Compounding Long-Only

[BlueHorseshoe 10]

Hello,

 

I've just begun trading a second account today (18 months in the planning and research), with rather different and more conservative goals to what I've done elsewhere, and I wanted to share some thoughts . . . Here's some basic info:

 

• The strategy is long-only a portfolio of 12 ETFs
  
• Positions are rebalanced monthly
  
• Allocations are the result of a rule-based adaptive algorithm
  
• Leverage is configured at 0.96
  
• A position is held in all 12 markets at all times
  

 

So here's the conundrum . . .

 

The strategy shows a greater edge, when the simplest of money management approaches (reinvestment of return, reinvestment of dividends) is applied, than when position sizing is based on a static account size.

 

How can this be?

 

Imagine investing in 4 shares of single stock. After a strong month, during which the stock rallied 100%, you liquidate your position. Your equity has doubled. Does this mean that you now go and buy 8 shares of this stock? Of course not: the stock now costs twice as much. You can only buy 4 shares.

 

Now imagine you split your equity evenly between purchases of shares of two equally priced stocks, with 2 shares in each. The first doubles in value; the second exhibits no change whatsoever. Your account has increased in size by 50%, and so some of this increase in available equity is passed on to your new position size in the second stock (the one whose value remained static).

 

You can only buy 2 shares of the first stock, but you can buy 3 shares of the second. The second is the "weaker" stock - the one that has demonstrated the least return for a long-only trader.

 

Now consider my long-only strategy, which is designed to benefit only from price increases. When returns are compounded evenly in allocation to all components of the portfolio, and the strategy becomes more profitable with this type of compounding, then the increase in alpha MUST COME FROM THE WEAKER PERFORMING COMPONENTS.

 

That's right: the strategy has been designed to benefit when price goes up, but the money management element will increase returns purely by increasing allocation to those markets that have gone up the least (or even fallen).

 

What do people make of this? In a strategy predicated upon relative strength, breakouts, trends and outliers, is the money management actually drawing out additional alpha from mean reversion, of all things?

 

Kind Regards,

 

BlueHorseshoe

[jpennybags 11]

What do people make of this? In a strategy predicated upon relative strength, breakouts, trends and outliers, is the money management actually drawing out additional alpha from mean reversion, of all things?

 

I think if there were a name that you could put to such a strategy it would be "trading the cycle" or possibly "value investing". If you watch a basket of individual stocks day-to-day it's fairly easy to spot the cycle in a market that is trending. This becomes more evident if the basket ranges across all sectors.

 

Reversion to the mean is a statistical fact of life as we know it, but playing that game can be frustrating. That's why I became a day trader... my account was always "reverting to the mean".

 

Edit: I'd like to think that I would be better at it now than I was then... not so sure though.

Edited July 2, 2014 by jpennybags

[MightyMouse 12]

How can this be?

 

 

Are the etfs equities markets biased?

 

Equities have done phenomenally well over the last 5 1/4 years. It was hard to lose money, long term, with anything that is a component of the s&p 500, or other related corporate index.

[BlueHorseshoe 10]

Are the etfs equities markets biased?

 

Equities have done phenomenally well over the last 5 1/4 years. It was hard to lose money, long term, with anything that is a component of the s&p 500, or other related corporate index.

 

Hi MM,

 

A few are equity indices, but it's pretty well diversified with things like metals, energies, interest rates, and timber woodland in there as well. The weighting of each component in the portfolio is a dynamic feature, so the allocation to equities may be minimal (as it was in 2008 for example).

 

The strategy has underperformed the S&P500 throughout the past few years.

 

Kind regards,

 

BlueHorseshoe

Edited July 6, 2014 by BlueHorseshoe

[SIUYA 11]

Hello,

 

 

What do people make of this? In a strategy predicated upon relative strength, breakouts, trends and outliers, is the money management actually drawing out additional alpha from mean reversion, of all things?

 

Kind Regards,

 

BlueHorseshoe

 

You have simply modified the portfolio strategy.....

if you rebalance every month you are making quantity to buy decisions based not on mean reversion, breakouts or any such changes in prices. You are making it based on your rule for rebalancing. There are many ways to do this, no rebalancing, rebalancing monthly, quarterly, at some change in % or absolute equity.....etc;etc.

 

Imagine it this way - you are a fund manager and you have to deal with subscriptions and redemptions each month.

You have to have a rule that either simply increases/decreases each quantity each month, OR you will have a change in each individual position proportionally - ie; investors either are diluted or get increased concentration to different positions for their returns going forward if you treat all investors the same with the same NAV calculation.

(One of the great reasons why similar styled funds have different returns even if they have similar entry exit triggers and many of them say its all in the money mgmt)

 

So in a nutshell its just another component to consider as part of an over all strategy for the portfolio simply as you have added the extra rule of rebalancing - (and why when many backtesting systems say they are able to test for a portfolio is BS, as it can get more complicated than they allow for)

 

......I also think that you are miss using the common usage of alpha. Your may or may not get extra alpha out of your strategy, depending on what the benchmark for your alpha measurement does. Hence while it might underperform the SP recently, it might be great over 10-15-20 years.

 

Your point might be better described as - if I apply this money mgt rule to a breakout type entry strategy does it massively change the returns over the long run, and deviate from the strategy I am implementing, or does it provide a better risk return profile--- at a complete guess it might smooth the returns and lower them.

[BlueHorseshoe 10]

You have simply modified the portfolio strategy.....

if you rebalance every month you are making quantity to buy decisions based not on mean reversion, breakouts or any such changes in prices. You are making it based on your rule for rebalancing. There are many ways to do this, no rebalancing, rebalancing monthly, quarterly, at some change in % or absolute equity.....etc;etc.

 

Imagine it this way - you are a fund manager and you have to deal with subscriptions and redemptions each month.

You have to have a rule that either simply increases/decreases each quantity each month, OR you will have a change in each individual position proportionally - ie; investors either are diluted or get increased concentration to different positions for their returns going forward if you treat all investors the same with the same NAV calculation.

(One of the great reasons why similar styled funds have different returns even if they have similar entry exit triggers and many of them say its all in the money mgmt)

 

So in a nutshell its just another component to consider as part of an over all strategy for the portfolio simply as you have added the extra rule of rebalancing - (and why when many backtesting systems say they are able to test for a portfolio is BS, as it can get more complicated than they allow for)

 

......I also think that you are miss using the common usage of alpha. Your may or may not get extra alpha out of your strategy, depending on what the benchmark for your alpha measurement does. Hence while it might underperform the SP recently, it might be great over 10-15-20 years.

 

Your point might be better described as - if I apply this money mgt rule to a breakout type entry strategy does it massively change the returns over the long run, and deviate from the strategy I am implementing, or does it provide a better risk return profile--- at a complete guess it might smooth the returns and lower them.

 

Hi SIUYA,

 

The strategy is best termed 'relative strength' rather than 'breakout' - I just tend to lump anything that is directional and not mean reversion together as 'go with' - trend following included.

 

I may be misusing the term alpha . . . I was considering an equal-weighted portfolio of the components (ie a 'passive' investment in the portfolio) as the benchmark. The portfolio as I have implemented it, with variable weights, shows (historically) a greater return than this benchmark - is this not usually termed alpha?

 

The key question I'm asking though is this: when profits can be traced to a particular portfolio component, should the "benefit" of increased position sizing be passed to just this one component, or to the portfolio as a whole?

 

What I have found in this instance is that performance is improved when the benefit is passed equally to every component in the portfolio. This must mean that the improvement in performance is the result of increasing position size for poorer performing components, surely?

 

Cheers,

 

BlueHorseshoe

[SIUYA 11]

hi,

re alpha yes - I thoughts you were simply talking about alpha 'generally' and not in direct ref to an equally weighted benchmark, and even then the benchmark needs to be standardised...but my confusion. (too much beer on the weekend in Prague) The way you are referencing it is good.

I think this is also one of the issues when people pigeon hole a strategy type as men reversion, breakout, trend etc; as it needs to be compared to a standard otherwise it is hard to compare apples and oranges.

''''''''''''''

"The key question I'm asking though is this: when profits can be traced to a particular portfolio component, should the "benefit" of increased position sizing be passed to just this one component, or to the portfolio as a whole?

 

What I have found in this instance is that performance is improved when the benefit is passed equally to every component in the portfolio. This must mean that the improvement in performance is the result of increasing position size for poorer performing components, surely?""

 

If you are rebalancing your whole porftolio, then you must be doing it taking into consideration all parts of the portfolio if you are doing it properly, otherwise its hindsight(???) How would you decide to not increase some or others....or are you simply doing the 'dogs of the dow ' theory and buying laggards but with extra leverage on top of a evenly balanced portfolio.

eg; buy 2 at $1, when one goes to $2 and the other is at $1, you dont sell any at $2, but simply buy the more of the one still at $1.

 

'''''''''''''''''''''''

I think I am getting at what you are talking about.....I remember doing a test once that had a portfolio limitation of not having too many correlated instruments and so it did not rebalance but it did not take new entries in what we determined to be highly correlated instruments if you were 'full up' in that sector. In a classic do your head in thing, during the Internet bubble approx 2000, because the portfolio was full of equities already it never took the nasdq buy signals. This meant massive underperformance of a benchmark.....and because we never rebalanced it never changed. When introducing variants (even small ones) for rebalancing, or spreading over further instruments, or even concentrating it more it had big changes in return profiles, and this also looked different over the time frame of the tests - eg; it never had the same issue when the naq burst and it even managed to put on a short.

So my guess is all you can do is simply test and see what happens over the long run and what you can live with....sometimes its great to be full tilt othertimes not.

[BlueHorseshoe 10]

...but my confusion. (too much beer on the weekend in Prague)

 

It's probably me not explaining myself clearly enough to be honest

 

If you are rebalancing your whole porftolio, then you must be doing it taking into consideration all parts of the portfolio if you are doing it properly,

 

Yes, but this isn't necessarily linked to the compounding aspect, is it? For example, suppose I draw off any profits so that the account always remained at a fixed amount by month end, but still rebalance the portfolio each month based on a consideration of all parts of the portfolio.

 

The allocation of reinvested profits could differ from the allocations of the rebalanced portfolio aside from these. For example . . . 10k account with 60% (6k) allocated to Stock A . . . 1k portfolio profit at month end . . . Monthly rebalance, 70% of 10k (7k) allocated to Stock A, but only 30% of profits (300) allocated to Stock A, so position size is 7,300, and not 70% of 11k (7,700).

 

....or are you simply doing the 'dogs of the dow ' theory and buying laggards but with extra leverage on top of a evenly balanced portfolio.

 

Quite the opposite - it's basically "relative strength" - increase position size for the leaders, reduce position size for the laggards . . . then a load of fancy machine-learning thrown at it (purely for my own gratification and entertainment - think basic strategy returns 13%, me playing around with code for a year might add at best 2% and smooth volatility of returns if I'm lucky!).

 

You can see the shifts in weighting of each component in the subpane of the curve I've attached. One component always has a weight of zero (at the minute, metals).

 

We're talking very long term outlook here

 

So my guess is all you can do is simply test and see what happens over the long run and what you can live with....sometimes its great to be full tilt othertimes not.

 

Sure. I've done this, and I'm pretty comfortable with what I'm doing.

 

In a previous thread here I argued that the individual components of a portfolio should be 'rewarded' with increased position size dependent on their unique performance; I now find myself doing the opposite and spreading the benefit from components that have performed strongly to 'reward' equally those that have underperformed. And yet that's what masses of testing tells me is the right thing to do.

 

Cheers,

 

BlueHorseshoe

[SIUYA 11]

Yes, but this isn't necessarily linked to the compounding aspect, is it? For example, suppose I draw off any profits so that the account always remained at a fixed amount by month end, but still rebalance the portfolio each month based on a consideration of all parts of the portfolio.

 

The allocation of reinvested profits could differ from the allocations of the rebalanced portfolio aside from these. For example . . . 10k account with 60% (6k) allocated to Stock A . . . 1k portfolio profit at month end . . . Monthly rebalance, 70% of 10k (7k) allocated to Stock A, but only 30% of profits (300) allocated to Stock A, so position size is 7,300, and not 70% of 11k (7,700).

 

........

 

In a previous thread here I argued that the individual components of a portfolio should be 'rewarded' with increased position size dependent on their unique performance; I now find myself doing the opposite and spreading the benefit from components that have performed strongly to 'reward' equally those that have underperformed. And yet that's what masses of testing tells me is the right thing to do.

 

Cheers,

 

BlueHorseshoe

 

Hi, I thought I got it but now I am confused again (too much wine tonight)....from your example at the end of the month, its like you are reweighting based on 2 different formula components.

A normal reweighting formula --- this takes you from 60% to 70% for stock A and

An inverse profit formula - you are taking the profit for the month (1k) and rebalancing on a % of that, as an inverse % of the total weighting.

Hence you are actually increasing your winners by a smaller amount if there is a profit (assuming the first normal reweighting formula is flat).

 

So I get that you are now reweighting across all portfolio instruments. Regardless of where the PL came from.

 

I dont think compounding makes much difference as to how you are getting it as it looks more like you are capturing a little bit of both trend following and mean reversion. Like having a mix of two strategies in one, and you are making the most of the only free lunch there is - diversification......which makes a lot of sense....and is an interesting take on it.

[BlueHorseshoe 10]

A normal reweighting formula --- this takes you from 60% to 70% for stock A and

An inverse profit formula - you are taking the profit for the month (1k) and rebalancing on a % of that, as an inverse % of the total weighting.

 

Hi SIUYA,

 

My lack of clarity again (and sadly I've had neither beer nor wine for several days).

 

The "Stock A/B" example I gave was purely by way of explaining that reinvested returns could be allocated with a separate criteria to the one used for rebalancing. So profits would not have to be distributed back to the poorer performing components.

 

This isn't what I'm doing though . . .

 

So I get that you are now reweighting across all portfolio instruments. Regardless of where the PL came from.

 

That's correct.

 

I dont think compounding makes much difference as to how you are getting it as it looks more like you are capturing a little bit of both trend following and mean reversion.

 

That's what I have concluded, although the intention was to capture the former (as my swing trading account focuses on mean reversion and I wanted diversification of styles across the two accounts).

 

As always, thanks for your thoughts and help.

 

Regards,

 

BlueHorseshoe