Leaving out the minnows...

[Quaggas]

You could equally say that including stats against them benefit those who succeeded against them.

The other angle is that runs and wickets against Bangladesh are, in the vast majority of cases, worth nothing in terms of helping your side win the match. Bangladesh haven't shown themselves to be capable of beating a test side in over 50 attempts, so it's reasonable to assume that South Africa will beat them whether AB De Villiers scores 250 or 0. In this case, why should De Villiers be penalised for scoring 0? His score, taken in isolation, is completely inconsequential, it neither helps nor hinders his team- the result is already close enough to predetermined as to make it reasonable to take it as inevitable, especially at home. So why should he be penalised for scoring 0?

Now, you could say a couple of things. Firstly, that it's not necessarily always the case, Vettori being the most obvious example. Secondly, that there are plenty of other inconsequential runs you might also want to exclude.

I'd argue that completely excluding Vettori's one-man show against Bangladesh would be unfair from a statistical perspective. You have to consider it to some extent.

As for the second part, runs that turned out to be inconsequential are a different thing from runs that were always going to be inconsequential. Bangladesh were, to all intents and purposes, always going to lose to South Africa. Not literally, but such an assumption is reasonable.

Don't take everything I say at face value, it's part of a wider point. Essentially, I'm asking, why should a player's failure against Bangladesh count against him? His failure doesn't affect his team's welfare even slightly.

So how about ABdV 250 vs 0, away, against India or Aus. Should that count for more than against a mid-table side? In the end, why go extreme (all or nothing) - why not count more or less heavily depending on strength of opposition, or even based on some measure of the importance of the score in the context of the game? An outing against Zim tells you something, you just might not want to attach that much importance to it. But why 0 and not 0.1?

The weight also depends on what question you're asking. If I wanted to compare ABdV's "mental strength score" against player X's, I'd want to weight their respective away performances in a 4th innings chase against Aus more than in a score in a 99.9% sure (from here on out) to win situation. If I want to compare their ability to smash the ball about "when it doesn't matter," the weights are different again and you could make a case for including warmup games (ideally conditional on instructions from the coach).

For purposes of casual conversation, I think one can always include/exclude/attach more/less importance to any innings provided a defensible rationale is given within the context of the discussion. The prior should always be an equal weighting for all innings, though. The "burden of proof" lies with the person arguing to "exclude."

Edit: Preempted by post just above.

 

[Prince EWS]

What about, rather than excluding totally the Bangladeshis (and Zimbabweans), is there any value in applying a weighting to Bangladesh's stats?

i.e. if, on average, Bangladesh wickets cost 10% less than the average cost per wicket over the time of their playing test cricket, should bowler's stats against them be inflated by 10%? This would ensure that players who have played badly against Bangladesh/Zimbabwe continue to have their records penalised accordingly, whilst the vagueries of players who have excelled against Bangladesh are somewhat normalised, whilst recognising that they did still perform well against a lesser standard team.

Yeah, I've always thought this had a lot of merit. I've actually attempted to do it a few times, tbh. The problem though is that you get to a point where you start asking how much weighting Australian First Class cricket should be given, and eventually you're querying as to how to incorporate Namibian 3nd grade club cricket. It just never ends. I eventually came to the conclusion that it was much easier, as a starting point, to look at a player's record against all the teams that are of a vaguely similar standard at the top of the tree, and then analyse the circumstances of the performances from that point, before then looking at their performances at lower levels (domestic cricket and Bangladesh games included).

If you're just interested in how good a Test player someone was, though, and not how good a player they were overall, giving a weighting to each team (Bangladesh included) based on how that team performed two years either side of the encounter or something similar would definitely be worth considering. Bangaldesh are a Test team so performing against them is part of what makes a good Test player - it's just significantly easier. The problem with this is that it's extremely difficult to measure on a large scale.

 

[Athlai]

But Will it isn't trying to incoporate every FC game played by a player just those that have been labelled as Tests. Surely applying a weighted average to knocks against Bangers and Zim wouldn't imply that you HAVE to even mention Australian FC stuff.

 

[Uppercut]

This is clearly a case of ridiculous extrapolation of a point as if you're willing to view this failure as inconsequential, the whole batting order may as well go out there and all get nought. I realise that's not what you're saying, but to view an innings as inconsequential as the result is 'predetermined' is total bollocks.

Yeah, that's why I said "when taken in isolation". De Villiers' innings alone plays no part in whether South Africa beat Bangladesh.

The result obviously isn't literally predetermined, but South Africa beating Bangladesh is likely enough that it's reasonable to take it as such. With a record like W 0 L 52, the chance of Bangladesh even drawing, barring weather, is somewhere <2%- and with a side of the quality of South Africa, it's at the lower end of that range. What odds do bookies give?

As I say, it's not to be taken at face value, I'm reducing it to the absurd to some extent. But you can see why, if trying to weight a De Villiers' innings against Bangladesh, the fact that his team had a 98+% chance of winning regardless of how many he scored means the weighting has to be very small- so small, in fact, that you might as well just leave it out.

 

[HeathDavisSpeed]

As I said, that wasn't even relevant to the point I was making anyway. Why not weight performances versus Bangladesh rather than exclude them?

 

[Uppercut]

So how about ABdV 250 vs 0, away, against India or Aus. Should that count for more than against a mid-table side? In the end, why go extreme (all or nothing) - why not count more or less heavily depending on strength of opposition, or even based on some measure of the importance of the score in the context of the game? An outing against Zim tells you something, you just might not want to attach that much importance to it. But why 0 and not 0.1?

The weight also depends on what question you're asking. If I wanted to compare ABdV's "mental strength score" against player X's, I'd want to weight their respective away performances in a 4th innings chase against Aus more than in a score in a 99.9% sure (from here on out) to win situation. If I want to compare their ability to smash the ball about "when it doesn't matter," the weights are different again and you could make a case for including warmup games (ideally conditional on instructions from the coach).

For purposes of casual conversation, I think one can always include/exclude/attach more/less importance to any innings provided a defensible rationale is given within the context of the discussion. The prior should always be an equal weighting for all innings, though. The "burden of proof" lies with the person arguing to "exclude."

Tbh, if you wanted a better idea of how good a batsman is, you could try that.

Personally, I wouldn't look at how good the attacks a player scored against were. Succeeding against Warne-McGrath but getting out a lot to Phil Tufnell isn't my idea of a good batsman. It might be worth looking at how useful the runs scored were to the team in context, but it would be extremely awkward to do. Taking out the near-worthless runs against Bangladesh wouldn't give you a fantastic statistic that would settle all X vs. Y arguments forever. But it would be a better measure than average with Bangladesh, no?

 

[Uppercut]

As I said, that wasn't even relevant to the point I was making anyway. Why not weight performances versus Bangladesh rather than exclude them?

Just because the weighting given to the runs would have to be so small that it would barely be any different to the figure excluding Bangladesh. How much would you say runs against Bangladesh are worth to the team compared to runs against someone else?

 

[Athlai]

Just because the weighting given to the runs would have to be so small that it would barely be any different to the figure excluding Bangladesh. How much would you say runs against Bangladesh are worth to the team compared to runs against someone else?

It actually opens the question about weighting runs against all teams. Which is kind of what the ICC rankings try to do.

 

[HeathDavisSpeed]

Just because the weighting given to the runs would have to be so small that it would barely be any different to the figure excluding Bangladesh. How much would you say runs against Bangladesh are worth to the team compared to runs against someone else?

Take the average per wicket against Bangladesh, and the average per wicket in all other test matches. The percentage difference would be your inflation/deflation factor to weight performances (for bowlers). Similar for batsmen, but you would take the average per wicket acheived by the team facing up against Bangladesh etc.

 

[HeathDavisSpeed]

"near-worthless runs" indeed. That's why everyone scores centuries against them. May as well line up with a team of Chris Martins as they'd own against Bangladesh. Really.

 

[Prince EWS]

Take the average per wicket against Bangladesh, and the average per wicket in all other test matches. The percentage difference would be your inflation/deflation factor to weight performances (for bowlers). Similar for batsmen, but you would take the average per wicket acheived by the team facing up against Bangladesh etc.

Yeah, I was devising a system that did exactly this at one point, although it attempted to do it for every single match of a player's career against any opposition. For example, a match played against Australia in 1997 would take the Australian overall batting and bowling averages from 1995-1999 and standardise them to meet the all-time global average.

I never did finish the program though.

 

[Uppercut]

Take the average per wicket against Bangladesh, and the average per wicket in all other test matches. The percentage difference would be your inflation/deflation factor to weight performances (for bowlers). Similar for batsmen, but you would take the average per wicket acheived by the team facing up against Bangladesh etc.

Well if you did, the figure would at most be the difference of one average point higher or lower than their average w/o minnows.

In any case, that's not how I'd do it. I'm not that interested in how difficult the runs are to score. A batsman who scores six sixes off McGrath then gets out to Cameron White is no better than a batsman who does it the other way round. I'm interested in how much their runs were worth to their team, how much their runs make some kind of a difference between winning and losing. Against Bangladesh, they barely ever make that difference. Vettori is a notable exception.

 

[Athlai]

Well if you did, the figure would at most be the difference of one average point higher or lower than their average w/o minnows.

In any case, that's not how I'd do it. I'm not that interested in how difficult the runs are to score. A batsman who scores six sixes off McGrath then gets out to Cameron White is no better than a batsman who does it the other way round. I'm interested in how much their runs were worth to their team, how much their runs make some kind of a difference between winning and losing. Against Bangladesh, they barely ever make that difference. Vettori is a notable exception.

Again this is what the ICC rankings do isn't it?

 

[Uppercut]

"near-worthless runs" indeed. That's why everyone scores centuries against them. May as well line up with a team of Chris Martins as they'd own against Bangladesh. Really.

No. But you might as well replace the single player in question with Chris Martin. It wouldn't make any difference.

De Villiers scoring a century for South Africa against Bangladesh helps them win the match in the same way me voting Conservative would help them win the next election. It wouldn't work if nobody did it, but whether one person does it isn't usually relevant.

 

[Uppercut]

Again this is what the ICC rankings do isn't it?

I don't rightly know

 

[Athlai]

It's mostly a form guide which is its biggest downfall. Good at determining who is the best in the world right now, crap at examining a players impact overall. Players that peaked get greater preference to ones who performed over a very long period of time which makes the all-time rankings not that helpful. Would be interesting if they could apply an average rating for a player over their career though.

 

[HeathDavisSpeed]

Against Bangladesh, they barely ever make that difference. Vettori is a notable exception.

What about the contributors to Pakistan's one-wicket win over the Bangas (Inzamam-ul-Haq scoring a century, for example).

Also, by keeping the stats in and applying an inflation/deflation factor ensures that those who have performed badly continue to have the stain on their record, rather than having it stricken from them as you would intend. Harby's average against them would increase from 40-odd to maybe 50-odd with an inflationary factor.

 

[Uppercut]

What about the contributors to Pakistan's one-wicket win over the Bangas (Inzamam-ul-Haq scoring a century, for example).

Also, by keeping the stats in and applying an inflation/deflation factor ensures that those who have performed badly continue to have the stain on their record, rather than having it stricken from them as you would intend. Harby's average against them would increase from 40-odd to maybe 50-odd with an inflationary factor.

So you'd have a system where failure against Bangladesh counts more and success against them counts less? Do explain it.

Had no idea that was what you were implying. I thought you meant that, say, if figures against Bangladesh account for 5% of a bowler's average, the deflation factor would mean that it only impacts on it to the tune of (for example) 2%- so 5/100 in the bowler's overall figures would become 2/40.

Oh, and yes, Inzamam et al is another notable exception.

 

[Top_Cat]

Just because the weighting given to the runs would have to be so small that it would barely be any different to the figure excluding Bangladesh. How much would you say runs against Bangladesh are worth to the team compared to runs against someone else?

Depends on the weighting criteria. Your multiplier might be a compendium of all sorts of stuff other than the strength of the opposition. Plus, any analysis of this kind would have to take into account the inherent limitations of sports stats as predictors, etc.

 

[HeathDavisSpeed]

So you'd have a system where failure against Bangladesh counts more and success against them counts less? Do explain it.

Had no idea that was what you were implying. I thought you meant that, say, if figures against Bangladesh account for 5% of a bowler's average, the deflation factor would mean that it only impacts on it to the tune of (for example) 2%- so 5/100 in the bowler's overall figures would become 2/40.

Oh, and yes, Inzamam et al is another notable exception.

What? No. Say, the average per wicket vs. all other teams is 50 and the average per wicket vs. Bangladesh is 40 (percentage difference of 20%) then I would inflate the cost of all players wickets against them by 20%. Vettori's would go up from 24-ish to 29-ish and Harbhajan's would go up from 40-ish to 48-ish (original average * 1.20)

This adjusted figure would be then included in their overall stats to give a Vettori overall average of, perhaps 35 rather than the 37 you get if you totally exclude Bangladesh's results; whilst Harbhajan's would hardly be impacted due to the small sample size.

What I really don't understand is why its great to take cheap wickets off Vettori to make him look worse whilst also taking away poor performances by Harby to make him look better.