home *** CD-ROM | disk | FTP | other *** search
- %OP%VS4.13 (28-Apr-92), The Bindon Family, R4001 0009 7093 7995
- %OP%TNN
- %OP%IRY
- %OP%PL0
- %OP%HM0
- %OP%FM0
- %OP%BM0
- %OP%PT1
- %OP%PDDotMatrix
- %OP%WC2,1558,42,1158,0,0,0,0
- %CO:A,80,0%%C%%H3%%H1%Analysis of Lottery results
-
- by Donald C Bindon (1995may30)
-
- 60 Mellstock Avenue, DORCHESTER, Dorset DT1 2BQ
-
- An article in "The Times" of Saturday 28th January 1995, reported that a
- chi-squared analysis of 12 years of the results of the Australian National
- Lotto had concluded that they were far from truly random. `Anyone wanting to
- improve their chances of winning could do so by following the trends'. The
- Times did not give any further details of this research.
-
- Suppose the results from the British machines are also less random than
- statistical expectation. We might find that some numbers come up more often
- than expected; or perhaps that some numbers tend to come up in pairs. We might
- then back the contention that the past history of the Lottery is slightly
- correlated to its future.
-
- Ignoring the bonus ball distinction we have seven balls selected from 49 each
- week. After seven weeks each ball will have appeared once on the average. But
- if after seven weeks each ball had come up exactly once, we would suspect human
- intervention. If on the other hand, one particular ball had come up every one
- of the seven weeks, we would suspect that insufficient care had been put into
- making the lottery machines random.
-
- This spreadsheet LotNos28 and its custom functions analyse historic lottery
- results to consider if they could be biased by inadequate lottery machine
- construction, or by human intervention.
-
- The lottery machines have no memory of past selections, and they are supposed
- to be completely random. So we would expect the distribution of ball success to
- be represented by a Poisson distribution, which relates the average occurrence
- of a random event to the expectation of less or more occurrences of that event,
- by an exponential series.
-
- 1 = e^x * e^-x = e^-x [ { x^0/0! , x^1/1! , x^2/2! , x^3/3! , ...}
-
- After first 28 weeks at 7 balls from 49 each week, average ball success must be
- 28*7/49 -> 4 Total balls 28 * 7 -> 196
-
- Success 0 1 2 3 4 5 6 7 8 9 10 11 Total
- success
-
- Observed 0 2 7 13 9 9 5 2 2 0 0 0
-
- Balls (O) 0 2 14 39 36 45 30 14 16 0 0 0 196
-
-
- 49 = 49*EXP(-4) * [ { 1 , 4^0/0! , 4^2/2! , 4^3/3! , ... }
-
- Poisson .90 3.59 7.18 9.57 9.57 7.66 5.11 2.92 1.46 .65 .26 .09
-
- Balls (P) 0 3.59 14.4 28.7 38.3 38.3 30.7 20.4 11.7 5.85 2.60 .99 195.53
-
- Balls
- Pois-Obs 0 1.59 .18 -3.4 .57 -1.3 .11 .92 -.54 .65 .26 .09
-
- The observed distribution of ball success peaks at 3 with 13 of the 49 balls
- having been this successful, a somewhat higher peak than the 9.57 expected. But
- there is little evidence yet either of human intervention or of poorly
- constructed machines. There is not yet any equivalent of the behaviour of the
- Australian Lotto.
-
- The custom functions CountSort and Fac are in the file c_LotMath. The former
- illustrates counting and sorting in custom function language.
-
- Custom function CountSort accepts a matrix with a 7-element row containing the
- successful balls for each week and as many rows as analysed. It creates a new 2
- by 49 matrix in which the total success to date for each ball is computed. This
- matrix is then sorted in descending order of ball success.
-
- A further 3 by 21 matrix is then created showing for each total number of ball
- success categories from 0 to 20, the observed and Poisson expected number of
- balls with this degree of success. Over the first 28 weeks no ball has
- succeeded less than 1 time or more than 8 times. The first and second columns
- are set in the custom function and the third column is set in the calling
- spreadsheet.
-
- Chart28 shows the observed and Poisson distribution of ball success.
-
- %H1%Appendix
-
- The file LotNos32 contains a further four weeks of lottery numbers making a
- total of 32 weeks results; and Chart32 shows the latest cumulative distribution
- of ball success. The leftward skew of the observed distribution of ball
- success, is beginning to look interesting, but it will probably disappear over
- the next 32 weeks.
-
- Our lives start with the lottery of the genes we inherit and the family
- circumstances of our birth: health and wealth and happiness are not ours to
- choose. Compared to this great "lottery of life", winning a fortune through
- Camelot is a small thing, so I do not see no trying to win as a moral failing.
- But I agree with Gerald that some may be tempted to gamble more than they can
- afford. Perhaps some may gamble money they should spend on food and clothes for
- their children, and this is a very bad thing. But I read that Britain was the
- last country in Europe to institute a state lottery, a few months behind
- Albania. Britain had a State lottery once before. In the early 19th century
- prayers were offered in churches for the success of parishioners in the British
- State lottery! This earlier lottery was stopped after allegations of cheating.
- I wonder how long the present one will last?
-