mindmoneymarkets

Big days & the 200 MA

Just a quick note to share something I've been looking at.

I'm sure this has been noted elsewhere but I thought it was interesting in reference to the research on selling extreme indicators such as the Williams Vix Fix. The research I've made on this so far is based on holding for one day only i.e. short term moves.

I believe VIX or VIX spikes or indeed many mean reversion indicators tap into a common stock market pattern whereby the biggest daily gains happen during down trends.

Naturally this is a function of the greater volatility of these periods, but I was still struck by the pattern.

I performed a quick study on the FTSE 100 and S&P 500, looking at the top daily gains based on the index closing values.

For both the FTSE 100 and the S&P 500, going back to 1987, all the top 10 biggest daily gains have occurred when the previous day's close was below the 200 period moving average.

Here's the FTSE top 10:

Top 25:

The S&P500 shows a similar pattern, but with more days in the top 25 that were above the 200 MA.

If you look at the top 10 and top 25 biggest losing days, you'll also find that all but two of these occur when the market is below its 200 period moving average.

So values below the 200 MA are certainly more volatile and offer more danger in the form of big losing days.

However there is another danger during these times: That investors pull their money out and miss the very large gains that can occur during the depths of such periods. 

This is where mean reversion/ capitulation indicators appear to come into their own, perhaps showing a longer term investor when not to cash in his chips (maybe just before a big rally day) or the short term nimble trader high risk/ high reward buying points.

William's Vix Fix Update

I've been away for a while hiking the West Coast trail on Vancouver Island. More a 6 day assault course than a hike, but rewarding to get to the end.

On returning I was pleasantly surprised to see that Michael of Market Sci has performed some further analysis on the Williams Vix Fix, with Trading The Odds also performing some fascinating analysis.

Frank of Trading the odds also reminded me that I haven't updated the 'about' page on this blog, something I'll rectify soon. Every time I try to do things like an about page or a blog roll I get lost in typepad so leave it for another time to do, so these parts of the blog are well over due.

So a quick 'about' - My name is Dave Evans, I've a background in psychology but somehow have found myself immersed in financial markets. I write about financial markets for a few publications and a broker as well as trading my own account. I'm a Brit who has been working from Vancouver in Canada with my fiance for the last 12 months, but am on my way back to the UK next week just in time for the lovely British winter.

I'll add more flavour soon.

I've recently been playing around with runs on the Vix fix where the market reaches extreme values for more than one day. Early days, but there seems to be something in this angle. I'll report back soon.

S&P 500 At Volatility Extremes

As a follow up to my last post looking at volatility extremes on the FTSE as measure by the Williams Vix Fix, here's how the S&P 500 measures up:

I used the Williams Vix Fix (WVF) as a Vix type volatility applied to the S&P 500. I then performed a 30 period percent rank of the WVF. I then looked at what would happen if you were to buy the next day after a particular level is reached and stay out of the market otherwise.

Firstly I looked at high volatility measures, when the WVF had a percent rank of 0.75 or above. You can see the results below:

S&P 500

Data is on the S&P 500 back to 2000. No costs, slippage of dividends. I'm just looking for trends here.

The results are similar to the FTSE 100 with buying at volatility extremes working out to be a profitable strategy, albeit with increased volatility.

However buying only during extremely quiet periods has been an poor strategy. Sure its less volatile, but its also less profitable.

 

Here's a comparison of two extremes:

Buying only when the WVF PR was above 0.95.

Sorry I've just noticed the graphs still say FTSE 100. They are not, they refer to the S&P 500 (or Russell, or Nasdaq later)

Buying only when the WVF PR was below 0.05.

Russel 2000

There is a similar effect on the Russell 2000. The improvement in performance at high volatility extremes is less marked, but the impact of buying at low volatility appears to be more consistent (i.e. worse).

Here's the results of buying the Russell 2000 any time the WVF Percent Rank dropped below 0.2. 

Sorry I've just noticed the graphs still say FTSE 100. They are not, they refer to the Russell 2000

Nasdaq 100

There is also a very similar effect on the Nadaq 100, though this is the opposite of the Russell 2000 with the benefit of buying at high volatility extremes increasing and the losses at low volatility extremes decreasing. Interestingly, buying the Nasdaq at volatility extremes actually decreased the average standard deviation of each trade.

Mean Reversion effect?

Is this another example of mean reversion applicable to just the last decade?

The effect is stronger now, but there was still some evidence of its impact between 1990 and 2000. The near relentless bull market during this period no doubt nullified the effect somewhat.

Here's the same test on the SPX from 1990 to 2000.

FTSE 100 at volatility extremes

Further to my last post on the Williams Vix Fix I thought I'd share some interesting angles I've been looking at.

I've taken a few passes at this, but this is the simplest way of expressing what I've found.

I used the Williams Vix Fix (WVF) as a Vix type volatility applied to the FTSE 100. I then performed a 30 period percent rank of the WVF. I then looked at what would happen if you were to buy the next day after a particular level is reached and stay out of the market otherwise.

Firstly I looked at high volatility measures, when the WVF had a percent rank of 0.75 or above. You can see the results below: 

Data is on the FTSE 100 back to 2000. No costs, slippage of dividends. I'm just looking for trends here.

As the volatility percent rank increased, so did the returns. Since the starting point, the FTSE is down 18.64%, but buying only when the WVF percent rank was greater than 0.75 produced a positive return from fewer trades.

There is a trade off as you can see. Trading at points of greater volatility implied by the WVF naturally produces more volatility so the standard deviation increases as you move up the extremes as does the average loss etc.

Now let's compare this to volatility extremes in the other direction: 

This time with the percent rank 0.25 or below in increments.

This I didn't expect, trading low volatility extremes would have increased losses. The standard deviation is reduced, but the strike rate (up days) drops from 51.24 to 50.66 then as low as 45.59% as you move down the volatility extremes.

To see how this worked out over time, here's the FTSE 100 going long only when the WVF 30 day percent rank was greater than 0.95 (top extreme score)

Now here is buying when the PR was below 0.05:

 

Not much time to wrap up with a conclusion, but I've noticed how the effect differs on other markets. I'll follow up soon.

The Williams VIX Fix

There have been a number of useful research posts put out there recently from ETF Prophet, Woodshedder and CSS.

I wanted to have a look at the FTSE and see how it compared. My eventual aim is to look at how it impacts Mean Reversion strategies, but first I wanted to see how it impacted returns on the FTSE 100 itself.There are various ways to measure volatility with proxy being the VIX volatility index. The problem is that there is no UK VIX and it would be a pain to apply the VIX to the FTSE as I'd have to manually align trading days/ holidays (unless some clever chap/ chappess knows a way of auto sorting dates in excel so they'd match)

As a measure of volatility I used Larry William's Vix Fix.

This calculation aims to mimic the VIX so it's particularly useful for markets that the VIX does not related to such as the FTSE. Calculation and notes are all in William's article. 

Here's the Vix fix on the FTSE 100. It's not perfect, but it appears to spike in the right places.Williams uses a 20/ 22 day lookback period hence why the value goes to near zero sometimes when the VIX does not. 

I've done some analysis on the FTSE 100 splitting the VIX Fix levels into quartiles but need to double check the data as there are some important conclusions.

The VIX Fix could be a useful volatility proxy for any market. I've put the calc in a spreadsheet here with the data from the FTSE 100:

Download VixFix

New Quant wizards?

Well there was no bounce on Friday, but a few things seemed a little odd to me. Not that the markets 'should' have bounced, indeed the trendline break and the potentially game changing news announcement were a red warning lights in retrospect.(S&P futs up 0.5% Monday as I write this though)

However, there's something else that doesn't smell right to me. Firstly, as you can see on the chart below, in recent times, the Aussie and New Zealand dollar have been a good proxy for risk taking. These two currencies have the highest interest rates in the Western world and are the weapon of choice for carry traders.

 

Correlations/ relationships like this wax and wane but I did find it odd that the risk proxy currencies finished the day unchanged with a muted reaction across other currencies as well. 10 year notes also showed a muted reaction relative to that of the Dow & S&P 500. It could imply that the move on Friday was not a general flight from risk that the VIX jump implied (imperfect as this is), but a concentrated slump in equities and in the financial sector in particular.

It's a small sample set (A slump in oil prices argues against this for example) so these are just musings, which lead me to one conspiratorial thought. In Barry Rithholz book on credit crisis he writes about the mythical plunge protection team which some have speculated may have links with Goldman Sachs. To take this on the fly conspiracy theory to its extreme, could certain players have thrown the metaphorical rattle out of the pram on Friday in order to influence the final outcome of the still to be solidified Volcker plans?

Speaking of the unquantifiable, Friday's plunge did make me think about pure quant trading vs adding discretionary layers. Had I been actively trading on that day, I would have been wary of buying the dip on Friday because of the trendline break and unknown impact of the Obama/ Volcker plan. This is of course in retrospect as I was entirely out of the market so unencumbered by thoughts of a position or potential position, but it does raise some interesting thoughts.

In the original Market Wizards book by Jack Schwager, few of the traders interviewed were pure quants. This may be due to the lack of readily available computing power back then, but it made me think. Does the human brain's ability to adapt and sense what it cannot articulate mean that discretionary trading or adding some discretionary layer to quant trades will always have an edge over pure quant? Or does quant trading allow us to override our worst instincts and buy into dips like Friday? The human input is therefore the discretion in knowing when to drop a quant strategy and the research and hunches that go into its creation.

There does seem to be an emerging new wave of quants, at least in the blogosphere. Is it time for a 'Quant wizards'?

By Quant here I'm not referring to dark pool HFT type stuff. 

The Vix is very predictable

Inspired by this post I took at look at applying DV's Super smooth stochastic to the VIX.

 

The question is, can be turn this into a profitable strategy?

It's well known that the VIX is negatively correlated by the vix, but this correlation is not fixed, the correlation varies between near 100% and randomness.

Here's the spreadsheet. For the vix I've exported the data from Esignal.

Download VIX

Fairly consistent in being negative, but this correlation waxes and wanes, could this correlation be used to fine tuned any vix related strategy?

I hope to find out soon.

Some related strategies.

http://marketsci.wordpress.com/2008/09/22/test-of-condor%E2%80%99s-vix-based-trading-strategy/

http://davesbrain.blogs.com/mindmoneymarkets/2009/12/trading-edges.html