Moved Thread maximum Discrepant Anomalies In Coach OpinionS Index(D.A.I.C.Os index)

[McKuursma]

At the risk of defending Meteoric Rise, they are talking about maximum v minimum discrepancy, rather than absolute or actual discrepancy. The GWS v Pies example is a good one because the maximum discrepancy is +4/-4 (note it can go either way, an educated guess using team bias from the coach can be helpful but lacks reliability and repeatability) while the minimum discrepancy is 0 (2&2)

So while it can express the maximum possible discrepancy over a sample size of a season, the discussion on maximum upper discrepancy should be balanced by the maximum lower bound and minimum possible discrepancy.

It would also need to be applied at scale across the competition, then standardised to generate any useful comparison. It is then possible to generate the probability of funny buggers using variance against the mean, and then draw an inference using the data against environmental context

EDIT: I should add that the lack of transparency from what I've bothered to read in generating the maximum upper bound indicates expectation bias, while lack of consideration for the minimum and maximum lower indicates a likely confirmation bias

 

[WalshistheGOAT]

That outlier I keep banging on about - Daicos's 5-0 in the carl v coll match from round 21 last year raises huge questions about Michael Voss's integrity.

What round do the stop publishing the coach's votes? Presumably it was Daicos, Heeney, Cripps as the top 3 at that stage.

So in a crucial game both Cripps and Daicos step up to the plate. General consensus seems to be Cripps BOG, Daicos next best.

McRae votes 5 Daicos, 4 Cripps. You know what? I forgive him that. He's allowed to come up with a perfectly valid, perhaps a little forced, justification to see it that way. Collingwood won after all.

But Voss goes 5 Cripps, 0 Daicos. A crucial late round game with Cripps and Daicos neck and neck. And Voss votes like that.

At the end of round 20 it was Cripps 86, Daicos 94. At the end of round 21 it was Cripps 95, Daicos 99.

Of course all the Blues are off in hiding at the moment understandably, but WalshistheGOAT is a top contributor to this thread. Do you remember the game? Can you justify (presumably) Voss giving Naicos nothing?

Yeah, he wasn't really a factor that game outside the first quarter. Genuinely disappeared when Cripps and co started clawing back the margin in the 2nd half and had a few weak efforts in the 4th.

Cripps, Newman, Hoskin-Elliot, Hewett, Hill, Cameron and Howe were all far more influential from my memory of it.

Daicos should have been in the running for 1-2 votes but nowhere near a 5 from either coach. Of course we can infer pretty easily that it was Fly that gave him the 5 that game.

0 votes was far closer to fair than 5 if you watched that match.

 

[SC_Power]

You are free to suggest a better way to measure the integrity of coach votes from what is available to us.

Do you know who gives what votes in the Brownlow? The AFLPA? Any media award?
Really banging a non existent drum here, but if you believe it, it must be true.

 

[sabso]

At the risk of defending Meteoric Rise, he is talking about maximum v minimum discrepancy, rather than absolute or actual discrepancy. The GWS v Pies example is a good one because the maximum discrepancy is +4/-4 (note it can go either way, an educated guess using team bias from the coach can be helpful but lacks reliability and repeatability) while the minimum discrepancy is 0 (2&2)

So while it can express the maximum possible discrepancy over a sample size of a season, the discussion on maximum upper discrepancy should be balanced by the maximum lower bound and minimum possible discrepancy.

It would also need to be applied at scale across the competition, then standardised to generate any useful comparison. It is then possible to generate the probability of funny buggers using variance against the mean, and then draw an inference using the data against environmental context

EDIT: I should add that the lack of transparency from what I've bothered to read in generating the maximum upper bound indicates expectation bias, while lack of consideration for the minimum and maximum lower indicates a likely confirmation bias

Minimums will be 0s and 1s in most configurations- Zac Bailey necessarily having 2 difference in the example Mr Meow raised is going to be pretty rare I guess.

 

[HeathySure]

Cant believe people are taking this thread seriously. For one thing its based purely on speculation of what the OP thinks the coaches votes were issued so the premise is highly flawed already,
Then we have how the coaches vote and I am sure coaches see the best players of their side in different ways to what commentators and supporters see the game based on what they want their players to do. So how a supporter might view the game is completely different to how a coach might view the game and who the best players are.
And i am certain this thread was set up to shit on Daicos from a poster who has posted many times how Daicos is over rated, He is entitled to his opinion, but Daicos got Brownlow votes in 18 games last year, more than any other player including Cripps, and given the voting system being a 5 to 1 vote i think maybe he probably got voted correctly...

 

[Mr Meow]

That outlier I keep banging on about - Daicos's 5-0 in the carl v coll match from round 21 last year raises huge questions about Michael Voss's integrity.

What round do the stop publishing the coach's votes? Presumably it was Daicos, Heeney, Cripps as the top 3 at that stage.

So in a crucial game both Cripps and Daicos step up to the plate. General consensus seems to be Cripps BOG, Daicos next best.

McRae votes 5 Daicos, 4 Cripps. You know what? I forgive him that. He's allowed to come up with a perfectly valid, perhaps a little forced, justification to see it that way. Collingwood won after all.

But Voss goes 5 Cripps, 0 Daicos. A crucial late round game with Cripps and Daicos neck and neck. And Voss votes like that.

At the end of round 20 it was Cripps 86, Daicos 94. At the end of round 21 it was Cripps 95, Daicos 99.

Of course all the Blues are off in hiding at the moment understandably, but WalshistheGOAT is a top contributor to this thread. Do you remember the game? Can you justify (presumably) Voss giving Naicos nothing?

I feel like this was an instance where a 5 overall felt about right, while a 5 and a 0 individually were controversial. Based on MR's interpretation, the massive discrepancy in this case could only show extreme McRae favouritism.

Then any players who polarise opinions would automatically be considered their "coaches pet" based on discrepancies. It might be true in some cases but MRs method at its core is simply a method to determine which players split opinions more over a given season. Which you could show by running the same analysis on any two posters club MVP voting results (e.g MR and one of his peers).

There are some controversial coaches votes results just like Brownlow votes but over the course of the season it just about evens out. It's not like Daicos finished 1st in the coaches votes but 20th in the Brownlow.

 

[Mr Meow]

Cant believe people are taking this thread seriously. For one thing its based purely on speculation of what the OP thinks the coaches votes were issued so the premise is highly flawed already,
Then we have how the coaches vote and I am sure coaches see the best players of their side in different ways to what commentators and supporters see the game based on what they want their players to do. So how a supporter might view the game is completely different to how a coach might view the game and who the best players are.
And i am certain this thread was set up to shit on Daicos from a poster who has posted many times how Daicos is over rated, He is entitled to his opinion, but Daicos got Brownlow votes in 18 games last year, more than any other player including Cripps, and given the voting system being a 5 to 1 vote i think maybe he probably got voted correctly...

I am glad it is being ripped to shreds rather than having its "key findings and conclusions" accepted by default. You can do the same with any convoluted analysis MR spends hours putting together and defending. He has his hypothesis and then slaps a terrible method onto it until it just about fits.

 

[CharlieMortdecai]

Yeah, he wasn't really a factor that game outside the first quarter. Genuinely disappeared when Cripps and co started clawing back the margin in the 2nd half and had a few weak efforts in the 4th.

Cripps, Newman, Hoskin-Elliot, Hewett, Hill, Cameron and Howe were all far more influential from my memory of it.

Daicos should have been in the running for 1-2 votes but nowhere near a 5 from either coach. Of course we can infer pretty easily that it was Fly that gave him the 5 that game.

0 votes was far closer to fair than 5 if you watched that match.

I assume you would agree with this take Chronz , Cyclops

 

[WalshistheGOAT]

I assume you would agree with this take Chronz , Cyclops

Mind you Hill was probably more just threatening than actually positively impactful. Seemed like he was going to score every time he got the pill but continually fluffed up opportunities or tried being a bit too pretty with it.

But he was certainly more involved in the game as a whole.

 

[Chronz]

I assume you would agree with this take Chronz , Cyclops

WalshWoat is always crying about Naics. He deserves every vote he gets, it's more like oppo coaches don't give him enough!

 

[doppleganger]

I think it is pretty clear that in 2016 he played the type of high accumulation, lower impact style of footy that can catch the eye(& votes), but the ratings don't reward highly.

I think it is pretty clear that anyone who believes in the Player Ratings as a good judge of impact, would have to conclude that Dusty was massively overrated.

Outside top 40 in player ratings, Dusty was a low impact over-rates player for majority of his career according to PlaYer RaTinGs.

Nobody else comes close.

 

[Crankyhawk]

The only thing we know for sure is that a player receiving 10 votes was voted BOG by both coaches. That means two people agreed. It doesn't mean on that particular day the coaches had stronger integrity.

There are all sorts of combinations below that and there will be disagreements at times, as there would be between any two random people you might get to complete a 1-5 voting system. If there is stronger agreement between those people it does not mean they had higher integrity or vice versa.

It can be done to an extent with the weekly votes. Then highlight the seriously discrepant (where a 3 or more vote gap, particularly if it is 3-0 as a 4-1 or 5-2 is probably less extreme)

 

[Meteoric Rise]

At the risk of defending Meteoric Rise, they are talking about maximum v minimum discrepancy, rather than absolute or actual discrepancy. The GWS v Pies example is a good one because the maximum discrepancy is +4/-4 (note it can go either way, an educated guess using team bias from the coach can be helpful but lacks reliability and repeatability) while the minimum discrepancy is 0 (2&2)

So while it can express the maximum possible discrepancy over a sample size of a season, the discussion on maximum upper discrepancy should be balanced by the maximum lower bound and minimum possible discrepancy.

It would also need to be applied at scale across the competition, then standardised to generate any useful comparison. It is then possible to generate the probability of funny buggers using variance against the mean, and then draw an inference using the data against environmental context

EDIT: I should add that the lack of transparency from what I've bothered to read in generating the maximum upper bound indicates expectation bias, while lack of consideration for the minimum and maximum lower indicates a likely confirmation bias

Let me go through your points one at a time.

Your opening sentence is completely true and there are plenty on this thread making posts on the basis that it is compeltely false. I haven't made any serious statements claiming to know the actual discrepancies. At this point those cannot be known in many cases. We can of course just focus on the cases where it is known that there is a discrepancy, or where the actual level of the discrepancy is obvious, say where a player receives 9 votes, which can only be 5 from one coach & 4 from the other. And that would be useful. But what it would also do is leave a whole lot of less certain, but potentially just as instructive data unscrutinised. At the end of the day we could take the maximum, the minimum, or whatever, we are looking for obvious outliers. Or we are looking for the players/coaches at the top and bottom end of ranges, to give us some clues as to where to look further.

In your second para, I am interested in what you mean by "maximum lower bound" in this context. If you could explain that it might be useful. I do agree that providing both of the maximum & minimum discrepancies would be ideal, but it adds significantly to the workload. Also, we are not drilling for oil at this point, we are just searching for the terrain that is a higher risk to be productive.

Your 3rd para outlines conditions that are both beyond my capacity & expertise. Would love to see it done. But it wouldn't be my top priority. That honour would fall to Richmond's astonishing free kick situation since 2017.

Your 4th para if I am reading it correctly is asserting the figures I am providing are influenced by my expectation bias, & confirmation bias. I am human so I will be just as subject to any type of bias as the next person. There is almost certain to be errors in the figures I have provided. But remember, we are scanning the landscape at this point, rather than trying to mine it with precision.

But to make things more transparent my raw data comes from:

AFLCA Champion Player of the Year - AFL Coaches Association Awards - AFL Coaches Association

aflcoaches.com.au

My method in divining the maximum possible discrepancy is to:

1. look at how many votes the player has received.

2. If it is say 6 votes, I start from 5+1 and see if I can get the rest of the votes to add up. If they do, then that is the mximum possible discrepancy. If the rest of the votes don't allow for 5+1 then I move to 4+2 and so on until I find the highest discrepancy that is possibly true.

It is open to anybody to check for errors and correct them. To be any more transparent than that would require posting things that people are not going to be too excited to be reading, let alone the fact I can't be bothered.

There is really no need to be more transparent than that.

 

[Meteoric Rise]

I think it is pretty clear that anyone who believes in the Player Ratings as a good judge of impact, would have to conclude that Dusty was massively overrated.

Outside top 40 in player ratings, Dusty was a low impact over-rates player for majority of his career according to PlaYer RaTinGs.

Nobody else comes close.

Do they?

Show us the mid-forwards who were rated higher than him year by year.

 

[Mr Meow]

Let me go through your points one at time.

Your opening sentence is cmpletely true and there are plenty on this thread making posts on the basis that it is compeltely false. I haven't made any serious statements claiming to know the actual discrepancies. At this point those cannot be known in many cases. We can of course just focus on the cases where it is known that there is a discrepancy, or where the actual level of the discrepancy is obvious, say where a player receives 9 votes, which can only be 5 from one coach & 4 from the other. And that would be useful. But what it would also do is leave a whole lot of less certain, but potentially just as instructive data unscrutinised. At the end of the day we could take the maximum, the minimum, or whatever, we are looking for obvious outliers. Or we are looking for the players/coaches at the top and bottom end of ranges, to give us some clues as to where to look further.

In your second para, I am interested in what you mean by "maximum lower bound" in this context. If you could explain that it might be useful. I do agree that providing both of the maximum & minimum discrepancies would be ideal, but it adds significantly to the workload. Also, we are not drilling for oil at this point, we are just searching for the terrain that is a higher risk to be productive.

Your 3rd para outlines conditions that are both beyond my capacity & expertise. Would love to see it done. But it wouldn't be my top priority. That honour would fall to Richmond's astonishing free kick situation since 2017.

Your 4th para if I am reading it correctly is asserting the figures I am providing are influenced by my expectation bias, & confirmation bias. I am human so I will be as subject to any type of bias as the next person. There is almost certain to be errors in the figures I have provided. But remember, we are scanning the landscape at this point, rather than trying to mine it with precision.

But to make things more transparent my raw data comes from:

AFLCA Champion Player of the Year - AFL Coaches Association Awards - AFL Coaches Association

aflcoaches.com.au

My method in divining the maximum possible discrepancy is to:

1. look at how many votes the player has received.

2. If it is say 6 votes, I start from 5+1 and see if I can get the rest of the votes to add up. If they do, then that is the mximum possible discrepancy. If the rest of the votes don't allow for 5+1 then I move to 4+2 and so on until I find the highest discrepancy that is possibly true.

It is open to anybody to check for errors and correct them. To be any more transparent than that would require posting things that people are not going to be too excited to be reading, let alone the fact I can't be bothered.

There is really no need to be more transparent than that.

Your method hasn't revealed obvious outliers.

To do so you would need to calculate your maximum discrepancy for all players league wide, plot them on a chart and actually show that Daicos' or candidate X's value was way out of line with the rest of the cluster.

If anything, from your extremely limited and cherry picked data set, Heeney and Martin are closer to outliers than the rest. Then you seem to truly believe 25 from 106 (Daicos) is hugely different to 19 from 113 (Cripps). In a 22-23 game season, it really isn't.

"So overall the table looks like this so far:

2023 Daicos 25 votes max possible discrepancy from 106 votes total(65 v 41)
2024 Daicos 23 votes from 117 total, (72 v 49 )
2023 Butters 23 votes from 109 total(66 v 43)
2016 Dangerfield 23 votes from 121 total(72 v 49)
2024 Cripps 19 votes from 113 votes total(66 v 47)
2024 Heeney 12 votes from 112 votes total(62 v 50)

2017 Dusty 10 votes from 122 vots total(66 v 56)

It's beginning to look like Dusty & Heeney were the only ones with honest coaches! But there is a long way to go with this."

Then look at that insane conclusion at the end. You truly believe you are uncovering coaches votes dishonesty in a quantitative fashion.

 

[Meteoric Rise]

Do you know who gives what votes in the Brownlow? The AFLPA? Any media award?
Really banging a non existent drum here, but if you believe it, it must be true.

We do know who gives the votes in the Brownlow, of course we do. The match field umpires(who are known entities) meet post-game and collectively agree the votes is my understanding.

The AFLPA we do not know and in my case couldn't care less.

Media awards I don't follow but historically at least some of them we have been told which journo or "expert" cast the votes.

Norm Smith Medal we know exactly who votes for who, all 5 judges are listed individually. This of course is the most similar to the Coaches vote regime, where 2 parties each individully and separately judge the votes without influencing each other - so there is room for disparity in the votes.

As for banging on a non-existent drum, several posters on this thread alone have called for disclosure as to who is voting for who. So the drum exists outside of my banging on it.

Bit of a room of mirrors post from you on the whole.

 

[Meteoric Rise]

Your method hasn't revealed obvious outliers.

To do so you would need to calculate your maximum discrepancy for all players league wide, plot them on a chart and actually show that Daicos' or candidate X's value was way out of line with the rest of the cluster.

If anything, from your extremely limited and cherry picked data set, Heeney and Martin are closer to outliers than the rest. Then you seem to truly believe 25 from 106 (Daicos) is hugely different to 19 from 113 (Cripps). In a 22-23 game season, it really isn't.

"So overall the table looks like this so far:

2023 Daicos 25 votes max possible discrepancy from 106 votes total(65 v 41)
2024 Daicos 23 votes from 117 total, (72 v 49 )
2023 Butters 23 votes from 109 total(66 v 43)
2016 Dangerfield 23 votes from 121 total(72 v 49)
2024 Cripps 19 votes from 113 votes total(66 v 47)
2024 Heeney 12 votes from 112 votes total(62 v 50)

2017 Dusty 10 votes from 122 vots total(66 v 56)

It's beginning to look like Dusty & Heeney were the only ones with honest coaches! But there is a long way to go with this."

Then look at that insane conclusion at the end. You truly believe you are uncovering coaches votes dishonesty in a quantitative fashion.

This post is hilarious or sad Meow, unsure which.

Your post reveals things that are somehow familiar to me....maybe because I have already posted words to that effect.

You caught me out good with the insane conclusion at the end though. You may have missed the exclamation mark denoting mock excitement.

Welcome to the colourless world of Mr Meow, the man who does not know how to banter.

 

[Mr Meow]

This post is hilarious or sad Meow, unsure which.

Your post reveals things that are somehow familiar to me....maybe because I have already posted words to that effect.

You caught me out good with the insane conclusion at the end though. You may have missed the exclamation mark denoting mock excitement.

Welcome to the colourless world of Mr Meow, the man who does not know how to banter.

Ah so it's just a banter thread, no worries. It's not hilarious but it is sad, so I guess mission accomplished?

Once again failing to address any point. I'm guessing your profession didn't involve numbers, or if it did it was a lot more basic than trying to put together a quantitative assessment that made any sense whatsoever.

 

[craffles]

Awesome thread, would read again.

 

[Meteoric Rise]

Ah so it's just a banter thread, no worries. It's not hilarious but it is sad, so I guess mission accomplished?

Once again failing to address any point. I'm guessing your profession didn't involve numbers, or if it did it was a lot more basic than trying to put together a quantitative assessment that made any sense whatsoever.

Maybe this has never been explained to you Meow, but analysis & banter can(and very often do) exist side by side in the same discussion on this site, in football repartee in general, in Melbourne, perhaps even in Geelong, & in life in general. At the first sign of light-heartedness it doesn't automatically make all that preceeds or follows "banter."

My employment history has varied, thanks for asking.

I repossessed cars, that was about numbers, the number of cars I repossessed.

I did time in the Public Service. That was about numbers, the number of ways you could pretend something that went wrong was somebody elses fault.

And for 20 years I made my living betting on sports. I specialised in finding rapid imprecise ways to get the odds in my favour. That was very much about numbers. Sadly when the generation arrived who specialised in finding rapid precise ways to get the odds in their favour, I was out of business.

But why the sudden irrelevant interest in my vocationl history?

 

[sabso]

lol the projected ad hominem

 

[Mr Meow]

lol the projected ad hominem

You do realise you're contributing nothing other than ad hominem to the thread yourself, with posts like this?

 

[sabso]

You do realise you're contributing nothing other than ad hominem to the thread yourself, with posts like this?

You appear to be misunderstanding what ad hominem is.

 

[Mr Meow]

You appear to be misunderstanding what ad hominem is.

Your contributions in this thread have been more of a personal attack or commentary on me and my posting style, as opposed to engaging in the on topic points I have consistently raised. If you were here in good faith you would simply dispute whatever point I raise and leave it at that.

Basically you are just existing in this thread to talk about me. If the topic itself bores you, nobody is forcing you to participate.

 

[Mr Meow]

View attachment 2262210

Thanks for proving my point.