Tuesday’s Monster Rally Significant?
I heard two opposing reactions to the huge 6.4% rally on Tuesday. One was the CNBC’esque “the bottom is here” battle cry. And the second was the (rather sophisticated I think) conclusion that these type of big up/down days tend to come in bear markets NOT bull markets, and thus, portend nothing.
I’ve touched on this subject before, looking at volatility and overnight gaps in up vs down trends (the conclusion was that, by all measures, the market is more volatile in bear markets), but in this report, I want to look at the likelihood of big up or down days like we saw Tuesday.
For this report, I’ll use the zig-zag idea I’ve used in the past to define the trend.
[click to zoom]
[logarithmically-scaled]
The graph above shows a sample snapshot of the zig-zag on the S&P 500 from 1960. Bear trends are red and bulls green. The zig-zag switches between the two trends whenever the market moves against a high or low by 20% or more. Note that the zig-zag “peeks” into the future, meaning you couldn’t use it to trade, but that’s not the point. The point is to understand what the broader trend was at any point in the market’s history, even if trend-following indicators (ex. 50/200-day moving average crossovers) didn’t yet realize it at that moment.
ANALYSIS
Volatility has changed pretty drastically over the market’s history, so it doesn’t make sense to define a “big” day in percentage terms. Instead I’m going to use each day’s % change relative to the standard deviation (SD) of the previous five years. So for instance, Tuesday was up +6.4%. Using this approach, in today’s market that’s +4.5 SD (because at the moment, 1SD = +/-1.4%).
The table below looks at the percentage of all days on the S&P 500 during bullish and bearish trends that exceeded 2, 3, and 4 SDs, since 1960.
How to read this table: take the first line for example. 2.8% of all days during bull trends were up at least 2 SDs (a big up day), but counter-intuitively, 5.2% of all days during bear trends were up by a similar amount.
The table shows that both big up and big down days, regardless of how we defined them, were far more likely to happen on any given day during bear markets than bull markets. That would lend credence to the idea that Tuesday’s rally was actually a bearish sign.
But that’s only part of the story. The zig-zag I used for determining up vs down trends, tagged about 80% of all days as a bull trend. Even though on any given bear market day, big up/down days are far more likely to occur, because bear markets happened so infrequently, any given big up/down day is about equally likely to occur during a bull as a bear market.
Whew…that was confusing to write (and probably impossible to read).
In a nutshell, I think anyone who walked away from Tuesday’s rally thinking that in and of itself it meant absolutely nothing was absolutely correct. Yes, big up/down days are more likely to occur on any given bear market day, but because most of the market’s history is up trending, big up/down days are about equally split between bear and bull markets.
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Stock Market Mechanics, VIX & Volatility | 14 Comments
I’d be curious to see if big up/down days had any correlation with an impending change in trend. In other words, were these big up/down days more or less likely to occur near the end (or any other temporal segment) of a particular trend?
Great blog, by the way. I’m a new reader.
RE to newb: great question – on the to do list.
Part of the answer (but not your whole answer) is…
Because the zig-zag study peeks into the future, if for example a big up day in a down trend resulted in the start of a new up trend, it would be counted as part of the up trend (assuming the market didn’t retrace the big up day). If it didn’t result in a new up trend, it would still be counted as part of the down trend. So…implicitly, down trend big up days like Tuesday haven’t historically usualy resulted in a reversal of the trend (otherwise they’d all be counted as up trend big up days).
I hope that made sense. That part of the discussion seemed a bit too abstract for the blog, so I left it out. But I don’t think that’s your whole answer…more to follow.
michael
I’m also interested in whether large moves tend to occur near trend changes. Perhaps you could divide the up and down trends into pieces and see if the distribution depends on how far along you are. So you could divide each downtrend into 2-5 pieces (not necessarily equal in size) and see if large moves occur more frequently at the beginning, middle, or end of bear markets. Useful divisions are likely to be first and last 10-25% of the move (by time or price) and the middle 50-80%.
Alternatively, you could assign each large move a score that represents how far through the current trend move it is (so a large move on day 20 of a 200 day move would get a score of 0.1). Then you could look at those distributions to see if there is anything interesting about them. In principle they will be uniform if these days are randomly distributed. The sample size might be too small to make any actual conclusions.
Michael,
An extremely minor comment about an otherwise excellent post: I think the sentence that reads “The table below looks at the percentage of all days on the S&P 500 during bullish and bearish trends that exceeded 2, 3, and 4 SDs, since 1960″ would read better reformatted as “The table below looks at the percentage of all days on the S&P 500 that exceeded 2, 3, and 4 SDs during bullish and bearish trends since 1960.”
It may just be me, but when I read the original I thought you were talking about trends that had certain volatility, not days.
As I said, a pretty minor point indeed…
Keep up the good work — I’m looking forward to Part 2 on the performance of the SOTM report.
Great blog. I am daily reader for two weeks since I ran into this blog.
wrt your reply to newb, there is additional possibility that, implicitly, big up days in down trend may not have resulted in reversal of the trend because may be big down days in up trend that resulted in reversal of trend nullified the ‘imbalance’ of occurance. Historically (general perception) trend change occurs because of some drastic change in persception of socio-economic situation resulting in bigger than normal ‘reactionary’ move in stock market.
I am not sure I am clear…but you got the gist…
Combining big UP days with big DOWN days is confusing. Market behavior is different for these two types. Having said that I can understand why there are more big UP/DOWN days per your analysis. Bear markets are typically shorter than bull markets. So price changes have to happen in a shorter time frame. There has to be more big DOWN days in Bear market and also more big UP days as part of ‘bear market’ rallies (the final one eventually pulling it out of the bear market). The bull markets are more mellow by comparison except when a bubble (NASDAQ, oil, Real Estate,..) has formed.
Hi Michael,
You’ve probably seen this, but I just came across it and thought I’d pass it along. Mark Hulbert’s latest column (http://www.marketwatch.com/news/story/story.aspx?guid={78B75F06-E03A-4BE2-AA36-060703DC33E4}) refers to one of his columns last Fall (http://www.marketwatch.com/News/Story/seasonality-timing-system-has-acquitted/story.aspx?guid={53DF58BD-B4AC-4220-A3B5-81DAFD474155}) in which he describes Norman Fosback’s Seasonality Timing System (STS). It claims that turn of the month days and the two days preceding holidays are bullish.
– Russ
RE to jkw: you read my mind. Expect something out on this shortly. michael
RE to Devanathan: I agree when you say “bear markets are typically shorter than bull markets so price changes have to happen in a shorter timeframe”. Expect a follow up post showing something like the slope of those bearish vs bullish periods in the zig zag study. michael
RE to Russ: agreed…thanks for the link. I did my own take on turn-of-the-month seasonality here:
http://marketsci.wordpress.com/2008/12/19/turn-of-the-month-strategy/
I’ve added to the to do list pre-holiday seasonality.
michael
Michael, I searched for more on the “zig-zag” thesis and found nothing. Is there an article expanding on this hidden in your blog?
Thanks
John
RE to John French: unfortunately no – I probably talked about it more deeply in this post than in any previous one. But it’s not really that exciting. It’s just looking at tops and bottoms in the market defined by a move of at least X%. In this post that was 20%. In previous posts I’ve used 10%. I’ve tried to use the zig-zag whenever the point of the study is to understand what’s happening during a given trend rather than trying to predict a given trend. Hope that helps.
michael
If there are pre-holiday effects, it seems likely that there would be pre-weekend effects. Have you looked into that? I’m surprised that the effect is positive. Most traders presumably don’t like to be in positions during long non-trading periods. What about the pre-holiday/pre-weekend effects and their relationship to the underlying market trend at the time? Do the pre-holdiay/weekend effects magnify the trend? Or are these effects always positive? Why would that be?
RE to Russ (Blue): I’ve looked at weekend seasonality – I can’t remember the specific numbers, but trading based on the day of the week didn’t bear fruit (or at least not enough to justify a post). Having said that, I did a post a while back that looked at daily follow-through between the last and first day of the week vs other times (http://marketsci.wordpress.com/2008/10/07/weekend-vs-weekday-follow-through-does-friday%e2%80%99s-sentiment-carry-over-to-monday/). In a nutshell, unlike this purported pre-holiday effect (which I haven’t tested), the weekend follow-through effect has evolved over time as broader daily follow-through has (I explained that badly…the post will make it clearer).
I like your observation that the holiday and weekend affects should be the same (whatever that observation is), assuming that you don’t buy into the “the traders are happy because it’s the holidays” explanation. I’ll keep that in mind to discuss inthat future post on pre-holiday seasonality.
Thanks for the great comments Russ,
michael