can numbers catch criminals

ive been watching two programs on AXN quite regularly Numbers and CSI . Numbers has a couple of mathematicians(the main guy keeps thinking out loud for the benefit of the viewers with enough jargon to impress anyone) who help out FBI catch criminals using math, mostly talk on random numbers, probability, and related stuff. CSI is similar thing about catching criminals but they showcase a complete investigation right from the analysis of every piece of evidence gathered at the scene of crime to linking it up with suspects. One thing i appreciate about the makers is the research they do on a subject before they make any program, unlike our Hindi serials where everything is possible. These guys show many techniques in detail like using triangulation for finding out exact position from where a pic is taken given the pic and control points (like a landmark or building in the pic), finding out possible location from where a bullet was fired by finding the depth to which bullet pierced a wall and the angle it made while piercing and of course dna fingerprinting.

I am not sure if american crime investigators really are as efficient and intelligent as shown in CSI, but surely our own Indian police or CBI is far from being there. Chaos rule places of accidents and murder where most of the initial evidences can be gathered .. everything gets ignored in this chaos. weve got narco analysis tests done on criminals but the results cannot be used in court of law unless the criminal later on accepts the version he uttered during the test.

CSI is convincing , but watching Numbers leaves me trying to justify the methods, guess it is because the mathematician leaves off the details about the problem and gives a fuzzy picture of his technique.. but uses enough jargon to keep you impressed.

Like in a recent one there is talk about finding out patterns in seemingly random and unrelated traffic accidents (which the lady agent is not convinced are a result of road rage),here he speaks of possibility of some chain of events to be too random.. now this talk about qualifying a random process to be 'too' or 'less' random would lead to a discussion on meaning of randomness. When you try to generate random number from a finite state machine like a computer , you get what is known as pseudo random number (due to the finite states that a computer can be in and any thing generated using any of these states may not qualify as random in pure theoretical sense , but may suffice to be enough random for practical usage) and may be an attempt to qualify randomness of these computer generated randomness is allowed .. but if it is occurence of natural events then the qualifier may not be apt. Is the sequence of births in different parts of a country in a given time frame being male of female ... or the distribution of weight of the babies taken on a scale be qualified as too random or less random.

Numbers fails to convince on many other details , like this one ....collecting input data from traffic signals on traffic load , satellite monitoring, and feeding it to a modeling software which will do your work on building intelligence from these inputs .. i doubt if someone has really developed this kind of s/w. If it is a known distribution and you are trying to take the data and fit it to this then it is understandable but otherwise as in this case .. i keep looking for answers. Or am i unnecessarily taking this program too seriously :).