Price movements in the same direction?

Ever since I started to trade, I used to wonder if the price movements in one direction increased the probability of price moving in the same direction the next day? With the help of computers, I was finally able to test this.

Data used:
Dow Jones Industrial Average from 1897 - Daily data. Total bars = 28,975 bars of data. It is well known that larger the sample size, more reliable the statistic.

Condition:
a = If Yesterday's close > (or lesser than) Close the day before And Today's close is > (or lesser than) Yesterday's close, 1 is plotted, otherwise, 0 is plotted.
The sum of a is added and divided by the total sample size (bars used).
I have used closing price here because I have only data for close in DJIA from 1897 - 1970.

Inference:
From 1987-2002, there existed a probabibility of 52.1914% that price will move in the same direction it did the previous day. Putting it in other terms, 47.8086% of the time, it moved in the opposite direction.
Assuming the next day's price is completely random, the probability it will move in the same direction should be 50% and the probability that it will not move in the same direction should be 50%.

Though by mere sight, it does not seem to be a significant difference from the expected figure of 50%,I have not done a statistical hypothesis testing to validate if it is a significant from the 50% mark or not. I have attached the test result.
I request the members to comment on this.

Obliged.

hi,

what if we only consider

yesterday's close > close of day before and today's close > yesterday's close

since from 1987 , your study market is up with minor reactions?

i think than the probability would increase.

my point is that once we establish a trend, we can ask computer to take the signals in line with trend and increase the probability!

~Winston

Winstonn , i came to the mkt at Nifty 800 points,see with just (+)2.19 % winning chance it has traversed to > 500 % ,that is his point.Any system having > 53 % may outperform mkt.
Likewise [ sum(ref(c,-1) > ref(o,-1),500)/500 ] & see chances of open over close ,for better Exit / Entry with any 2 systems "With the Trend" or "Counter Trend" .

Good find.

The data is tested from 1897 and not 1987.

One thing I have not taken into consideration is the magnitude of the rise or the fall. If the rise is sharp and the fall is small, the direction of price movement would be up with blips but that is not what I intended to show through this study.
Through this study, the question I put forth was, is today's price movement a variable which affects tomorrow's price movement?
Assuming no other variable affecting the event, the probability that the price will move in a particular direction is 50%. Now, if we assume direction of today's price change as a variable, the probability that the price will change in the same direction (tested with 100years of Dow data) is about 52%...not significantly different from the 50% mark (by look of the naked eye).

I think the probability would increase with

yesterday's close < close of day before and today's close > yesterday's close

Regards,
Kalyan.

In a national poll (for a vast electorate like India) a mere % swing results in land slide victories for parties operating at national/state level.
A 65 % Win (pyramided) results into enormous % rise in overall equity .
A 2 % positive bias brings Nifty to + 500 % level.
The key here is time dimension,in election it is vast number,so a little percentage (probability) edge leveraged properly skews the rusult.
That's the essence.

This one might also be worth a try

Today's close > Yesterday's close AND Today's close <= 0.65*(H-L)

Regards,
Kalyan.

Nope. That one will catch the reversal days as well. Try this one instead

Today's close > Yesterday's close AND Today's close >=0.35*(H-L) and <= 0.65*(H-L)

Regards,
Kalyan.

its a good thread. carry on friends.