using RANDOM + MOD is not a good Random nummber generator

drawfail.sh

# Prove using RANDOM + MOD is not a good Random nummber generator
#
# CATEGORIES == the amount of "winners" in the experiment
# @author freakingtux
CATEGORIES=707
EXPERIMENTS=100000


if ! which gnuplot 2>&1 > /dev/null
then
    echo "This script requires gnuplot"
    exit 1
fi


# problem # $RANDOM returns a different random integer at each invocation.
# Nominal range: 0 - 32767 (signed 16-bit integer).
# if you assume that every number is draw as often as the other you will find
# that the first people will be "more lucky" as there is a bigger chance the number
# end with a value lower then MAX % CATEGORIES in this case 32767 % 707
# the first 245 contestants have a bigger win factor !
COUNT=0
while [ $COUNT -lt $CATEGORIES ]
do
    data[$draw]=0
    COUNT=$[$COUNT +1]
done

COUNT=0
while [ $COUNT -lt 32767 ]
do
    draw=$[ $[ $COUNT % $CATEGORIES ]]
    data[$draw]=$[${data[$draw]} +1]
    COUNT=$[$COUNT +1]
done



cat <<EINDE

echo This shows the people at the top of the list have more chance of winning.
echo because theire number will be drawn more often.

EINDE

(
COUNT=0
while [ $COUNT -lt $CATEGORIES ]
do
    echo $COUNT ${data[$COUNT]}
    COUNT=$[$COUNT +1]
done ) | gnuplot -e "set term dumb ; plot \"-\" with steps "



#NOTE if you remove this next line somehow (I don't know why
#the last value
#########echo REMOVE ME and see the last value
# Perform the real exeriment
COUNT=0
while [ $COUNT -lt $CATEGORIES ]
do
    data[$draw]=0
    COUNT=$[$COUNT +1]
done



COUNT=0
while [ $COUNT -lt $EXPERIMENTS ]
do
    draw=$[ $[ RANDOM % $CATEGORIES ]]
    data[$draw]=$[${data[$draw]} +1]
    COUNT=$[$COUNT +1]
done


(
COUNT=0
while [ $COUNT -lt $CATEGORIES ]
do
    echo $COUNT ${data[$COUNT]}
    COUNT=$[$COUNT +1]
done )  | gnuplot -e "set term dumb ; plot \"-\" with steps "

gistfile2.eclass