# Predicting market movements

Posted by Owl on February 22, 2009 (12:09PM)

Here's a question I was wondering about: If you know what the Dow did on any given day, does that tell you anything about what it will do the next day?

Background on the practical interest of this:

• If you could predict the direction of movement of the Dow (or of any stock) on a day-to-day basis, you could reap enormous profits.
• You don't have to be 100% accurate. As long as you can predict better than chance, you can make profits.
• So all we need is a statistical correlation between Dow activities on succeeding days.
Three theories:
• One theory is that the Dow movements on succeeding days should be positively correlated: i.e., if the Dow goes up on one day, it will more likely go up the next as well. This might be true because the market has "momentum"--once people start buying, this tends to encourage more people to buy; and similarly for selling.
• Another theory is that Dow movements should be negatively correlated: i.e., if the Dow rises one day, it will more likely drop the next, and vice versa. This might be true if we assume that the Dow has a fair value, based upon the intrinsic worth of the Dow 30 companies, and that prices are attracted to that value. If so, then when the market moves above that value, it will be more likely to move down on the next day. When it drops below fair value, it will more likely move up the next day.
• Still another theory is that Dow movements should be completely uncorrelated: the Dow is totally unpredictable. This would be true on the Efficient Market Hypothesis.
• These are interesting theories. I got historical data from Yahoo to test them (PM me if you want my data).

Q: On any given day, what is the probability that the Dow moves in the same direction that it moved in the previous day?

Results: The answer depends on what time period you look at; over the last five decades, the percentage has decreased. Here are the percentages for different time periods:
• 1959-1969    57.0%
• 1969-1979    56.6%
• 1979-1989    49.9%
• 1989-1999    50.2%
• 1999-2009    46.6%
• 2005    45.2%
• 2006    49.0%
• 2007    48.2%
• 2008    42.7%
• 2009    50.0% (Jan - Feb)
Conclusions?
• Most of the deviations from 50% (the result to be expected from pure randomness) are statistically significant; they indicate a systematic phenomenon. Analogy: If you flipped a coin 2500 times, and 46.6% of the flips were heads, that would prove that the coin was not fair. Likewise, the 46.6% statistic for 1999-2009 is not due to chance.
• The 46.6% for the past 10 years, and even more so the 42.7% for the last year, support the second theory mentioned above. However, in 1959-1979, the correlation was in the opposite direction, and from 1979-1999, there was no signifant correlation. I don't know why this should be.
• The deviations are big enough to be useful. If you could predict the Dow's movement correctly 57.3% of the time, you could be a rich man. And you would in fact have done so, if you had made a habit over the last year of always betting on opposite-direction movement.
• Unfortunately, the correlation changes from year to year. If you are only correct 51% of the time (as you would have been in 2006), that is slightly useful, but it won't make you rich.
What do you think--can we make use of this information?
And why has the percentage changed over the last five decades?

Posted by incubus on February 22, 2009 (12:57PM)

One glitch.
quanitity, if on a given year the average day moved upward 55% of the time and you bet on it with an even percentage of your equity you should theoretically do well.

The problem..it would only take one erroniously down day of to wipe out your .05% average gain.

Posted by OldFart on February 22, 2009 (12:59PM)

Many financial/market blogs note that the short term mean reversal is becoming stronger in recent years, in other words, fade the Dow. Here is one of the blogs on the subject - http://marketsci.wordpress.com/2009/02/17/roundup-short-term-stock-market-mean-reversion-becoming-stronger/

Posted by Prussian on February 22, 2009 (02:48PM)

I'd still like to know whats "wrong" with using overnight indicators as sort of a predictor for the next days USA trading session. The overnight futures indices, world stock exchange indices. When they are ALL green our markets seem to have a good day. When they are ALL red it is a good day for PUT options. Granted-overnight indicators would not be good for anything long term.

Posted by Owl on February 22, 2009 (02:58PM)

Inc,
If you were to use a contrarian trading strategy, I think you should bet on down movements as well as up. So you should also, on average, profit from the market drops.
I'm not sure if that was your point. Maybe your point is simply that short-term trading is risky.

Posted by Owl on February 22, 2009 (03:01PM)

Prussian,
I think you're right. But I'm worried that by the time you have this information, it may be too late to use it: when you predict an up day, the market may just open higher, rather than rising throughout the day. So your prediction is right, but you don't get to use it. I don't know whether this is true or not. I haven't yet done a systematic study of this.

Posted by incubus on February 22, 2009 (03:18PM)

Owl no.
My point was that simply going by the average daily performance of the market as a guide to buy or sell might be too rudimentary.

You'd do well to factor in things like witching Fridays, economic calenders regarding CPI, PPI, GDP, earnings, imports, oil...etc.

Then maybe factor in that average regarding the previous days performance.

Something to that effect.

Posted by incubus on February 22, 2009 (03:25PM)

Owl, I just copy/pasted this because you have a serious point there.

Prussian,
I think you're right. But I'm worried that by the time you have this information, it may be too late to use it: when you predict an up day, the market may just open higher, rather than rising throughout the day. So your prediction is right, but you don't get to use it. I don't know whether this is true or not. I haven't yet done a systematic study of this.

I've lost money on days the market had gone up substantially as I bought into what appeared to be a rally and watched the day lose a portion of it's gains.

I also make money on days when the market goes down, Friday was a perfect example.

The lowest day since 1997, I made a nice gain going long at around 1:30.

Posted by incubus on February 22, 2009 (03:28PM)

Prussion, which overnight indicators are you referring to?
You mean futures?

Posted by Owl on February 22, 2009 (03:33PM)

Oldfart,
Thanks, that blog you mentioned is just the sort of thing I wanted to read. Very interesting.

Inc,
Yes, you're right. This is only one factor among many.
Prussian is probably thinking of: http://money.cnn.com/data/world_markets/

Posted by incubus on February 22, 2009 (03:42PM)

Owl, I have that in my favorites along with http://www.cnbc.com/id/15839171/  and  http://www.bloomberg.com/markets/stocks/futures.html .

These are futures and world markets, indicators would be something else, unless prussion used the term loosely.

Posted by Prussian on February 22, 2009 (05:40PM)

Owl is correct in my thinking-cnn money example- also, the after hours and pre-market hours XDow, NAS, SP futures indices

Posted by incubus on February 22, 2009 (05:54PM)

Prussion, I find futures are an especially good way to gage market reaction to after hours news, earnings reports and other late developements.
An example was two Fridays past, the final negotiations on the stimulus were complete, but at the last minute Senator Dodd threw in additional measures to further decrease execs salaries who sought aid from the package.

I wasn't completely sure, but I figured that would cause a negative reaction.

Sure enough the market plummeted over friday night and this past Tuesday morning.

I even stated we had a "wild week" coming, I knew what to expect this week within a reasonable level.

Posted by gumbo on February 23, 2009 (02:43AM)

There are a number of technical indicators that can assist you. Specifically, bollinger bands, full stochastics, CCI and candles.  Candles are the best indicator of where the market's mentality is.

An example is where the DOW has been since 01/01/09.  Candles gave confirmation that the DOW was gonna tank on 01/02.  CCI and full stochastics said there was no confirmation of a reversal and it went sideways until 02/09 when it showed that doji.  That spelled sell off for the next day which it did.  Since the week of 02/09 there has been no confirmation but you can see how some of the chart guys have been ready to jump in with both feet to be the first in an impending rally.

If we get any kind of a green candle on Monday, 02/23, it's off to the races.  The full stochastics is projecting a nice run but one super negative story can alter that in a single instant.

But to answer your question, yes you can have a pretty good idea where the market is heading when you combine technicals and overall news imo.

Posted by gumbo on February 23, 2009 (02:50AM)

Correction:  on 01/02 confirmation was given by full stochastics and CCI that a reversal was gonna happen.  On Friday, 01/16 you see a big red candle but Monday you got a doji.  Combine that with no confirmation of full stochastics and CCI, that's when the market went sideways until 02/09 ish.

Posted by townj on February 23, 2009 (02:17PM)

I read a good point which I think makes good advice, particularily at this point in economic history. According to a post on the Stock Research Portal Blog (SRPB), short-term upticks in financial indices cannot sensibly be explained by transitory economic news, nor are they of interest to serious students of the market. For example, a 1% increase in the U.S. market was attributed by the Associated Press (quoting a “senior investment strategist”) as being due to an optimistic housing report, some decent corporate earnings and a talking-up of the U.S. stimulus package by Treasury Secretary Timothy Geithner. But SRPB counters that one-day changes are largely inexplicable and that investors are better off considering likely changes over the forthcoming six to 12 months. Good Luck