Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Question fro you math wizards

Author: Andrew Dados

Date: 07:57:44 02/08/00

Go up one level in this thread


On February 08, 2000 at 09:25:32, Tijs van Dam wrote:

>On February 08, 2000 at 08:36:27, Andrew Dados wrote:
>
>>On February 08, 2000 at 08:05:40, Les Fernandez wrote:
>>
>>>Does anyone know given a string of "x" numbers (ie 237496) how large must a
>>>string of random numbers be so that you can have a 95% chance of finding that
>>>exact string within the larger string of numbers.  If there is a way that I can
>>>calculate it please provide info.  Keep in mind although the above example
>>>contains 6 numbers it is used only for an example so that I may understand the
>>>concept.  I am interested in being able to calculate how long the length of the
>>>longer string needs to be for numbers up to about 60 in length if this is
>>>possible. Thanks
>>>
>>>Les
>>
>>My naive approach: given x: length of your substring and y: length of string...
>>(1-1/x^10) - probability that there is *no* match in one slot;
>>(1-1/x^10)^n - probability of *no* match in n slots;
>>1-(1-1/x^10)^n - probability of at least one match in n slots...n slots means:
>>y=n+x;
>>so you need to solve equation:
>>1-(1-1/x^10)^(y-x)>=0.95
>>given I didn't have my coffee yet it's all correct with some 0.15% chance..:)
>
>I think you mixed some variables and numbers here. The probability of having a
>number the same is 1/10, so for a string of x numbers it is 1/10^x and not
>1/x^10. Without coffee, I would have made the same mistake :)
>
>When finding a string of 10 in 100, there are 91 possible starting points (there
>a 90 numbers before the last, so the last one is the 91th). So n=y-x+1.
>
>Therefore, you need to calculate y for
>
>1-(1-1/10^x)^(y-x+1)>=0.95 for the smallest integer y that is possible
>
>I did have my coffee :) and i think the answer should be
>
>y=x-1+log(0.05)/log(1-1/10^x) and round up.
>
>If you want another chance than 0.95, say p, you need to replace 0.05 with
>(1-p).
>
>To find a string of 6 with a chance of 0.95, you need a larger string of 2995736
>numbers. To find a string of 60 with equal chance, you need a larger string of
>about -[E- numbers. My handcalculator wouldn't do 1/10^60 :)
>
>Greets,
>Tijs

Below are string lengths for 0.95 chance of finding at least one match for
substring lengths 1-10: (above formula run in Mathematica):

28.4332
299.073
2996.23
29958.8
299576.
2.99574 10^6
2.99573 10^7
2.99573 10^8
2.99573 10^9
2.99573 10^10

Looks like 3*10^n is a good approximation of our formula.

-Andrew-



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.