STATISTICAL Data Analysis : Comparing Stocks ??? Help....

[Ian McVicar <icm63@xxxxxxxxxxx>]

I am building my own Stock scanner...need help with stock data comparison
theory.

Is there any one with PhD of maths out there ????

Sample base : 400 stocks over Average volume 1.5 M (50 days), 5 minute data

Goal : Compare stocks trending form. To find nice trending stocks, with little
swings and fluctuation for current day only.

Method : Draw a linear regression line for todays data, take the difference
from stocks close for each 5 min data point from the linear line,  from these
difference values find the standard deviation. Use the standard deviation to
compare stocks, lowest StDev will be the one with least fluctuation.

QUESTION : Normalisation of data : I could normalise all stocks 5 min close
data between 1 and 100. Would this produce better results in comparing a stocks
that have values between $20-$30 to one that is btw $50-$60 ?
However I find that some stocks with good form (low swings, nice trend on a
trendline) do not get the best StdDev, as there minor swings are expanded, this
is the same for stocks with a bit of swing the normalisation shrinks the true
effect. So should I normalise or not ? 

QUESTION : Method : Do I have a sound approach, is there a better way, please
pass ideas on .

Thanks In advance

Ian McVicar