correlation=0). Note that the distance is not required to be zero. It can be anything. They don't have to cross. The two series are also allowed to wander away from each other for some time. The distance is constant in average only.

Both can regress on each other but the cointegration can only exist if something external links them. This "something" is the statistical edge. You needn't know what this "something" is.
Grains like corn and soy have in common that they both feed livestock.

@GoldTheHun
I like the pragmatic approach taken by 7bit in his thread http://www.forexfactory.com/showthread.php?p=4198691. He argued with old_dog on how important it is to stastically validate the cointegration. Old dog's point was that it is paramount. But tests like ADF or Johansen have a hard time with financial time series (near unit root, volatility clustering, fractional integration). So 7bit's approach was "let's pretend it works despite we know it doesn't always and when it gets too bad let's rebalance". Nothing is perfect anyways. As long as the rebalancing doesn't happen to often.

@kprsa
You are right for the example I gave because they both go upward. By skipping samples you'll more and more only take their trend into account. Each individual move would represent less and less in percentage. Since they go up at each step you'll see that when blue goes up red does so all the time (correlation=1). But if there was no trend you would have each move independant again (correlation=0). Nonetheless you rise the good point that the sampling frequency is important see the two posts here http://www.forexfactory.com/showthre...92#post6916092.

@Tardigrade
No I wasn't referring to rebalancing but to the timing of entries and exists. But I thought about the rebalancing issue when I read the PDF you posted. Using a local regression I think the variance of the coefficients can be high (aka need to rebalance often). Especially if the implicit window (the bandwidth of the kernel) is narrow.
No greed. No fear. Just maths.
  • #25
  • Jan 30, 2016 8:46am Jan 30, 2016 8:46am
Quoting PipMeUp
@Tardigrade No I wasn't referring to rebalancing but to the timing of entries and exists. But I thought about the rebalancing issue when I read the PDF you posted. Using a local regression I think the variance of the coefficients can be high (aka need to rebalance often). Especially if the implicit window (the bandwidth of the kernel) is narrow.
@PipMeUp

Don't know if you are comfortable with semi-discretionary style, but I use technical analysis to devise entry/exit rules as if the basket is just another ordinary spearding combo without any "neutrality" assumption; of course, no BB as usual

Regards
No Highs? No Lows? Must be Bose!
  • #26
  • Jan 30, 2016 3:25pm Jan 30, 2016 3:25pm
In theory "any" fx trade is a spread trade since you are opening up a long position in one countries currency vs a short position in another countries currency.

Lets take a long trade in eurusd for example. You are long euros and short dollars.

What you are trying achieve by becoming "market neutral" is to trim excess volatility affecting your desired outcome. Nominal price alone has always been a terrible way to evaluate a given market.

A perfect example of late has been the Crude decline. Instead of playing the weakness directly in Crude it was a far better to trade USDCAD. If I were trading nominal Crude my position would have been hard to hang onto every time Opec opened there mouths or the Saudis started threatening Iran. By instead trading short CAD - based on Crude weakness and strong USD - strengthening economy I was able to exploit the weakness in Crude with far reduced volatility. Take a look at the USDCAD chart over the last year.

This has got to be the best way to find Alpha.
  • #27
  • Edited 5:32am Jan 31, 2016 4:05am | Edited 5:32am
Quoting PipMeUp
Care is to be taken when using correlation coefficient. You may well find fake relationships due to a "third party" correlator, usually USD. For instance you would easily find a high correlation between USD/CAD and WTI oil. Yet this correlation is mainly coming from the strength/weakness of USD. These two assets used to be cointegrated but they no longer are despite what a lot of people keep believing. This is an image I made long ago (there is no data for 2015) that clearly shows it. It shows the returns of Oil vs the returns of CAD/USD (reversed...
I meditated on what you said about spurious relationship for a few days and still feel the importance of it could not be appreciated enough; a highly-correlated pair/basket indicated by statistics alone does not insinuates fundamental tandem of long turn outlook, let alone co-movement in distance. A logical way to put it, an intrinsic-linked asset basket may not guarantees its individual components behave in tandem, but an intrinsic-no-relationship-whatsoever combo will be subjected to more model-breakdown risk due to external market forces beyond mathematical relationships.

Few ways to solve the asset selection problem as I can think of:

1. Increase window length of the statistics being used
Pros: May indicate sustainable long term relationship
Cons: Curve-fitting to the balls, may get ED on extrapolation

2. Use fundamental relationships
Pros: Parameterless
Cons: Unquantifiable

3.Reduce window length of the statistics being used (Whaaaaat?)
Pros: Subdued influence from geopolitical risk
Cons: Need method to detect structural break

I think what's been discussed here barely even scratched the surface of spreading let alone market-neutral (Asset Selection, Stationary Process(Test), Hedge-Ratio Model, Order Management, etc) with some of the folks just idea-fishing, but I thank you @PipMeUp @All who did contribute and make this world a better place. That being said, I acknowledge the highly specialized nature of such approach, augmented by the lack of realistic statistical tools & increased difficulty of multi-asset management that may prevent it from appealing to the mass but few inquisitive.

Here's a anecdote for your effort

Once a group of students decided to study empirically the causes of drunkenness.
They drank vodka and water. Got drunk.
They drank rum and water. Got drunk.
They drank scotch and water. Got drunk.
They drank bourbon and water. Got drunk.
They drank gin and water. Got drunk.
They looked at the patterns of their data and saw that water appeared in every instance in which they got drunk, while no other factor appeared to remain in every case. They concluded that water made them drunk.
No Highs? No Lows? Must be Bose!
  • #28
  • Jan 31, 2016 4:08am Jan 31, 2016 4:08am
http://epchan.blogspot.com/2011/06/w...eaks-down.html
No Highs? No Lows? Must be Bose!
  • #29
  • Jan 31, 2016 4:18am Jan 31, 2016 4:18am
Quoting redbaron1981
In theory "any" fx trade is a spread trade since you are opening up a long position in one countries currency vs a short position in another countries currency. Lets take a long trade in eurusd for example. You are long euros and short dollars. What you are trying achieve by becoming "market neutral" is to trim excess volatility affecting your desired outcome. Nominal price alone has always been a terrible way to evaluate a given market. A perfect example of late has been the Crude decline. Instead of playing the weakness directly in Crude it was...
Not a conventional view but alright, what about stocks traded in NYSE quoted in USD(ex. MSFT/USD)? Are they spread trade in your definition?
I like the "trim excess volatility affecting your desired outcome" part.
A fundamentalist's approach of "Search-Validate-Leverage-Destroy"?
No Highs? No Lows? Must be Bose!
  • #30
  • Jan 31, 2016 4:56am Jan 31, 2016 4:56am
Quoting Tardigrade
{quote} Not a conventional view but alright, what about stocks traded in NYSE quoted in USD(ex. MSFT/USD)? Are they spread trade in your definition?
Yes they are.

When you are buying AAPL you are shorting the USD. It's AAPL/USD. If I lived in Europe and I bought shares of AAPL, they would be purchasing dollars and selling Euros. Technically speaking you could say they were buying "Apple Dollars" against Euros. When you buy shares of AAPL you are essentially willing to give up the interest rate you are getting by holding dollars in exchange for the return of holding AAPL shares.

This example might help. Let's say you have $1,000 in cash. You have two choices. You could invest that 1000 in a interest bearing account for one year that pays say 2%. Or you could forgo that 2% and take your chances with $1,000 worth of AAPL stock. The opportunity cost for owning AAPL is 2%. Essentially you are giving that away by selling the dollar in exchange for shares of AAPL. Hope that makes some sense.
  • #31
  • Jan 31, 2016 5:13am Jan 31, 2016 5:13am
Quoting PipMeUp
But if there was no trend you would have each move independant again (correlation=0.
You are right, it is conceivable to have cointegration with two flat signals, where the correlation will be weak on any tf.
k
  • #32
  • Jan 31, 2016 5:25am Jan 31, 2016 5:25am
Quoting redbaron1981
{quote} Yes they are. When you are buying AAPL you are shorting the USD. It's AAPL/USD. If I lived in Europe and I bought shares of AAPL, they would be purchasing dollars and selling Euros. Technically speaking you could say they were buying "Apple Dollars" against Euros. When you buy shares of AAPL you are essentially willing to give up the interest rate you are getting by holding dollars in exchange for the return of holding AAPL shares. This example might help. Let's say you have $1,000 in cash. You have two choices. You could invest that 1000...
Thank you for chipping in.

You would've need at least two instruments to construct a spread.
No Highs? No Lows? Must be Bose!
  • #33
  • Jan 31, 2016 5:31am Jan 31, 2016 5:31am
Quoting Tardigrade
{quote} I like the "trim excess volatility affecting your desired outcome" part. A fundamentalist's approach of "Search-Validate-Leverage-Destroy"?
Excuse me for using a dated image from one of my earlier posts but to provide a better example would mean having to start my trading machine up. In the below image you are seeing in the upper chart flat price of a continuous crude oil future vs the lower chart which is a simple calendar spread. You can recognize visually that the lower chart has less vol just by looking at wicks and range.

Exchanges recognize this by carrying out volatility scans and adjusting margin accordingly. Look at the following link here for CME margins on inter market spreads. If you select Nymex Crude (NY-CL) from the drop down you will see that trading the spread on prompt months can yield you an 87% margin discount. Prompt month Crude margin is 4000$ per contract so the 87% margin discount makes a hell of a discount. So basically for 1040$ you can own 2 contracts of Crude.

http://www.forexfactory.com/attachme...1&d=1444205902
  • #34
  • Jan 31, 2016 5:51am Jan 31, 2016 5:51am
Quoting Tardigrade
{quote} Thank you for chipping in. You would've need at least two instruments to construct a spread.
I don´t think you are understanding me correctly. Every trade is kind of spread whether you choose to recognize it or not.

For example in the 1980's a killing was made with the purchase of Brady Bonds. These were debt instruments issued in Brazil and other South American Countries. The killing was made threefold. Their debt was paying huge interest rates. The principal value of their debt increased and lastly the value of the currency shot up.
  • #35
  • Jan 31, 2016 6:51am Jan 31, 2016 6:51am
Quoting redbaron1981
{quote} I dont think you are understanding me correctly. Every trade is kind of spread whether you choose to recognize it or not. For example in the 1980's a killing was made with the purchase of Brady Bonds. These were debt instruments issued in Brazil and other South American Countries. The killing was made threefold. Their debt was paying huge interest rates. The principal value of their debt increased and lastly the value of the currency shot up.
1. I agree you for saying "every trade is kind of spread"---no instrument can be quoted without a fiat currency "price" assigned to it, which essentially means for every transaction, a trade involves not 1 but 2 instruments: buys/sells an instrument while simultaneously sells/buys a quoted currency.

2. What this thread differs is the goal & deliver mechanism of a strategy, the "search" & "leverage" part. Spreading itself is usually a directional bet, I 'm longing & shorting Japan right now with a spread. Market neutral tends to be spreading that strives to eliminate directional bias.

3. Your insights on fundamentals are very welcomed.
No Highs? No Lows? Must be Bose!
  • #36
  • Jan 31, 2016 12:03pm Jan 31, 2016 12:03pm
Quoting Tardigrade
{quote} 1. I agree you for saying "every trade is kind of spread"---no instrument can be quoted without a fiat currency "price" assigned to it, which essentially means for every transaction, a trade involves not 1 but 2 instruments: buys/sells an instrument while simultaneously sells/buys a quoted currency. 2. What this thread differs is the goal & deliver mechanism of a strategy, the "search" & "leverage" part. Spreading itself is usually a directional bet, I 'm longing & shorting Japan right now with a spread. Market neutral tends to be spreading...
For part 2 of your last post, I think I have most probably explained sufficient of you to grasp what SPAN margin is and how to use this to reduce margin requirements.

With regard to part 3 my main trading style is Price Action. The term itself Price Action really has been applied out of concept on this forum so much, that the true meaning of Price Action has been lost. Indeed most posters on here believe that Price Action is the use of trend lines, OHLC patterns in conjunction with support/resistance etc. The Price Action term has constantly been rehashed into commercial trading method names etc. I can tell you with 100% certainty that non of these systems are PA based.

The truth is that price action trading takes a lot of time and effort to procure and there is certainly no free lunches. What I can stress is that the market gives out a constant stream of information. I hear so many traders starting the week by saying the EUR should be bearish, the USD should be bullish. Expecting the market to act in a logical manner is very naive. Just doesn't work like that never has.

Instead of taking the arbitrary logic in our head and forcing it upon the market, we should be listening to what the market is saying and forcing it upon our trade decisions. If you know how to listen correctly, the market can tell you a lot during the day. The best starting point I can recommend for any trader is to build a framework in order for one to 'listen' to the market.
  • #37
  • Feb 1, 2016 8:04am Feb 1, 2016 8:04am
A good starting point for the inquistives

Trading Spreads and Seasonals
The Handbook of Pairs Trading
Pairs Trading: Quantitative Methods and Analysis

Blindly following trading methods in educational materials without checking their premises won't help you surviving in this niche
No Highs? No Lows? Must be Bose!
  • #38
  • Last Post: Feb 3, 2016 10:33am Feb 3, 2016 10:33am
Quoting redbaron1981
{quote} For part 2 of your last post, I think I have most probably explained sufficient of you to grasp what SPAN margin is and how to use this to reduce margin requirements. With regard to part 3 my main trading style is Price Action. The term itself Price Action really has been applied out of concept on this forum so much, that the true meaning of Price Action has been lost. Indeed most posters on here believe that Price Action is the use of trend lines, OHLC patterns in conjunction with support/resistance etc. The Price Action term has constantly...
Normally a trader would put a piece of arbitrary claim such as "Stat Arb won't work" or "market neutral strategy is doomed" into a paper shredder, enjoying the Whirrrshhhhhhhcccchhh sound as it goes, LOL. I am not that kind of trader and long passed the stage of seeking ego boost from a online forum.

What I do like to point out is that the financial market as one of the most complex synthetic structure ever created, is inherently the sum of expectations from all speculators, hedgers, arbitragers across the globe, hence an "arbitrage" would not exist if there hadn't been one already inside the market micro-structure itself.

As a market-neutral strategist I ask myself the same question everyday: am I listening to the market right now or am I just forcing it? There is no black and white answer.

I thank you for your input on this subject matter.
No Highs? No Lows? Must be Bose!
0 traders viewing now