Trading is about risk estimation that is to say probability of extremes. Today statisticians working in financial field try now to focus on extremes studies instead of mean studies using for example fractal models like Mandelbrott see his article: "A Multifractal Walk Down Wall Street" http://www.elliottwave.com/education/SciAmerican/Mandelbrot_Article2.htm "These techniques do not come closer to forecasting a price drop or rise on a specific day on the basis of past records. But they provide estimates of the probability of what the market might do and allow one to prepare for inevitable sea changes. The new modeling techniques are designed to cast a light of order into the seemingly impenetrable thicket of the financial markets. They also recognize the marinerâs warning that, as recent events demonstrate, deserves to be heeded: On even the calmest sea, a gale may be just over the horizon." Further more a trend is defined by higher high or lower low by traditional technical analysts and my model confirms their point of view although before I discovered this model I thought it was rather ridiculous because probability law of highs and lows is highly unstable and normally difficult to estimate (that's why mandelbrott and other researchers work on this subject ) compared to the probability law of the mean ... but this suppose as premisce for this logical statement that the underlying law is truly stochastic and it is not the case finally my model shows that the law is or at least has a predominant deterministic law component hidden inside the stochastic apparent noise. More generally many things can be discovered by looking at the implicit premisces. People generally focus on the explicit conclusion they rarely care about the premisces so the inevitable errors when they apply the conclusion whereas the premisce is not true. The practical usefulness of high and low of the day is for example the famous ATR entering into volatility estimation for breakout system. Why use ATR instead of standard deviation ? Because standard deviation is underestimating volatility if people usually determine statistical zone with a coefficient of 2 standard deviations thinking that it will encompass 95% of all points. This is roughly true for mean of a sample but as said above about risk we are less interested by mean than by extremes in trading. So ATR is not used practically only because it is simpler to calculate than standard deviation although this could be the justification in the past when computers had high cost
I think a more plausible answer is that Yahoo data is basically crap. I noticed some discrepancies in their option data once and sent them an e-mail. The reply I got back essentially said "well, we're really sorry but it is a known fact that we're having problems with our data...".
>"it is a <B>known fact</B> that we're having problems with our data" they answered that ?!!! Next time I will ask them what algorithm they use to produce such error if it is more difficult than take the real high since they show quote in real time also how can they get a quote that has never been quoted during the day as a sofware engineer I have really some difficulty to imagine The bug is not on Yahoo alone it is also on Bigcharts. So two big sites at least and perhaps others. It is since at least 3 YEARS that their quote are crapped periodically even Microsoft doesn't produce such bug such a long time . It is no more a bug it could be called a functionality In fact it is even since 4 years because the first time I remarked the crap was the morning of the eleventh September and what is funny is that the evening before that day the quote was correct. I remember that very well because I was doing calculation and it projected a crash so I looked at the quote and saw they were false. If I knew that there would be the terrorist attack I would have taken a screenshot before . So how is it that the problem persist since 4 years and they never corrected it huh ? The high and low can be very important for chartists in fact once their analysis were false and they said it was because of false datas from Comstock and that Comstock alerted them after that they have corrected the false quote. Something funny I remembered also: errors reproduced frequently after and I finally sent them an email for their errors and they replied that the errors was because of the terrorist attack . I didn't told them that just the morning of the attack day so before the attack the quote was already crapped. And how can a terrorist attack affect historical quotes ... until today .