mindmoneymarkets

Combo templates - When two is better than one

In a past post I blogged about the advantages of combining a short term mean reversion indicator with an intermediate term indicator of the same bent. Click here to view the post. 

As a recap, the DVI/B combo method helped smooth the ride with the longer term DVI modulating the shorter term and more erratic DVB. These indicators are available from DV Indicators.

This year the DVI will have had its work cut out in modulating the shorter term indicators like the DVB which came unstuck in August (though have enjoyed a good month or so recently). 

To help assess this better I've created two templates for the DV Indis Research package. The general rule is that you only trade if both indicators are signalling in the same direction. You stay flat otherwise. 

Template 1 Combines a short term and intermediate term indicator. 

Template 2 Combines a short term indicator with a longer term unrelated trend type indicator. 

Combo 1 - Short term and intermediate term indicators. 

I've used the template to test out various combination options. The first idea presented a few months ago was the DVB/ DVI combination. This helped soften the blow of the DVB's poor performance, but did not avoid it entirely. Following my recent post analysing the components of the DVI, I have been focusing my trading on the DVIM component and stripping out the DVIS. This combo option improves results further. 

Finally, using the DVO instead of the DVB has been another way to lift results. 

Combo 2 - Short term and longer term trend or skew indicators.

Logically combining separate indicators should offer better diversification and this indeed seems to be the case. The longer term trend filter, the DLTI has been a great companion for shorter term MR indicators helping to reduce drawdown and keep a good level of upside. 

Short term mean reversion isn't quite dead yet, especially when its used in conjuction with a wiser, slower moving companion. 

Templates available soon. 

Close to Close or Open to Open?

Much of my trading and system research is based on entering the trade at the close and exiting at the next day's close. 

There's one practical problem with this - Trading from the UK this means checking/ acting on trades in the evening. This is fine most of the time, but a) It can be hard to switch off for the evening if you know you've got to trigger a trade b) when you are out and about, its a pain to have to dash out and check a signal, especially if you're in the middle of a meal or an martial arts class! 

The more practical alternative for European traders is to trade based around the open, but will performance still hold up?

DVIM Performance (Intermediate term indicator)

I've been researching various indicators and have generally come to the same conclusion. The DVIM has performed well this year on a close to close basis (CtC) but has Open to Open (OtO) been much different?

This is based on the basic rules of going long below 0.5 and short above 0.5.

In turns out, there has been very little difference in 2011. I think this is primarily a function of the holding period being a lot longer for the DVIM than other shorter term indicators like the DVB. 

Speaking of which...

DVB Performance (short term indicator)

Speaking of which...

Trading the DVB OtO this year has actually been better than close to close. This is certainly against the long term trend for the CtC to out perform. 

I'm sure those pesky Europeans are playing their part in this year's performance differential. 

The Rolling Sortinos

In a recent post I previewed a neat service coming soon from my colleague Jeff Pietsch that ranks various indicators by their rolling 252 period Sortino ratio.

I raised the possibility of being able use these rolling Sortinos to switch between systems or manage position size.

It’s taken a while to get the models built but they are showing some initial promise.

The Rolling Sortinos

Despite sounding like a travelling circus act, this metric does appear to have some use as a filtering tool.

DVB’s Evil Twin & The DVI.

I’ve blogged recently about the change in market behaviour recently, particularly in relation to short term mean or range reversion.

Quite simply, the market aint behaving like it used to.

In some ways, we had it too easy in recent years, with the clear out performance of range reversion indicators such as the DVB.

Now, despite the choppy conditions, the DVB indicator continues to perform poorly. The typical way to trade the DVB has been to go long with a reading below 0.5 and short with a reading above 0.5. Yet over the last 100 days, doing the exact opposite would have brought in some rich rewards.

I’ve taken to calling this approach DVB’s evil twin.

Here are the results from the last 100 days on the SPY.

Rules:

• Go long above 0.5.
• Go short below 0.5.

Will this continue? I'm not sure, but I do believe the DVB's days of easy pickings are gone. 

Intermediate Mean Reversion

While the short term mean/ range reversion has flipped around, the intermediate measures are enjoying a good run of form.

There are numerous intermediate Mean Reversion indicators available within the DV indicators package, but the best known is probably the DVI.

The DVI is actually made up of two indicators with indicator explanations from David Varadi. All these indicators are available in the DV Plugin.

DVIS (Intermediate Stretch):

This is one component of the DVI that isolates the stretch in terms of net days up or down over a series of periods and is re-scaled to create an accurate and consistent measure. The stretch is not like RSI because it does not consider the magnitude of up or down days but rather the net difference. It also looks at the short, intermediate and long time frames with a net weighting on the intermediate. This is a very accurate and useful indicator for identify mean-reversion areas of value rather than actual turning points. It can also be used as a trend measurement in conjunction with the ADX (Average Directional Movement).

DVIM (Intermediate Magnitude):

This is the other component of the DVI that looks at the magnitude of up and down movements and is smoothed to reduce noise and also re-scaled to create an oscillator that appears to move in a “wave” format. The DVIM is a great complement to the DVIS because it measures the net percentage move over a variety of lookbacks weighted primarily on the intermediate time-frame. Like the DVIM, it does not attempt to identify turning points, and is instead a measure of relative value that does not change as frequently as other oscillators. The “wave” movement of the indicator is ideal for “hook” type mean-reversion strategies that buy or sell following a transition from overbought or oversold levels. It can also be lengthened to create an excellent trend or relative strength indicator.

DVI (Intermediate Oscillator):

The DVI is blend of the Intermediate Stretch and Magnitude Oscillators.

Performance of the DVI(s):

Trading the indicators as follows on the SPY:

• Long: Reading below 0.5
• Short: Reading above 0.5

DVI:

We can see that the DVI is enjoying a purple patch after an indifferent period. Still the 2517 day CAGR and 250 day CAGR both look respectable at around 15%.

This isn’t the full story though as the DVIS and DVIM have been running at different speeds of late.

DVIS:

The DVIS has not performed well in the last 250 days and appears to owe the bulk of its past performance to riding the credit crunch madness very well.

Short misery: When you isolate short trades only, we get to the crux of the problem. Here the indicator’s faults are laid bare. It caught the credit crunch short side very well, but has done little else before or since.

DVIM:

By contrast, the DVIM has been off at the races, and is clearly the driver behind the DVI’s recent performance.

When we isolate short trades only, we see that the indicator’s past performance still owes a lot to the credit crunch, but showed signs of life before this and is catching good moves now.

Summary: 

The implications of this are that on an intermediate term basis, magnitude offers a better measure of markets over extending than a time stretch measure.

All the indicators researched here are available within the DV Plugin. 

What's working right now?

There's no doubt markets are behaving in a different than they have in recent history. Some short term indicators like the DVB have fallen out of sync with markets. While there are some ways to filter and adjust to market conditions to turn off certain inputs, it’s useful to know what is actually working right now.

Luckily I’m privy to a great tool my colleague Jeff Pietsch has in beta right now. A performance matrix displays various indicators across different ETFS via 252 period Sortino ratio. It’s a very neat, visual way to see what is working on which market.

Here’s a sneak preview:

As you might have guessed, the shortest of the short term indicators have not been doing well, but looking a bit further out (DVI, DVIM etc), markets are still mean reverting – It’s just not the sort of mean reversion that the likes of the DVB thrives on though.

The longer term DVI, DVIM and DVIS have fared better on the main markets of late, but the real star has been the Livermore Trend Indicator which has shone while the DVB has faltered. All of these indicators are available within the DV Indicators package, a product I must disclose I am involved in.

The simple solution would therefore be to switch your attention to what is working and step away from what isn’t. But we all know this is incredibly easy to do with the benefit of hindsight, but rather more difficult in reality.

In the next few weeks I’ll be running a test of a 252 period Sortino rank method for indicator switching and will report back when I can on this rather mammoth undertaking.

One solution going forward might be to hedge bets by averaging indicator signals across fast and slow indicators with the addition of a trend indicator (Livermore Trend Indicator) as a final overlay.

Inputs:

• Fast: DVB (DV2 period)
• Fast: DV3 Period
• Slow: DVI (Intermediate)
• Slow: DVIM (Intermediate Magnitude)
• Trend: Livermore trend indicator

Buy on score below 0.5 sell on score above 0.5

Returns:

This still of course comes with a huge dollop of hindsight in the selection of the indicators to throw into the averages. Still it gives you an idea of what is working right at present.

Now where’s that time machine…

Equity curve surfing & other filters

Following my last post on the struggles the DVB indicator is having, a reader reminded me of this excellent post Jeff Pietsch's market rewind blog. 

Worth a read certainly. The main highlights from this post for me are these three suggestions: 

1. Know what environment you are trading in (reverting/ trending).

2. Trading methods themselves revert (addition by David V.).

3. Track your methodology's equity curve like it was a stock.

In the context of the current market, I thought it would be a good idea to put some flesh on these ideas applied to the DVB Indicator. 

Measuring Indicator volatility instead of market volatility

Question – Can you use volatility In DVB scores to modulate the indicators usage?

Historically, mean reversion type indicators have performed better in volatile conditions compared to trend following conditions (when the price does not pull back). Instead of using the market as a measure of volatility, can you use the indicator values themselves as a volatility proxy?

Trend following vs Mean Reversion Averages

Recently I've been attempting to come up with a superior system rotation model to the one I first discussed here. 

The general aim to find a method of switching between strategies as they blow hot and cold, so you don't continue to follow the strategies that aren't adapting to current market conditions. This has become more salient a problem as Mean Reversion general comes off the boil somewhat. 

There are no easy answers is my general conclusion after many hours searching!

I wondered if it was worth looking at from another angle.

What if you looked as the FTSE/ SPY/ S&P 500 etc as your primary trading vehicle (i.e. you buy and hold), but use trading strategies only when your chosen investment market has come off the boil. I.e. instead of focusing on the systems, you focus on the market primarily.

There has been a lot of great research put out there on the blogosphere on the merits of Mean Reversion strategies vs trend following. Mean Reversion generally means buying after a down day and vice versa. There are may variations on this theme but generally that is the gist of the idea behind my indicator strategies such as the RSI(2) and DVB.

I've not seen many compelling trend following strategies with the exception of things like the Golden cross or Ivy portfolios 10 month moving averages. 

As it stands though, for trend following systems or strategies, simply buying and holding the market has it's benefits. First and foremost - It's simple and when markets catch a trend, it can run and run.

Using DV Indicator's excellent tools + some excel crunching, I examined the returns of 6 Mean Reversion strategies on the FTSE 100 from 2001. 

These were: DVB, DVSC, Rolling RSI, DVO, 2 period RSI & the Ultimate Oscillator. 

I used the basic settings for each which meant selling above 0.5 or vice versa. Trade length was one day, rolled over if the signal continued.

 

As expected, the MR methods on average beat the FTSE 100. Of interest though was the fact that when taken as an average, the MR methods had a standard deviation, not only below the FTSE 100 (4.69%), but of all the MR systems individually. The MR method with the lowest Standard Deviation was the Rolling Relative Strength with 5.11%, while the biggest was DVSC at 6.38%. Averaged though, the MR methods had a standard deviation of 4.22%. 

The total returns from the average was lower than some of the top performers as you'd expect.

Next I looked at when you might turn on these systems vs when to simply ride the FTSE. 

I found that the bulk of the MR system's returns came when the FTSE's 5 month return was negative. Nothing special about 5 months, returns for 6 months were similar. 5 months was just the time period I first chose.

 

You can see that virtually all of the MR system's performance came when the FTSE 100's 5 month return was negative.

So this sparked to following methodology:

• Invest in the FTSE 100
• Unless the FTSE 100's 5 month returns are negative.
• If so then trade multiple Mean Reversion strategies.
• The only exception is if last month's TVA is below 1 (As explored here). Go to cash.

This produces the following returns with the DVB included for comparison.

 

The method beats the FTSE and individual MR methods. The monthly Standard deviation is strangely exactly the same as the FTSE at 4.69%. However if we isolate downside volatility only, we get a more favourable comparison. Negative monthly vol on the FTSE is 3.46%, the average method is 2.23%.

Thoughts

The first thought is the real world practicality of following 6 individual methods at once. For testing I simply divided the total monthly returns from all 6 methods and divided them by 6. In reality you could divide your trading pot by 6 and have 6 separate trades at once (costly, but not all of the sub systems would trade on the same day). Alternatively you could take a daily net position based on the aggregate signal across all 6, for example if 4 of the methods signal a long and 2 signal a short, you go long 33%.

My brain is too tired to figure out if mathematically either of these should make any difference, but I'll look into it soon.

Other things I shall certainly be investigating is whether this approach works as well on the SPY (the king of MR (Mean Reversion) and a useful out sample test as such) and whether using extreme settings (0.8 and 0.2) helps or hinders returns.

Another point to consider is that there is a dollop of hindsight in this idea with regards to which systems work. Some have come off the boil like Ultimate Oscilator, but in general the selection of the systems to include is from a small biased sample set.

In general this would need to be combined with some sort of long term general MR filter.

I hope this is of interest.

Will mean reversion bounce back?

Excuse the pun with the post title (Unintentional honestly!)

I have to admit I'm absolutely stumped by the things I'm investigating at the moment.

I would normally give up and move on, but I believe these things have implications for robust adaptive system development.

It concerns the Agg M indicator which I've blogged about a few times.

Aggregate M Indicator

As a quick recap, the Agg M works on the principle that markets are mean reverting in the short term but trend following in the longer term. The Agg M blends these two principles, with the longer term trend helping to smooth the more profitable but volatile short term mean reversion.

I wanted to have a look at making the Agg M adaptive, mainly because it broke down in May.

This is based on the bog standard settings of opening trades at the next close and holding for 1 day. Buy above 0.5 and sell below 0.5. I tested the FTSE 100 and the SPX, results were roughly similar.

'Broke down' is a bit excessive, but especially on the SPX, there were indications in May that the Mean Reversion part of the system was not firing. May saw a huge upsurge in volatility, a market condition that is 'supposed' to good for Mean Reversion, yet the Agg M had its worst month for a long time.

Month	FTSE	Agg M
Jan-10	-4.71%	-4.71%
Feb-10	2.63%	1.95%
Mar-10	7.25%	7.47%
Apr-10	-0.88%	-0.88%
May-10	-7.38%	-7.38%
Jun-10	-1.77%	3.59%

An Adaptive Agg M

I got thinking, well if the Mean Reversion part is falling down, that could be easy to manage and isolate. The indicator is built on the premise that markets are mean reverting in the short term, but what if they stop becoming mean reverting and do something else instead (or the type of mean reversion changes etc?).

So how do we track if the system is starting to become unprofitable?

My intimidate thought was to try the Expected Value levels on a rolling basis to see if it provides an means of switching the entire system on/ off or at least the mean reversion part.

To cut a long (and very boring) story short, I tried many different combination and rolling EV lengths, but applying the filter based on system performance tended to decrease overall performance, not increase it.

huh?

The chart below shows the problem visually. The problem with the Agg M, even with the recent under performance is that it is never consistently unprofitable. If anything, there is a rhythm whereby it bounces from weak performance to strong performance and back again. The data is of the FTSE 100 from 1997 to 2010.

A score of 0.5 or above means the system has been profitable for the last 50 periods.

I ran a rolling 50 period correlation comparing today's Expected Value Score with the Expected Value score from 50 periods ago.

The correlation strength waxes and wanes, but the clear bias is for the correlation to be negative. In other words a strong performance 50 periods ago is more likely to mean a weak performance today. Conversely, a weak performance 50 periods ago is more likely to mean a strong performance today.  

This makes using an Expected Value filter very hard because each time the system looked like it was about to break down, it was actually more likely to go on and make bigger profits. 

Correlation does not equal causation, but there appears to have been a relationship there.

Can we use this?

I ran a filter as follows:

Trade the indicator as normal except when today's rolling Expected Value is 50% greater than 50 periods ago. In addition, today's Expected Value also has to be greater than yesterdays. Logically with a negative correlation, this should improve performance because when the EV is 50% higher than the past, the more likely outcome is for the system to under perform, so we cut these trades out.

 

Month	FTSE	Agg M	AggMA
Jan-10	-4.71%	-4.71%	-4.71%
Feb-10	2.63%	1.95%	1.95%
Mar-10	7.25%	7.47%	7.47%
Apr-10	-0.88%	-0.88%	2.10%
May-10	-7.38%	-7.38%	-6.23%
Jun-10	-1.77%	3.59%	3.59%
Jul-10	0.00%	0.00%	0.00%

Month	FTSE	Agg M	AggMA
2001	-14.94%	24.42%	26.49%
2002	-25.78%	16.15%	23.15%
2003	15.53%	-0.70%	6.62%
2005	16.06%	13.66%	14.59%
2006	10.94%	5.80%	6.40%
2007	5.24%	-6.50%	-4.36%
2008	-31.76%	4.95%	24.08%
2009	23.36%	23.55%	33.35%
2010	-4.85%	0.05%	4.17%

AggMA = The Adaptive Agg M.

Slightly better performance in 2010, but a better total performance.

Thoughts

While interesting and potentially useful, the findings to not answer my original question on how to tell when Mean Reversion is falling apart. Neither does it provide a good way to finally say that a system has bit the dust. It seems a good way of working out when a method has over performed in the short term, but not as a switch on or switch off, which is what I was looking for.

Other thoughts:

• Indicators/ methods, just like markets have their own rhythms. Perhaps a completely unprofitable method could be made profitable by exploring its rhythm.
• Perhaps measuring the number of consecutive days a method goes with an EV under 0.5 might help as an off switch.
• Building adaptive systems is bloody difficult!

Any thoughts/ suggestions/ comments appreciated.

Position sizing with TVA

I've been playing around with the Targeted Volatility Analysis ideas presented on CSS.

Some interesting results.Check the link for more details, but basically TVA adjusts position size according to a relatively simple volatility calc called TVA. You can apply this to a market like the FTSE or in theory a trading system.

I'm not sure I've got the calc a 100% right, but here's the results of trading the FTSE 100 with it.

Note I've used it more of binary indicator which isn't the intention, but in the real world unless you are an institution, you're not going to be able to adjust your market position every day. So the TVA below 1 basically gets you out of the market. TVA of one means you stake 1 X the TVA. I didn't double exposure beyond this.

It does quite a good job:

 

The worst drawdown is cut to 1700 index points from 2930 index points.

I then wondered what would happen if you look at an indicator like the RSI2 period and traded this long only when you also got the go ahead from the TVA.

Rules: Buy only if the 2 period RSI is less then 30 and the TVA is above 1. Position size is 1 X the TVA.

 

The filter adds quite an improvement to RSI2 on a long only basis with the staking sizing helping to increase profits.

Unfortunately it hasn't helped the RSI2 method make a profit in 2010, but the losses have been marginal.

I've also combined this with an additional EV filter, but I'll put that up another day. Hope this is of interest.

A big thanks goes out to Jerry and everyone involved for sharing such an interesting idea.