The Deception of Bars

[kut2k2]

What is a bar? In the context of price data I mean, not in the context of booze and bouncers.

Wikipedia is no help at all.

Thankfully, there is Investopedia:

"Definition of 'Bar'

A graphical representation of a stock's movement that usually contains the open, high, low and closing prices for a set period of time.

For example, if a technical trader is working with daily data, one bar is the set of quotes for one day. In the case of one-minute data, it is the price data for one minute."

In short, a bar is an attempt to summarize price activity over a time span by reducing said activity to 4 key data points: the first price, the last price and the extrema.

The problem is that a bar is a semi-ordered data set that gives the illusion of full information. We know our starting point and our ending point but the only things we know that happened inbetween is that a given high was reached and a given low was reached. As to times or even order (did the low precede the high or vice versa?), we are completely in the dark in the vast majority of bars.

It has been my experience that the TA that works will most often resemble time-series analysis, where there is no question as to the order or the times of all the data points. There is no guesswork about the data in a true time series.

So mixing time-series analysis with bar analysis as some TA methodologies do will typically lead to inferior results IMO. Time series analysis is not designed to deal effectively with semi-ordered data sets, and jamming these two different constructs together either leads to failure or a new type of analysis that is rare or nonexistent in technical analysis today.

Constructive feedback welcome.

 

[dbphoenix]

I wouldn't call it "new" as it dates back to the nineteenth century. As to "rare", yes. "Non-existent"? Just about. The pioneers of technical analysis began as tape readers, though one could go farther back to the days before tape, when quotes were posted on a board. Reading this continuous flow of price became second nature, so that at the end of the day, if they chose to summarize their observations in the form of a bar, they knew exactly what that bar represented: all the individual trades that took place that day. Thus when newspapers started publishing charts, these traders didn't see "bars" but clusters of transactions, just as someone who understands art sees not a "painting" but colors and tonalities and brush strokes and composition. This is not what beginners see today, particularly those who've been told that a bar of less than a certain interval is "noise".

Candlesticks are much worse, giving the impression that the action takes place in the "body" and that the wick -- or "shadow" -- is only of minor importance. Some candlestick traders plot only the bodies. But if they were to look at the transactions that created that candle, they might find that the bulk of the trades took place in the "wick" and only a few in the body. In fact, most of the transactions may have taken place at the high and the low during that interval with very little inbetween. Whether or not this contributes to the general inability to trade price is another subject.

 

[Sergio77]

So mixing time-series analysis with bar analysis as some TA methodologies do will typically lead to inferior results IMO.

I think you are giving too much credit to TA. Most TA is ignorant of time-series analysis. An average is not time-series analysis. You must go up to Fourier analysis and beyond for that. I think for longer term trading analyzing series with the close is enough. But in principle you are correct in that what happens "inside a bar" is important but mainly to short term traders.

 

[dom993]

In short, a bar is an attempt to summarize price activity over a time span by reducing said activity to 4 key data points: the first price, the last price and the extrema.

The problem is that a bar is a semi-ordered data set that gives the illusion of full information.

Duh?!

In some market models I do reconstruct the bars from lower timeframes so that I can have more insight re. what's going on in a bar, the H/L order being the most obvious.

In other market models, I just get rid of the notion of bars, go to a very small timeframe to get enough accuracy when processing those bars into the model.

 

[murray t turtle]

[Partial QUOTE=dbphoenix;3954859]I wouldn't call it "new" as it dates back to the nineteenth century. ....................................................
Candlesticks are much worse, giving the impression that the action takes place in the "body" and that the wick -- or "shadow" -- is only of minor importance. Some candlestick traders plot only the bodies. But if they were to look at the transactions that created that candle, they might find that the bulk of the trades took place in the "wick" and only a few in the body. In fact, most of the transactions may have taken place at the high and the low during that interval with very little inbetween. Whether or not this contributes to the general inability to trade price is another subject.[/QUOTE]
===================
Good points;
IBD/charts [William O Neil]still,after all these years, prefers barcharts over candles.

And @ the end[close] of the day..... end of the month;
everybody voted.There is a market for selling burned out candles[long, burned wick @ top, most of candle is burned- gone].

Some may prefer to buy burned copper ,, rather than burned out candles.But it takes all kinds to make a market;with enough increasing volume, enough days, even burnt out candles can be bought cheap/sold @ profit