How good was that batsman?

[thongale]

There was plenty of arguments over Bradman vs SRT. Bradman fans argued that Bradman's average was out of ordinary stuff. SRT fans argue that SRTs weight of runs is unmatchable. There was some discussion on quality of cricket they played, but measuring the "quality" is just guess work, so I am just dropping that. BradmanÃÔ average of 99.94 was regarded as an extreme outlier in sport, which comes to about 4 standard deviations (SD). But when calculated properly this looks like to be around 6 SDs. But there are some limitations in describing him using SDs Firstly, to use SDs, the distribution has to be Gaussian. And here is the distribution of batting averages of all batsmen. Since some batsmen have remained not out in few innings they played, they will not have an average. Hence I calculated runs-per-innings (RPI) to include them as well. The following histograms show the distribution of Averages and RPIs. The mean batting average is 20.27 with a SD of 13.94. This makes Bradman 5.72 SDs away from mean, and his next competitor Tendulkar 2.52 SDs away from the mean, a massive difference. When RPIs taken in to account, mean RPI is 17.52 with SD of 12.90. Once again Bradman with RPI of 87.45, is 5.42 SDs away while Tendulkar (RPI = 49.74) is 2.5 SDs away. Once you see the distributions it is evident that the distributions are heavily negatively skewed. SD and mean of do not describe such a distribution well. Instead, in such a distribution, the better measure would be percentiles. (See attachments)

Percentile	Average	RPI
50		17.71		14.87
90		40.07		35.86
95		46.36		42.00
99		58.45		51.00

When percentiles are calculated the differences become much lesser than earlier. When Bradman is considered, he is in 99.96th and 99.92th percentiles. Tendulkar is in 98.5th and 98.7th centiles. This correlates to 3.35 and 3.16 SDs away from mean if the above negatively skewed distribution is fitted in to a normal distribution. For Tendulkar, its 2.17 and 2.23 SDs away from the mean respectively. This shows that the differences of averages are not as massive as it is shown to be with absolute numbers. Batting average is only one of the statistics we can use in judging a player. Number of runs scored, centuries scored, centuries per innings are the other criteria which will be helpful making a batting average more meaningful. So I've looked at total number of runs scored. (attachment 3) and calculated percentiles (as it is also a skewed distribution). Here no need to say that Tendulkar is in the 100th percentile while Bradman is in 97.6th percentile. (Due to a count error SRT's P is shown as 99.25, and the same proportion applied to other batsmen as well. And it will not make any difference in final out come) Now we have a measure to compare difficulty of maintaining an average vs difficulty in scoring a lot of runs. I just multiplied them and ranked with Average and RPIs

[thongale]

OK According to the Aggregate vs Batting Average here are the top 20. Here is Aggregate vs Runs per Innings I think you can improve by adding other parameters as well after seeing their frequency distributions. For the moment, I am too lazy

[Crookbond]

:hatsoff: Your first post, especially, is real insightful. I liked the way you used percentiles. It really makes sense.

[Sidhoni]

[Guest BossBhai]

-- Removed on request of the user --

[JaFanatic]

both are gr8 batsmen but from different eras so no comparison

[velu]

This is some new kind of analysis

[Crookbond]

No Arthur Mold and no videos yet! :omg:

[velu]

Data taken by OP is not proper or it includes some unwanted data. Bradman's RPI is 87.75 which should have the percentile score of 100 or atleast close to 100 ( iirc only one batsmen has a average of over 100 but he played only 1 match ). Also Bradmanns percentile average should be close to 100 and not 96.9

[velu]

This analysis is completely flawed. It doesnt give any weightage to exceptional stats. If a batsmen averages 500 Runs per innings , he will get the percentile score 100 , but the next guy who is averaging just 100 will get a percentile score of 99.99 or close to it. If we multiply any parameters with this percentile values , the batsmen who averaged 500 gets unfair treatment. :(( There is no way kallis or sanga can be better than sachin , and nor does anyone better than Bradman.

[panther]

inzi and anwar best batsman to come out of subcontinent.

[velu]

inzi and anwar best batsman to come out of subcontinent.

no.. it has to be afridi :finger:

[Crookbond]

inzi and anwar best batsman to come out of subcontinent.

:thefinger: Appam Chu****

[panther]

Thongle i see you have put in a lot of work but i disagree with moyo and miandad being in the top 20. both of these were great batsman but also ftb's.

[thongale]

So we simply assume that a Bill Bowes is equal in skills to Allan Donald ( because their Avgs are almost same ) ' date=' Ernie Toshack > Waz for same reasons . Makes perfect cricketing sense. [/quote']Nah, you don't get the point. It's easy to bowling average 15 off 5 matches compared to 50. To have an idea we have to use both. And you may have seen that some big names such as Sutcliffe (60.73), B.Richards (73), G. Pollock (62) are out of the list. Once again two way. If you tend to do well in one country, then you'll be benefited (ex. Shewag vs Pakistan or Sri Lanka). But if you happened to suck against them the impact will be massive. (Ex. Cullinan vs Australia)

[thongale]

Thongle i see you have put in a lot of work but i disagree with moyo and miandad being in the top 20. both of these were great batsman but also ftb's.

Probably yes. There are MANY MORE parameters to be added. I am just lazy to do that. You are welcomed to improve on them, I can even post the Excel and SPSS woksheets to start with.

[thongale]

Data taken by OP is not proper or it includes some unwanted data. Bradman's RPI is 87.75 which should have the percentile score of 100 or atleast close to 100 ( iirc only one batsmen has a average of over 100 but he played only 1 match ). Also Bradmanns percentile average should be close to 100 and not 96.9

Bradman's RPI is actually second in the list while average wise he is third. And Bradman's RPI percentile is 99.2. The problem is you cannot set filters in such analysis. Then the level of filter will become an issue.

[thongale]

This analysis is completely flawed. It doesnt give any weightage to exceptional stats. If a batsmen averages 500 Runs per innings , he will get the percentile score 100 , but the next guy who is averaging just 100 will get a percentile score of 99.99 or close to it. If we multiply any parameters with this percentile values , the batsmen who averaged 500 gets unfair treatment. :(( There is no way kallis or sanga can be better than sachin , and nor does anyone better than Bradman.

You have a point. But the raw stats and standard deviation just ignores the distribution. What we have to know is that average of 90 is special and so is average of 0.25. I admit this is not the perfect system, but only and effort to look at things bit differently. We know Tendulkar is better than Kallis or Sanga due to other reasons. But statistics wise, they are ahead.

[velu]

You have a point. But the raw stats and standard deviation just ignores the distribution. What we have to know is that average of 90 is special and so is average of 0.25. I admit this is not the perfect system, but only and effort to look at things bit differently. We know Tendulkar is better than Kallis or Sanga due to other reasons. But statistics wise, they are ahead.

IMO your analysis is really good in analyzing players with similar stats .. but i agree that setting filters will be a very big issue when u try to optimize it further anyway u have put lots of efforts .. may be ur jobless :two_thumbs_up:

[thongale]

The above took me only about one and half hours. And if you are interested, I am a Consultant Physician, so I have quite a good job too.