Removing statistics against Bangladesh and Zimbawe

[Uppercut]

The thing is, they will start winning games at one point, just like they beat Australia in an ODI in 2005 and India in the bloody WC. At that point, you can't suddenly flip a switch and start counting performances against them.

Of course you can. You start counting performances from the date when they first win a game.

 

[vcs]

Of course you can. You start counting performances from the date when they first win a game.

Haha, that's about as worthless a way of doing it as excluding them altogether, or taking the stats at face value.

 

[Uppercut]

Why?

 

[vcs]

Totally arbitrary. They don't transform into a different team competence-wise, the day they win a game.

 

[Prince EWS]

The thing is, they will start winning games at one point, just like they beat Australia in an ODI in 2005 and India in the bloody WC. At that point, you can't suddenly flip a switch and start counting performances against them.

I discount performances against them in ODIs as well though, but I do it for a different reason to Uppercut. I think they're basically an outlier on the performance chart - the standard they play at is so far removed from the rest of cricket that performances against them are generally a terrible indicator of likely performances against every other team. It wouldn't be so much of a problem if some players didn't play a higher percentage of games against Bangladesh than others - but they do.

When I do things 'properly' I just standardise matches against them (and every team) by statistically recognising exactly what the difference between them and the rest of the world is and adjusting the numbers accordingly, but it's a very slow process. As Uppercut said, neither including them or removing them is perfect, but removing them gives a more realistic outlook so if you're just looking for a quick indication of someone's career record it's the best starting part. Irfan Pathan's Test bowling average of 32 or so is IMO wholly misleading when you consider that if you take out the four Tests he played against Bangladesh and Zimbabwe, he played 25 matches for an average of 45. For 86% of his career he played against genuine opponents and produced a rubbish average - that he dominated weak batting lineups for four matches means pretty much nothing to me but it has a massive effect on his average.

 

[vcs]

I discount performances against them in ODIs as well though, but I do it for a different reason to Uppercut. I think they're basically an outlier on the performance chart - the standard they play at is so far removed from the rest of cricket that performances against them are generally a terrible indicator of likely performances against every other team. It wouldn't be so much of a problem if some players didn't play a higher percentage of games against Bangladesh than others - but they do.

When I do things 'properly' I just standardise matches against them (and every team) by statistically recognising exactly what the difference between them and the rest of the world is and adjusting the numbers accordingly, but it's a very slow process. As Uppercut said, neither including them or removing them is perfect, but removing them gives a more realistic outlook so if you're just looking for a quick indication of someone's career record it's the best starting part. Irfan Pathan's Test bowling average of 32 or so is IMO wholly misleading when you consider that if you take out the four Tests he played against Bangladesh and Zimbabwe, he played 25 matches for an average of 45. For 86% of his career he played against genuine opponents and produced a rubbish average - that he dominated weak batting lineups for four matches means pretty much nothing to me but it has a massive effect on his average.

That's a pretty good explanation, I can definitely understand that.

Out of interest, what measure do you use to standardise the level of teams, statistically?

 

[Uppercut]

Totally arbitrary. They don't transform into a different team competence-wise, the day they win a game.

The whole point of doing it the day they win a game is that it's not arbitrary. Otherwise I'd need to pick a date out of thin air using my own ****ty judgment. In hindsight we might be able to agree on a better date, but that really isn't the point of my argument.

Anyway, there's another interesting angle. To turn the question around, why do you include performances against Bangladesh? If Richard were here he would ask why a match against Bangladesh becomes a test match when a higher-standard NSW/Victoria game is merely a first-class match. Do you just count them as tests because that's what the ICC arbitrarily claim they are?

 

[bagapath]

I discount performances against them in ODIs as well though, but I do it for a different reason to Uppercut. I think they're basically an outlier on the performance chart - the standard they play at is so far removed from the rest of cricket that performances against them are generally a terrible indicator of likely performances against every other team. It wouldn't be so much of a problem if some players didn't play a higher percentage of games against Bangladesh than others - but they do.

When I do things 'properly' I just standardise matches against them (and every team) by statistically recognising exactly what the difference between them and the rest of the world is and adjusting the numbers accordingly, but it's a very slow process. As Uppercut said, neither including them or removing them is perfect, but removing them gives a more realistic outlook so if you're just looking for a quick indication of someone's career record it's the best starting part. Irfan Pathan's Test bowling average of 32 or so is IMO wholly misleading when you consider that if you take out the four Tests he played against Bangladesh and Zimbabwe, he played 25 matches for an average of 45. For 86% of his career he played against genuine opponents and produced a rubbish average - that he dominated weak batting lineups for four matches means pretty much nothing to me but it has a massive effect on his average.

bang on prince ews... this is exactly my line of thinking as well.

 

[vcs]

The whole point of doing it the day they win a game is that it's not arbitrary. Otherwise I'd need to pick a date out of thin air using my own ****ty judgment. In hindsight we might be able to agree on a better date, but that really isn't the point of my argument.

Anyway, there's another interesting angle. To turn the question around, why do you include performances against Bangladesh? If Richard were here he would ask why a match against Bangladesh becomes a test match when a higher-standard NSW/Victoria game is merely a first-class match. Do you just count them as tests because that's what the ICC arbitrarily claim they are?

My point is using your own judgment to pick a date wouldn't necessarily yield a worse (or better) date than just picking the day they win a game. Teams don't perform, or improve linearly.

I'm not sure whether they should be included or excluded, but I find it pretty simplistic to claim that all performances against them mean squat in terms of influencing the result.

As for whether performances in FC matches should be considered, they often are (for players of previous eras). It's hard to compare players on FC stats though.. because they don't all play in one FC competition, that's why it's hard to read much into it beyond using it as a confirmation of a player's quality/longevity.

 

[Uppercut]

I'm not sure whether they should be included or excluded, but I find it pretty simplistic to claim that all performances against them mean squat in terms of influencing the result.

Just to be clear, that's not what I'm claiming. I'm merely suggesting that the performances are sufficiently unlikely to influence the result as to render stats minus Bangladesh a better measure than stats including Bangladesh.

 

[Prince EWS]

That's a pretty good explanation, I can definitely understand that.

Out of interest, what measure do you use to standardise the level of teams, statistically?

I generally use team home and away bowling/batting averages over a set period either side of the year of the match. Not perfect, but better than just leaving it IMO.

For example, if you batted against Team X in Team Xland in 1992, and that team averaged 40 with the ball at home between 1990 and 1994, I'd conclude that the team was operating at 1.33 times the global all-time average (30.02) at home, and divide your innings score by 1.33. A ton would be worth 75.

Basically it attempts to statistically hypothesise what you'd average if every team was the same. It's my no means a perfect measure, but it's a better measure than just straight scorebook averages or even just removing minnows IMO.

 

[Cevno]

It's interesting whether Cricketers would have liked to face Bangladeshi spinners on dry ,spinning tracks in recent times at home or the west indies struggling pace attack(begore roach and co.) on some of the most flattest wickets in the world in west indies.

And at the same time do we also remove the srilankan stadiums that are really flat?
And what about removing India when Zaheer and Harbhajan are not playing?

With Newzealand having lost in Bangladesh,do we remove them as well?

 

[bagapath]

My point is using your own judgment to pick a date wouldn't necessarily yield a worse (or better) date than just picking the day they win a game. Teams don't perform, or improve linearly.
.

india vs aus. kolkata 2001
leeds 2002
adelaide 2003
windies 2007
england 2007
no.1
australia in india 2010
india in SA 2011

 

[Ikki]

My position is that they are so weak that it really can skew a comparison between players. Especially players who have played them so much and some so little. It is manipulating stats but that should not take a pejorative meaning IMO. It is manipulation to come to a more accurate comparison IMO and that is fine.

However, my position is that if a player has played them enough times and has failed...it should blot their record.

 

[vcs]

india vs aus. kolkata 2001
leeds 2002
adelaide 2003
windies 2007
england 2007
no.1
australia in india 2010
india in SA 2011

India losing in Zimbabwe a few months after Kolkata 2001 and in WI a year or so after that puts a big hole in that theory..

 

[bagapath]

India losing in Zimbabwe a few months after Kolkata 2001 and in WI a year or so after that puts a big hole in that theory..

those are aberrations as big as india thrashing england twice in 1986 and beating west indies once in 1987. those wins didnt change the fact that the indian teams of that era were mediocre.

and these losses you have mentioned dont change the fact that india of the 00s were improving steadily.

 

[Top_Cat]

It's open for debate, sure, but I'd include them. The arbitrary definition 'Test standard' is one reason why. Think it goes without saying that a few FC sides (NSW, for example) would have had higher standard of play than some international teams, definitely Zim/Bangladesh and at various times WI and Pakistan. If you're going to omit them, why not include runs scored/wickets taken against FC teams like that? They're not playing in Tests because Tests are between countries (an arbitrary distinction for many strong FC teams) but runs scored/wickets taken against them, if played at an agreed-upon 'Test standard', should be counted if we're going to discount Zim/Bang for their standard of play, no?

tbh, dealing with this in data analysis is actually pretty straightforward. If you're going to insist on using the virtually worthless measures of 'runs scored' and 'wickets taken' without sampling, just weight different teams at different times differently. Come up with a multiplier for different teams and/or at different times in their history and apply it to any game any player plays against them. 250 runs against Bangladesh for Jan 2003 becomes worth less than 250 runs against England the next month. Easy. If you take a sample, it becomes even easier.

Of course, this is predicated on the assumption that any player doesn't improve or decline. Yikes @ taking that into account .

 

[bagapath]

My position is that they are so weak that it really can skew a comparison between players. Especially players who have played them so much and some so little. It is manipulating stats but that should not take a pejorative meaning IMO. It is manipulation to come to a more accurate comparison IMO and that is fine.

However, my position is that if a player has played them enough times and has failed...it should blot their record.

this is a dream come true. always wanted to stay on ikki's side and fight the bad guys as a team; like richard burton and clint eastwood coming together in "where eagles dare" or t.cruise and v.rhames in the MI series. and here is the moment.... i totally completely agree with your words here, mate. now, i hope my support doesn't make you change your mind.

 

[Prince EWS]

tbh, dealing with this in data analysis is actually pretty straightforward. If you're going to insist on using the virtually worthless measures of 'runs scored' and 'wickets taken' without sampling, just weight different teams at different times differently. Come up with a multiplier for different teams and/or at different times in their history and apply it to any game any player plays against them. 250 runs against Bangladesh for Jan 2003 becomes worth less than 250 runs against England the next month. Easy. If you take a sample, it becomes even easier.

Yeah, that's exactly what I was talking about in the few posts just above yours.

 

[Top_Cat]

Wouldn't use global averages as my weights, personally. Just manipulating the same poor measure to explain other poor measures.