• Welcome to the Cricket Web forums, one of the biggest forums in the world dedicated to cricket.

    You are currently viewing our boards as a guest which gives you limited access to view most discussions and access our other features. By joining our free community you will have access to post topics, respond to polls, upload content and access many other special features. Registration is fast, simple and absolutely free so please, join the Cricket Web community today!

    If you have any problems with the registration process or your account login, please contact us.

New feature: Highest Impact - the all-time most impactful Test cricketers

viriya

International Captain
I was checking your article on how you assigned ISPs (Series Points ? A New Way of Ranking Test Players | Cricket Web), and was wondering how you valued batsmen and bowlers using the same measure. For example:

Say the total series points is 500. The batting average during that time period was 30 and the wickets per match for the period was 3/match. Bradman made 500 runs in the series while Grimmett took 15 wickets. Assuming those are the only two players, how will the series points be assigned between the two?

Maybe I missed some detail elsewhere - apologies if so.
 

MartinB

School Boy/Girl Cricketer
Another Bowling impact analysis is available on this article on Lillee

* THE COMPLETE FAST BOWLER | 100 HIDDEN CRICKET FACTS

I do not know how good there Impact analysis is. They seem to rate

  • Significant performances that impact match/series results
  • Top seven wickets per test
  • Tail end wickets (less important than top 7)
  • Economy
  • pressure building
  • failure rate
  • partnership breaking
 

watson

Banned
Based on 'ALL-TIME BEST IMPACT TEST PLAYERS – WEIGHTED AVERAGE, ADJUSTED', and picking the lower ranked Kumar Sangakkar because the team should have a keeper, we get;

01. Jacques Kallis
02. Sachin Tendulkar
03. Don Bradman
04. Brian Lara
05. Garry Sobers
06. Ian Botham
07. Kumar Sangakkara
08. Richard Hadlee
09. Shaun Pollock
10. Shane Warne
11. Sydney Barnes


A bit surprised that none of the West Indian quicks, Imran, or Lillee are in the top 11.

Also, there seems to be relatively few specialist openers featured in the 16 lists. Jack Hobbs, Bobby Simpson, and Virender Sehwag seem to have the highest impacts of those listed.
 
Last edited:

chasingthedon

International Regular
I was checking your article on how you assigned ISPs (Series Points ? A New Way of Ranking Test Players | Cricket Web), and was wondering how you valued batsmen and bowlers using the same measure. For example:

Say the total series points is 500. The batting average during that time period was 30 and the wickets per match for the period was 3/match. Bradman made 500 runs in the series while Grimmett took 15 wickets. Assuming those are the only two players, how will the series points be assigned between the two?

Maybe I missed some detail elsewhere - apologies if so.
First it's allocated between batting and bowling/fielding at the team level depending on series performance and era, then shared between fielders and bowlers accordingly.
 

chasingthedon

International Regular
Based on 'ALL-TIME BEST IMPACT TEST PLAYERS – WEIGHTED AVERAGE, ADJUSTED', and picking the lower ranked Kumar Sangakkar because the team should have a keeper, we get;

01. Jacques Kallis
02. Sachin Tendulkar
03. Don Bradman
04. Brian Lara
05. Garry Sobers
06. Ian Botham
07. Kumar Sangakkara
08. Richard Hadlee
09. Shaun Pollock
10. Shane Warne
11. Sydney Barnes


A bit surprised that none of the West Indian quicks, Imran, or Lillee are in the top 11.

Also, there seems to be relatively few specialist openers featured in the 16 lists. Jack Hobbs, Bobby Simpson, and Virender Sehwag seem to have the highest impacts of those listed.
As I mentioned, it depends on how you summarize the various data. Imran and Lillee are both mentioned at some point, as are Walsh, Marshall and Croft. As regards not being in the top 11, I wouldn't necessarily expect that - at least, not to be able to pick a full team from the top 11, if that's what you mean.

As regards openers, that could be an artifact of the system - it's more difficult to impact the win probability at the start of the match, though not impossible; it's more likely in the team's second innings, but if the opener does nothing else (i.e. fielding or bowling) he won't rate highly. Hobbs probably the exception that proves the rule, while Goddard and Simpson were all-rounders who opened.

Thanks for taking the time to read and comment.
 

smash84

The Tiger King
not a bad list of 20 in the end, certainly can be argued for and against but a great effort overall
 

watson

Banned
As I mentioned, it depends on how you summarize the various data. Imran and Lillee are both mentioned at some point, as are Walsh, Marshall and Croft. As regards not being in the top 11, I wouldn't necessarily expect that - at least, not to be able to pick a full team from the top 11, if that's what you mean.

As regards openers, that could be an artifact of the system - it's more difficult to impact the win probability at the start of the match, though not impossible; it's more likely in the team's second innings, but if the opener does nothing else (i.e. fielding or bowling) he won't rate highly. Hobbs probably the exception that proves the rule, while Goddard and Simpson were all-rounders who opened.

Thanks for taking the time to read and comment.
I guess I was surprised at Shaun Pollock's high status more than anything else. Sure he can bat a bit, but so could Imran and Kapil Dev. Initially I thought that their results might be affected by poor team performance, especially with Kapil as India won little during the 80s. But that doesn't seem to be the case,

Therefore, the actual outcome of the match is not taken into account in the assessment of the player’s performance.
So it seems that as far as Shaun Pollock is concerned it's a case of "cometh the hour, cometh the man" as they say. Or as you neatly put it;

For example, when Ian Botham had Jeff Thomson caught by Geoff Miller to win the Boxing DayTest of 1982 against Australia in Melbourne by just three runs, it shows on the scorecard as simply one of his two wickets in the innings. A return of 2/80 at first glance wouldn’t seem to represent a great performance, but In the context of the match that wicket had enormous significance as regards England’s chances of winning before and after the dismissal. Similarly Doug Ring – his 32* turned a likely loss into an unexpected win when Australia beat West Indies in 1951-52, again at the MCG around the New Year – those 32 precious runs were worth more than many a century in terms of their impact on Australia’s chances of winning.

Given Shaun Pollock's more obvious status in ODI cricket it would be interesting to crunch the numbers and see whether he had more impact than assumed No.1s like Garner, McGrath, Richards and Tendulkar. Or even a comparable player like Lance Klusener.
 

viriya

International Captain
First it's allocated between batting and bowling/fielding at the team level depending on series performance and era, then shared between fielders and bowlers accordingly.
But how exactly do you fairly allocate team level batting and bowling performance using Series Points without an assumption on how much a X runs is worth in terms of wickets?
 

chasingthedon

International Regular
I guess I was surprised at Shaun Pollock's high status more than anything else. Sure he can bat a bit, but so could Imran and Kapil Dev. Initially I thought that their results might be affected by poor team performance, especially with Kapil as India won little during the 80s. But that doesn't seem to be the case,
It shouldn't be, as the impact on win probability is taken at the time, rather than being based on whether they actually went on to win or not. There are a lot of events though, so I haven't been able to check that's always the case.


So it seems that as far as Shaun Pollock is concerned it's a case of "cometh the hour, cometh the man" as they say. Or as you neatly put it;




Given Shaun Pollock's more obvious status in ODI cricket it would be interesting to crunch the numbers and see whether he had more impact than assumed No.1s like Garner, McGrath, Richards and Tendulkar. Or even a comparable player like Lance Klusener.
It would be interesting, but I'd first have to develop a database of match status to generate the win probabilities, which takes a long time. I'll probably take some time off :)
 
Last edited:

chasingthedon

International Regular
But how exactly do you fairly allocate team level batting and bowling performance using Series Points without an assumption on how much a X runs is worth in terms of wickets?
I can go back and check the method, as it was a long time ago and I can't recall all the details.

How come you're so interested in that method? I did propose that the impact method is an improvement :)
 

chasingthedon

International Regular
As I mentioned, it depends on how you summarize the various data. Imran and Lillee are both mentioned at some point, as are Walsh, Marshall and Croft. As regards not being in the top 11, I wouldn't necessarily expect that - at least, not to be able to pick a full team from the top 11, if that's what you mean.

As regards openers, that could be an artifact of the system - it's more difficult to impact the win probability at the start of the match, though not impossible; it's more likely in the team's second innings, but if the opener does nothing else (i.e. fielding or bowling) he won't rate highly. Hobbs probably the exception that proves the rule, while Goddard and Simpson were all-rounders who opened.

Thanks for taking the time to read and comment.
Also what I haven't done is post the relative levels of batting, fielding and bowling for each player, partly because it goes against what I was trying to do. It might be interesting though, particularly with all-rounders such as Shaun Pollock, to see what the proportions were. Or for a WK like Knott versus Jack Blackham and Adam Gilchrist.
 

viriya

International Captain
I can go back and check the method, as it was a long time ago and I can't recall all the details.

How come you're so interested in that method? I did propose that the impact method is an improvement :)
Never mind - I thought of starting from the first article, didn't realize you switched methods midway.

Interesting idea to use FOW to calculate win probabilities. I've done a some what similar project with win shares where I do over by over calculations but obviously this is not possible for matches pre-2001 or so. Your method seems like a good way to get are nd that but I've found a few issues with doing win shares for tests that maybe you have some insight on:

- tests are notoriously hard to predict. Even if you look for "similar" match situations based on the score and runs to get, you will find that past matches are a weak predictor for the future.

- based on what I read on how you calculate win probabilities, it seems like you don't factor in the "time left" in the match. For example two 4th innings chases of 250 where the score 150/5 where one of them is at the 4th morning and the other is on the 5th day with 10 overs to go should not have the same win provability. This problem of figuring out the "ovets left" in a match makes win provability a difficult measure to get right.

- since you are using FOW, I would make an educated guess that for a lot of the FOW situations you don't find that many similar matches. If you only have 2-3 similar matches to use, the probability value you get is not going to be that accurate because of sample size issues.

Interested to know what you think about the above issues.
 

chasingthedon

International Regular
Never mind - I thought of starting from the first article, didn't realize you switched methods midway.

Interesting idea to use FOW to calculate win probabilities. I've done a some what similar project with win shares where I do over by over calculations but obviously this is not possible for matches pre-2001 or so. Your method seems like a good way to get are nd that but I've found a few issues with doing win shares for tests that maybe you have some insight on:

- tests are notoriously hard to predict. Even if you look for "similar" match situations based on the score and runs to get, you will find that past matches are a weak predictor for the future.

- based on what I read on how you calculate win probabilities, it seems like you don't factor in the "time left" in the match. For example two 4th innings chases of 250 where the score 150/5 where one of them is at the 4th morning and the other is on the 5th day with 10 overs to go should not have the same win provability. This problem of figuring out the "ovets left" in a match makes win provability a difficult measure to get right.

- since you are using FOW, I would make an educated guess that for a lot of the FOW situations you don't find that many similar matches. If you only have 2-3 similar matches to use, the probability value you get is not going to be that accurate because of sample size issues.

Interested to know what you think about the above issues.
I didn't only use the FOW point - say the wicket fell at 214/7, I would use all matches where there was a 214/7 status, i.e. the 7th wicket at or before 214 and the 8th wicket fell at or after 214, so at some point the score passed through 214/7 (e,g, from 213/7 to 215/7). That gives us a much bigger sample size and accordingly more confidence in the win probability.

EDIT: also I don't use the batting score or wickets down per se, rather how many runs behind and how many wickets to give.
 

viriya

International Captain
I didn't only use the FOW point - say the wicket fell at 214/7, I would use all matches where there was a 214/7 status, i.e. the 7th wicket at or before 214 and the 8th wicket fell at or after 214, so at some point the score passed through 214/7 (e,g, from 213/7 to 215/7). That gives us a much bigger sample size and accordingly more confidence in the win probability.

EDIT: also I don't use the batting score or wickets down per se, rather how many runs behind and how many wickets to give.
Yes, expanding the window that way makes sense.

This doesn't resolve the #2 and #1 issues however. How do you handle "time/overs left"? If you don't, your probabilities won't make sense for especially the 3rd/4th innings where the win probability depends a lot on time left.
 

viriya

International Captain
Another thing I noticed is that you were giving 50% of the dismissal credit to the fielder. I think that's too much - going by the average drop rate % which is around 15%, the fielder should get around that much for a regulation catch with 85% going to the bowler.

I think that would be a little more objective than just giving a flat 50%.

Great job btw - I'm trying to figure out what you've done to get a better underatanding of whether it is actually a viable method.
 

Top