Contribution Analysis Done!

[Nirvanam]

People, The project is finally done! You can view the article on my blog. I have also provided links to a pdf version and the data file in the blog. Needless to say, would love to know your thoughts about the project, the analysis, were the results as expected? surprises? etc? http://perceptz.blogspot.com/2010/11/contribution-analysis-of-atg-test_25.html To sum up... The hypothesis that, “The overall contribution, in terms of runs scored, of the 16 ATG Test batsmen to their teams’ causes, are similar in nature across their careers” has been objectively disproved. The project also establishes the following distinction between the 16 ATG players as far as contribution, in terms of runs scored, to their teams’ cause is concerned. Donald Bradman, Brian Lara, and Sachin Tendulkar tower over the ATG batsmen. Barrington, Dravid, Gavaskar, Hobbs, Hutton, Kallis, Ponting, Richards, and Sobers epitomize ATG-ness. Chappell, Inzamam, Hammond, and Miandad are among the lesser immortals of this elite ATG gang. The Final Graph ps: How do I get the graph inserted into the post itself?

[Nova]

My thought on this is that this is a waste of time (with due respect). May be the likes of Bossbhai will get aroused looking at this but I find such analysis boring. :yawn2:

[akshayxyz]

Imagine each batsman can be defined accurately with just three attributes. X,Y,Z. for the sake of simplicity, each attribute is rated on a scale of 1-100 ( or say 1000). Then all of the batsmen can be plotted as points in this three dimensional space. For retired player - the coordinates will be fixed - for current players, some attributes may be modified over the years. Now there comes a critic/statistician - Who is fine with relative scaling of the coordinate system. He wants to find the best batsman. So it comes down to searching for the "best" point in this 3-d space. So which is best - the (1000,0,0) Or (0,1000,0), (0,0,1000), (700,500,800), (800,500,700) or (700,800,500)....- it is highly unlikely to have a point ( 1000,1000,1000) - that should be clear winner in that case. Now imagine another critic comes and he questions the relative weights of the axes. He will shrink certain axes, and expand certain other. All the points will have to be transposed to this new coordinate system. I have not yet talked about other bunch of statisticians - who does not even agree with relative grading of batsmen on a given axis! Now come to real world - there are not 3 dimensions, but 1000s of them that define a batsman and his performance. And there are 100s of batsman that you need to compare with each other. Statisticians, at their convenience and/or to prove some perception of theirs, drop certain dimensions, change weight-age of certain dimensions, change gradings of certain batsman on certain attributes. Then come up with xyz>pqr!!! My point - well just to establish the complexity of the problem - lot of people claim to solve every other day with different results of course. Enjoy what you are enjoying doing.

[Nirvanam]

Imagine each batsman can be defined accurately with just three attributes. X,Y,Z. for the sake of simplicity, each attribute is rated on a scale of 1-100 ( or say 1000). Then all of the batsmen can be plotted as points in this three dimensional space. For retired player - the coordinates will be fixed - for current players, some attributes may be modified over the years. Now there comes a critic/statistician - Who is fine with relative scaling of the coordinate system. He wants to find the best batsman. So it comes down to searching for the "best" point in this 3-d space. So which is best - the (1000,0,0) Or (0,1000,0), (0,0,1000), (700,500,800), (800,500,700) or (700,800,500)....- it is highly unlikely to have a point ( 1000,1000,1000) - that should be clear winner in that case. Now imagine another critic comes and he questions the relative weights of the axes. He will shrink certain axes, and expand certain other. All the points will have to be transposed to this new coordinate system. I have not yet talked about other bunch of statisticians - who does not even agree with relative grading of batsmen on a given axis! Now come to real world - there are not 3 dimensions, but 1000s of them that define a batsman and his performance. And there are 100s of batsman that you need to compare with each other. Statisticians, at their convenience and/or to prove some perception of theirs, drop certain dimensions, change weight-age of certain dimensions, change gradings of certain batsman on certain attributes. Then come up with xyz>pqr!!! My point - well just to establish the complexity of the problem - lot of people claim to solve every other day with different results of course. Enjoy what you are enjoying doing.

I htink you misunderstood the analysis. It is not talking about the best batsman. It is trying to say how often a particular batsman (from the 16 ATGs) contributed at least 50 runs in a test match or top scored. Nothing more nothing else

[thongale]

Good work. But like to see what Andy Flower, the Good Doctor and Allan Border has done. If Inzi is regarded as ATG, Flower and AB can be regarded as ATG too.

[Nirvanam]

Good work. But like to see what Andy Flower' date=' the Good Doctor and Allan Border has done. If Inzi is regarded as ATG, Flower and AB can be regarded as ATG too.[/quote']Thongale, Flower's number comes up with a rough calculation to around 215. This is a very rough calculation...haven't looked at each of his scorecards. So take it with a +- 50 points range....lesser than the worst of the ATGs who is 303.

[Nirvanam]

Another rough calculation for Flower. First, facts - he played 63 Tests, 12 cents and 27 fifties. So sup.per is minimum 27...i.e. 27/63 = 0.43 sig.per is minimum 12...i.e. 12.63 = 0.19 In the previous post I considered all his 12 cents as top score for team, as well...hence his top-per = 0.19. But now let us consider that he top scored in say 50% of his Tests. so his top.per = 0.5 Un-normalized OCI = 1.25*0.5 + 1*0.19 + 0.75*0.43 = 1.1375 Normalized (without RFF) = 1.1375 / 3 = 0.379. RFF of Bradman was 0.8...he played 52 Tests...so let's say Flower's RFF will be 0.85. So his final OCI will be, OCI = 0.85 * 0.379 * 1000 = 322 So his score will be 322 assuming he top scored for his team in 50% of his games.....now that is a huge assumption....so huge that even Bradman doesn't have such a number...Bradman's was 48%. So, I am guessing Flower's actual score will be between 250 and 300