|
|
|
|
 |
Daily Speculations The Web Site of Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter; a forum for us to use our meager abilities to make the world of specinvestments a better place. |
Write to us at:
(address is not clickable)
2/3/2005
Do Charts Lie?, by James Sogi
Thanks Mr. Gardiner for recommending Edward Tuft, The Visual Display of
Quantitative Information and the two other books in the series about
statistical graphing. "Escaping Flatland" recalled one of my first
articles here. Does your chart communicate necessary information or
contain extraneous misleading elements? What information do charts
convey? Do they mislead?
Tufte says "modern data graphics can do more than substitute for small
statistical tables. At their best, graphics are instruments for
reasoning about quantitative information....Excellence in statistical
graphics consists of complex ideas communicated with clarity, precision
and efficiency." At their worst, graphs can mislead, present false
conclusions and hide valuable information. Show the necessary data,
maximize the data ink ratio, erase non data ink and redundant ink.
Remove the "chartjunk". First, remove all the distracting grids. They
convey no necessary information and obscure the relevant data and cause
the whole chart to 'vibrate' by optical illusion.
Take the simple bar. Studies have shown that people often misperceive
the size of an object, due to context, or the comments of fellow
observers, or due to illusions such as the Necker Illusion. A bar may
look long or short depending on the scale or context. For example, a 1
minute bar and 1 day bar bar look the same, but they do not convey the
same amount of information. A series of 1 minute bars appears to show a
trend, but may in fact mislead one into taking an apparent 1 minute
trend, when the day trend is the opposite direction, resulting in
consistent losses. Been there? The 1 minute chart leaves out the
context, what happened last week, or yesterday. Including a week or
two or more of price action on the screen provides much more
information than 1 minute charts, even if your trades are only 5 minutes
long. Context in a chart conveys more information than just the prior
prices, but reveals the relationship between prior price action and
current action. When combined with the data export function to export
the tables to statistical programs such as R, you have potent tools.
In Practical Speculation, Kenner and Niederhoffer deflate the
effectiveness of a number of technical chart indicators. Clogging up a
chart with indicators conveys no useful information, obscures the
necessary information and misleads. The eye has a tendency to detect
patterns even in random data and the illusion causes losses. I've
removed indicators from my chart.
The Candle has redundant information and unnecessary ink. Each side of
the candle can be replaced by a single line. The top and bottom by a
single bar. The filling adds no information beyond the open and close,
and takes up visual space and adds clutter. Lines miss the highs and
lows and for that reason I find are impossible to trade with misleading
one into turning, when it is only a false turn.
What information do charts convey? The change in price. When price heads
down, the bears and media conclude, economy set for destruction, slide
starts. That is the conclusion. But the context is Dimson's century, Sogi's millennia. Looking at four months, instead of one week, one might
conclude that January's 4% drop was a merely reversion to the mean.
Other's with a foot in the grave look at the drop as the opportunity to
pull out their canes, and hobble to the brokers office. Show the
necessary data, maximize the data ink ratio, erase non data ink and
redundant ink. Remove the "chartjunk". It will improve your bottom line.