Why....

[nightprowler10]

Zac, do you go through the Battle of the Australians thread to pick out avatars? [/attempt to take thread further off-topic]

 

[Perm]

Zac, do you go through the Battle of the Australians thread to pick out avatars? [/attempt to take thread further off-topic]

No, can't say I do.

 

[Pratters]

Why do stats against minnows get removed to prove a point, like those runs shouldn't count because the oppo is poor but when a player fails against said poor opposition little is made of it.

I've seen this question raised before and it's a valid question. Fine if you want to say a ton against Bangladesh or Holland shouldn't count that much because of their quality, but then the players that don't perform shouldn't it be counted as double failure? Seems like we only see one side of the story here.

Thoughts?

This is a very interesting topic Zac. If we look at the Biased Sample theory of stats, it says the following:

A biased sample is a statistical sample of a population where some members of the population are less likely to be included than others. An extreme form of biased sampling occurs when certain members of the population are totally excluded from the sample (that is, they have zero probability of being selected).

By that logic, one would have to include all teams (equally no less). However, as cricket fans, we know that this isn't right as we are concerning ourselves with finding a qualatative ascertainment of runs and wickets. If we give equal importance to matches versus Bangladesh and Australia, an Irfan Pathan could be considered a far better bowler at a stage of his career when he clearly wasn't comparative to another player.

The solution where I am concerned is giving weights to teams according to their difficulty level. So if it is felt that Zimbabwe and Bangladesh are very easy oponents, give them very low weights. Give a team like Pakistan and New Zealand middle level weights and say an Australia lot of weights. A West Indies would command a similarly high weightage points in the 70s and 80s.

How to adjust these weights? Maybe over a 0-3 (4 maximum) year period.

So weights right now could look as follows:

Australia .8
South African .58
India .55
....
Bangladesh .01
Zimbabwe .005

(This is just an example)

How to determine weights? ICC ratings or win-loss-draw percentage extractions (giving due importance to home and away) maybe.

Excluding Zimbabwe or Bangladesh completely isn't right because they are test runs after all.

There are other stuff some one can look at - giving weights according to adjusted strengths of the bowling and batting line ups apart from the teams a player is facing. Would not have a Tendulkar or Dravid tonne in Australia in 2003-04 been all the more difficult if Warne and/or McGrath been playing in the series? That would provide a clearer picture (if one can call it that because stats never really tell the truest picture or even necessarily a true picture) but would make the process all the more complicated. We haven't yet calculated the weight adjustments to be done because of wickets or conditions or so millions of other things cricket comes up with.

So how much would Hayden's average go down if we give the 380 lower weight and how much would a 100 versus the West Indies in Barbados in the 80's be worth?

 

[nightprowler10]

No, can't say I do.

Was talking to the Jew.

 

[pasag]

Haha, was confused at first then went and looked it up. Gun thread, should do it again (mainly because I don't think I was around CC much at the time ). Anyways, nah, was looking for Miller in a St Kilda gursey (he played 50 matches of Aussie Rules, WAG), couldn't find one and settled with this found on google.

Nice post Pratyush, btw.

 

[silentstriker]

This is a very interesting topic Zac. If we look at the Biased Sample theory of stats, it says the following:

A biased sample is a statistical sample of a population where some members of the population are less likely to be included than others. An extreme form of biased sampling occurs when certain members of the population are totally excluded from the sample (that is, they have zero probability of being selected).

By that logic, one would have to include all teams (equally no less). However, as cricket fans, we know that this isn't right as we are concerning ourselves with finding a qualatative ascertainment of runs and wickets. If we give equal importance to matches versus Bangladesh and Australia, an Irfan Pathan could be considered a far better bowler at a stage of his career when he clearly wasn't comparative to another player.

The solution where I am concerned is giving weights to teams according to their difficulty level. So if it is felt that Zimbabwe and Bangladesh are very easy oponents, give them very low weights. Give a team like Pakistan and New Zealand middle level weights and say an Australia lot of weights. A West Indies would command a similarly high weightage points in the 70s and 80s.

How to adjust these weights? Maybe over a 0-3 (4 maximum) year period.

So weights right now could look as follows:

Australia .8
South African .58
India .55
....
Bangladesh .01
Zimbabwe .005

(This is just an example)

How to determine weights? ICC ratings or win-loss-draw percentage extractions (giving due importance to home and away) maybe.

Excluding Zimbabwe or Bangladesh completely isn't right because they are test runs after all.

There are other stuff some one can look at - giving weights according to adjusted strengths of the bowling and batting line ups apart from the teams a player is facing. Would not have a Tendulkar or Dravid tonne in Australia in 2003-04 been all the more difficult if Warne and/or McGrath been playing in the series? That would provide a clearer picture (if one can call it that because stats never really tell the truest picture or even necessarily a true picture) but would make the process all the more complicated. We haven't yet calculated the weight adjustments to be done because of wickets or conditions or so millions of other things cricket comes up with.

So how much would Hayden's average go down if we give the 380 lower weight and how much would a 100 versus the West Indies in Barbados in the 80's be worth?

We already have CW rankings that are being updated that you could use for this.

 

[Anil]

Has anyone ever thought that perhaps Bangladesh and Zimbabwe are both good sides but most people just happen to play good against them making their averages appear as though they aren't good or test standard..

I don't expect the answer to be yes completely, but with cricket being such a confidence sort of game, if you feel you've got a good opportunity to cash in against Bangas and Zims before a ball is bowled - then your in a positive frame of mind and more likely to perform better.

nice try, but no way are either of them test standard...

 

[nightprowler10]

Haha, was confused at first then went and looked it up. Gun thread, should do it again (mainly because I don't think I was around CC much at the time ). Anyways, nah, was looking for Miller in a St Kilda gursey (he played 50 matches of Aussie Rules, WAG), couldn't find one and settled with this found on google.

Wouldn't be a bad idea doing another battle of the Aussies tbh. With Warne and McGrath now retired on a high and Hayden hitting the form of his life, Ponting improving his form, the results could well come out completely different. Besides, its been more than a year. I ran it when I was a noob in posting terms and my randomizer caused Mister Wright quite a bit of frustration IIRC.

Dibs on running it again if we do it.

 

[Richard]

I have an idea on the next battle, TBH, that I need to keep under wraps for the time being but I'd ask that whenever this battle of the coms is finished (I presume that's soon if not already?), we lay off it for a short time.

It will be worth it.

 

[Son Of Coco]

Simple. People with what I might call an educated mind have realised that Bangladesh are not Test-class; that Bermuda are not ODI-class; etc. (there are quite a few teams who've never been ODI-class; there are a few teams who have at certain times been both Test- and\or ODI-class and not so at certain others).

With this in mind, anything - no picking and choosing, no removing runs but not removing innings; no removing innings above a certain score but leaving innings below a certain score - that happens in games such as these is not counted when we are assessing what we know is proper Test\ODI cricket.

Of course, someone should be disappointed not to score runs against Holland or Canada, but no more so than they should if they don't score them against Western Australia or Kent.

Usually (and not always correctly IMO) people are willing to completely write-off what happens at the level below Test\ODI if the player is a success at Tests\ODIs.

i.e...you

 

[Son Of Coco]

I'm doing it to prove a point about the value of statistics.

I refer to your post here: http://www.cricketweb.net/forum/showpost.php?p=1342476&postcount=29. You suggest that Paul Collingwood is indeed not as good as his current test average. In 50 years time, people might review Paul Collingwood's carrer and assume that he is in fact better than he was. This suggests that statistics do in fact mislead people, or "lie".

Statistics is the analysis of the numbers, not just the number itself. The danger lies in simply looking at a given player's average and drawing all your conclusions from that. Further analysis is required to come up with a true understanding of what they did or didn't achieve. This is why I have no problem with including all the innings a batsmen or bowler has played against all the Test nations at the time. The only time a player's stats become a problem is when you want an answer to how good/bad he was in 20 seconds.

 

[Son Of Coco]

This is a very interesting topic Zac. If we look at the Biased Sample theory of stats, it says the following:

A biased sample is a statistical sample of a population where some members of the population are less likely to be included than others. An extreme form of biased sampling occurs when certain members of the population are totally excluded from the sample (that is, they have zero probability of being selected).

By that logic, one would have to include all teams (equally no less). However, as cricket fans, we know that this isn't right as we are concerning ourselves with finding a qualatative ascertainment of runs and wickets. If we give equal importance to matches versus Bangladesh and Australia, an Irfan Pathan could be considered a far better bowler at a stage of his career when he clearly wasn't comparative to another player.

The solution where I am concerned is giving weights to teams according to their difficulty level. So if it is felt that Zimbabwe and Bangladesh are very easy oponents, give them very low weights. Give a team like Pakistan and New Zealand middle level weights and say an Australia lot of weights. A West Indies would command a similarly high weightage points in the 70s and 80s.

How to adjust these weights? Maybe over a 0-3 (4 maximum) year period.

So weights right now could look as follows:

Australia .8
South African .58
India .55
....
Bangladesh .01
Zimbabwe .005

(This is just an example)

How to determine weights? ICC ratings or win-loss-draw percentage extractions (giving due importance to home and away) maybe.

Excluding Zimbabwe or Bangladesh completely isn't right because they are test runs after all.

There are other stuff some one can look at - giving weights according to adjusted strengths of the bowling and batting line ups apart from the teams a player is facing. Would not have a Tendulkar or Dravid tonne in Australia in 2003-04 been all the more difficult if Warne and/or McGrath been playing in the series? That would provide a clearer picture (if one can call it that because stats never really tell the truest picture or even necessarily a true picture) but would make the process all the more complicated. We haven't yet calculated the weight adjustments to be done because of wickets or conditions or so millions of other things cricket comes up with.

So how much would Hayden's average go down if we give the 380 lower weight and how much would a 100 versus the West Indies in Barbados in the 80's be worth?

Yeah, a weighted average would be a solution. As you say though, it would require the weightings to be spread across all bowling attacks of all times to be fairly accurate. To simply give the West Indies a rating for their attack now would render the average innaccurate for the period they still had Curtly Ambrose and Courtney Walsh playing. If we were to go back to the 90's obviously it would have to be adjusted again. And then in the 80's they had the strongest attack going. Not all teams have an equally strong batting and bowling attack, so that would have to be factored in. Then subcontinental teams playing on their own tracks, and the Windies/Aus/Eng etc playing on their own wickets would be rated higher than if they played in the subcontinent, with the exception of when Warney was playing perhaps for Australia.

Any volunteers?

 

[Richard]

Statistics is the analysis of the numbers, not just the number itself. The danger lies in simply looking at a given player's average and drawing all your conclusions from that. Further analysis is required to come up with a true understanding of what they did or didn't achieve. This is why I have no problem with including all the innings a batsmen or bowler has played against all the Test nations at the time. The only time a player's stats become a problem is when you want an answer to how good/bad he was in 20 seconds.

Why not just go further and include every game of cricket he's ever played?

 

[Son Of Coco]

Why not just go further and include every game of cricket he's ever played?

Because it depends on what we're looking at Richard. If you want a batsman's test average you look at tests. If you want his first class average you look at tests and state/county games. If you want his lifetime average in all forms of cricket from U/12s onwards then you look at every game he's ever played from that point onwards.

I find it hard to believe even you can't tell the difference between including games played against all Test nations at the time, and every game a person's played. As I said, usually the problem with stats is what people try to do with them rather than the stat itself. We kind of proved the point in 'Hear, Hear,,,'.

 

[Perm]

Was talking to the Jew.

I know.

 

[Richard]

Because it depends on what we're looking at Richard. If you want a batsman's test average you look at tests. If you want his first class average you look at tests and state/county games. If you want his lifetime average in all forms of cricket from U/12s onwards then you look at every game he's ever played from that point onwards.

The point is, games are classed as something they shouldn't be. Are you seriously telling me that Bangladesh are a Test-class team?

If you want a batsman's Test or ODI average you look at games that deserve to be called Tests or ODIs, not those that some imbecile at I$C$C has called a Test\ODI because they don't have a clue how to promote the best interests of cricket.

If you want good value cricket statistics, you have to classify games properly. There are many cases of this not being done - "remove team X" is merely a way to rectify this.

 

[Son Of Coco]

The point is, games are classed as something they shouldn't be. Are you seriously telling me that Bangladesh are a Test-class team?

If you want a batsman's Test or ODI average you look at games that deserve to be called Tests or ODIs, not those that some imbecile at I$C$C has called a Test\ODI because they don't have a clue how to promote the best interests of cricket.

If you want good value cricket statistics, you have to classify games properly. There are many cases of this not being done - "remove team X" is merely a way to rectify this.

No, you do this because you can't be arsed spending the time to do it properly. And what you end up with is bad stats.

 

[Richard]

ODI stats including ODI-standard sides only >>>>>>> ODI stats including all teams I$C$C consider worthy of ODI-status.

If one is awful and the other is merely bad, I couldn't care less.

 

[Son Of Coco]

ODI stats including ODI-standard sides only >>>>>>> ODI stats including all teams I$C$C consider worthy of ODI-status.

If one is awful and the other is merely bad, I couldn't care less.

Ah, but one is awful in your opinion, and one is great in your opinion.

You come up with a good idea but implement it poorly. It's ok, not a lot of people are going to be bothered to tweak the stats and still have them remain significant unless they're getting paid for it.

 

[Ikki]

It all depends how you use them TBH. But I find people regularly take the stats out when comparing 2 players and seeing who is better. For example, one person makes the assessment that because Batsman A's average of 56 is higher than Batsman B's average of 51, he s better. It is very apt then to see when Batsman A has played 15 tests against these minnows and Batsman B has played only 1. Then you end up with averages like Batsman A: 49; Batsman B 50 - which gives a clearer picture IMO.