Snowballs statistics

[tlcs]

Snowball, I notice the commencement of your blizzard of statistics coincided with your well publicised recent illness. I wish you a speedy recovery.

DITTO

Gordon Browns job , is up for offers....a new challenge for you sir

[Royal Rother]

Look, if someone says "You don't understand statistics" you can either agree, or say, "OH YES I DO!" (and then they say "Oh no you don't." or you can make the point you have formal qualifications in statistics, worked for a market research bureau and taught stats for a while at Uni.

Well the problem is that to us real statisticians, your statistics are sometimes embarassing. I get enough sarcasm about making stuff up, without someone claiming to be a statistician making stuff up with statistics in a public domain.

Poor. Really really poor.

What's been made up?

[roberto_11]

As for the original post on the thread, I'm doing an econometrics module which has a lot on regression this semester and I had a mid term exam on it yesterday, so I'm quite knowledgeable at the moment. Whilst leaving variables out of the OLS estimator (regression line) can result in omitted variable bias and make results less accurate, the change is very rarely significant enough to reverse the trend that the current regression line is showing. Including more relevant variables which are correlated with the currently included variables will normally just show that X1 is not solely responsible for values of Y, and the coefficient of X1 will decrease.

In non-economics speak, the stats on Hunt and Kebe accounting for a lot of assists fails to take into account factors such as form, opposition and surrounding players. These factors, if included, could take some of the credit away from Hunt and Kebe. However an Economics graduate should know that these omitted variables are highly unlikely to have such a great impact that the initial findings that we can discredit the fact that Hunt and Kebe are simply two of our most creative players and best options on the wing. Even if the coefficient, i.e. the level to which Kebe and Hunt are responsible for creating our goals, is considerably lower than it was initially when we include all variables, so long as it is still positive there is evidence that they have been our two most creative players this season. I imagine the value would even be large enough that it could pass a test at the 1% significance level that it was accurate (i.e. we could be 99% sure that Hunt and Kebe had been our most creative players so far this season.)

Snowball's statistics cannot be written off on the basis of omitting certain factors. The truth is this is a football board, do you expect people to come up with a load of complicated equations and formulas before going into hypothesis testing? Of course bloody not. Keep up the stats Snowball. They don't tell the full story, but they certainly tell a significant part of it and start debate.

Without boring everyone to death and turning this into geekville, Im going to point out that altough some factors are obviously less important than others (the ones you have chosen to back up your point), there are a few which will change your coefficient (change the level to which Hunt and Kebe are responsible for our goals). The main reason why "60% of all our goals have been scored or assisted by Kebe and SHunt" is because Coppell plays in such a way that the wingers are our main source of creativity, coupled with the fact that SHunt and Kebe have been our first choice wingers this season. Coppell has in effect built his team aound those 2 being the main souce of creativity.

Snowballs stat of 60% goals / assists is pretty irrelevant. The Center backs and goalkeeper are hardly gonig to contribute much. Central midfielders and full backs will only occasionally the way Coppell plays. Plus the strikers are mainly going to contribute to goals. If you look at any team you will find that the players the manager utilises as "creative" players will contribute more to assists/goals than the defensive ones.

I think you would find the stats changing if Coppell started playing Hunt at Left back and Kebe up front, or if e played Hunt and Kebe as wing backs, or if he dropped either of them, or if he changed hi style of play and used Matejovski in a free role as our main source of creativity (for example).

The only way of comparing the performance of Kebe SHunt is to replace them in exactly the same games, playing the same roles, with different players and see the effect they had. Impossible in fooball. Hence why its a subjective game.

[Dr Hfuhruhurr]

Look, if someone says "You don't understand statistics" you can either agree, or say, "OH YES I DO!" (and then they say "Oh no you don't." or you can make the point you have formal qualifications in statistics, worked for a market research bureau and taught stats for a while at Uni.

Well the problem is that to us real statisticians, your statistics are sometimes embarassing. I get enough sarcasm about making stuff up, without someone claiming to be a statistician making stuff up with statistics in a public domain.

Poor. Really really poor.

What's been made up?

His THIS IS NOT A SAMPLE is a classic example of someone trying to fool you.
All his stats are Post the conclusion and are set by his parameters of relevant time. i.e. 'This is what I think, now here are the stats' Classically 'Post Hoc'.

Nice try, though, Rother. Youre losing your touch

[Hoop Blah]

Poor. Really really poor.

What's been made up?

Go back through the last few weeks of posters highlighting errors and flaws in his stats and you'll plenty of mistakes.

Are they made up to suit his ramblings or just human error? Make your own mind up on that one, I'll give him the benefit of doubt as he seems little in need of it.

[Ian Royal]

I've noticed IMO that he shows inconsistent application of stats across difference subjects, even similar subjects sometimes.

Not only that, but he seems to ignore a lot of other relevant data. His stats are the be all and end all. I'm guilty of chucking insults around, but usually only if someone posts something ridiculously stupid, or if they fail to respond to the actual point made against their argument or so on.

I've recently tried to be nicer, to only question his posts where I disagree with them. To put up my own analysis and sometimes to challenge his data with some of my own. All that I've got from this is shouting, abuse and a refusal to actually address what I've said.

Apparently it's ok for him to ignore cup games (when it pleases him) but it's not okay for me to use them at all.

He's as guilty as any on here of attributing arguments to people when that has nothing to do with what they've said. And I still say half of what he includes in his posts is completely unnecessary and that he generall omits detailed analysis.

He'll also find something (for example Long's goals per minute) and post it as if it's deeply relevant. Then when anyone tries to argue against it there arguement is belittled and it's against his rules to allow you to bring in anything relevant like overall performance, or control. There is no why, there is simple statige

[roberto_11]

So to summarise, hes a retard

[Snowball]

Apparently it's ok for him to ignore cup games (when it pleases him) but it's not okay for me to use them at all.

There is no why, there is simple statige

statige?



I ignore cup games, and have explained why. You pretend "niceness" and deliberately add cup games to further confuse the issues and raise the temperature.

I made one exception and only then because the discussion was about CONSECUTIVE starts (and it included a single game)


It is PERFECTLY VALID to have an agreement, "Let's talk about the league: appearances, minutes played, goals" because then we are all on the same page and there is no confusion. We don't suddenly have an 8-0 win versus Lesser-Spotted Rovers FC or a 6-0 thumping by Arsenal (eg Sheffield United this year)... we don't have arguments about how good Reading's 11 were (usually not so good) or whether some Premier side put out their reserves.

Whereas, in the league we all know that we aim to put out the strongest possible team and play against another strongest possible team, and we know that teams are closer together in class and motivation

[Snowball]

So to summarise, hes a retard

Absolutely.

Here's a tip. I was taught, "An insult from a fool is a compliment."

Thanks for the compliment.

[Snowball]

His THIS IS NOT A SAMPLE is a classic example of someone trying to fool you.

OK, SIR. Please define, in statistical terms.

Sample

Sampling

Population

[roberto_11]

I note that you havnt actually respnded to my intelligent post earlier on this thread about Kebe and Hunt...

[Snowball]

I note that you havnt actually respnded to my intelligent post earlier on this thread about Kebe and Hunt...

I probably went into shock

[roberto_11]

thats great, but untill you do actually reply to people who are criticising the basis of your argument with an intelligent response, and not just some random insult, then you cant really be taken seriously

[loyalroyal4life]

snowball puts effort in with his stat, i like them, some are very interesting keep them up snow

[roberto_11]

Andy Hughes put effort in but couldnt cross a ball for shit.

[sheshnu]

Andy Hughes put effort in but couldnt cross a ball for shit.

Bit like Stephen Hunt, eh Snowball?

[Skyline]

Statistics prove nothing; and relying on them is foolish.

I'll give an example - if a team scored 46 goals in the whole season, they would, statistically at least, be towards the bottom of the goal scoring charts BUT if those 46 goals had resulted in 46 1-0 wins, that makes a mockery of that statistic.

In other words, its about context. A final thought - knowledge is knowing that a tomato is a fruit; wisdom is knowing that you don't put tomatoes into a fruit salad.

No, you are wrong. Goal-scoring is NOT a good predictor. It's goal SUPERIORITY that predicts.

Goal superiority predicts nothing. It's a reasonable guide to which is the better team, but not a predictor by any means.

Imagine a team that wins half its games 3-0 and loses the other half 1-0. GF=69, GA=23. GD=46. Points = 69. Identical goal diff to Mr Angry's hypothetical team, but 23 points fewer.

[Dr Hfuhruhurr]

*Warning*
*This post is boring and can be happily skipped by most*

OK, samples and populations.

You are using statistics to show Shane is better than Leroy. (As it happens, I agree, but I don’t believe in your applications of stats).

You have taken one set of 7 games for Shane, and another set of 7 games for Leroy calculated a number of statistics, and used it to conclude Shane is better than Leroy. It’s the conclusion part that requires an understanding of sampling.

So lets give it a go.
Your comparison to the general election is valid to a point. There is no disputing that Labour gained more seats than the conservatives at the last election. It’s a fact – such as the facts that you have chosen to compare striking ability between Shane and Leroy in your chosen seven games.

No disagreement here. (Although it is worth pointing out that at the last general election, the Tories and the Conservatives were fighting for seats in the same country, but Shane and Leroy’s statistics were gathered from different games (It would have been far more interesting if it was the same seven games))

But where do we go from here, Oh yes, Shane is better than Leroy and that we should favour Shane over Leroy. This is where your population definition disappears and when you make conclusions or predictions, then what you call populations are actually samples. What are we sampling from? Well the population doesn’t actually exist, but it would need to be the same for Leroy and Shane to make their stats comparable (as your 7 games vs 7 games obviously aren). So we make a population up, which exists purely for us to apply a variation to Shane’s and Leroy’s statistics, so we can estimate what it would be if Leroy had played in Shane’s games and vice versa, or what Leroy’s statistics would be if he played another seven games, or Shane’s, or whatever.

Most of your criticisms are that your ‘populations’ don’t compare. Using your definition of population, that is blatantly obvious. Lita’s 7 games are different to Long’s 7 games. If you start accepting that those 7 games are in fact samples from the same population, and embrace the variation this produces (or even just recognise there is some variation to your statistic when you use it for any form of prediction) then your statistics will be more purposeful.

Statistics just isn’t as black and white as you are presenting it.

As it happens, and I repeat for emphasis, I do think Shane is better than Leroy and am happy to see statistics supporting this, but there is nothing definite about what those statistics prove or predict.

[Snowball]

Goal superiority predicts nothing. It's a reasonable guide to which is the better team, but not a predictor by any means.

Imagine a team that wins half its games 3-0 and loses the other half 1-0. GF=69, GA=23. GD=46. Points = 69. Identical goal diff to Mr Angry's hypothetical team, but 23 points fewer.

Oh I DO so love it when people spout without thinking or checking their facts

The correlation in this year's Championship between goal-superiority and points per game is .9939

That is 99.39% totally predictive

Even if you don't bother to allow for the fact of different numbers of games played and just
look at goal superiority, the prediction is a staggering 91.3%

Columns are :



(GS) Goal Superiority
(TT) True Table (Points Won per Game)
(CH) Actual table positions -
(04) Games Played

GS TT CH PL

02 01 01 36 Wolverhampton Wndrs
06 03 02 36 Birmingham City
01 02 03 34 Reading
04 05 04 35 Sheffield United
03 04 05 33 Cardiff City
09 06 06 36 Preston North End
11 07 07 36 Burnley
08 09 08 36 Bristol City
05 08 09 35 Swansea City
07 10 10 36 Ipswich Town
12 11 11 35 Queens Park Rangers
16 13 12 36 Sheffield Wednesday
13 12 13 35 Coventry City
18 14 14 35 Doncaster Rovers
10 15 15 34 Crystal Palace
14 16 16 34 Derby County
19 17 17 36 Plymouth Argyle
20 20 18 36 Nottingham Forest
15 18 19 34 Watford
21 19 20 34 Barnsley
22 21 21 36 Blackpool
23 22 22 35 Southampton
17 23 23 36 Norwich City
24 24 24 35 Charlton Athletic

[Snowball]

*

You are using statistics to show Shane is better than Leroy. (As it happens, I agree, but I don’t believe in your applications of stats). You have taken one set of 7 games for Shane, and another set of 7 games for Leroy calculated a number of statistics, and used it to conclude Shane is better than Leroy. It’s the conclusion part that requires an understanding of sampling. So lets give it a go. Your comparison to the general election is valid to a point. There is no disputing that Labour gained more seats than the conservatives at the last election. It’s a fact – such as the facts that you have chosen to compare striking ability between Shane and Leroy in your chosen seven games.

No disagreement here. (Although it is worth pointing out that at the last general election, the Tories and the Conservatives were fighting for seats in the same country, but Shane and Leroy’s statistics were gathered from different games (It would have been far more interesting if it was the same seven games)) But where do we go from here, Oh yes, Shane is better than Leroy and that we should favour Shane over Leroy. This is where your population definition disappears and when you make conclusions or predictions, then what you call populations are actually samples. What are we sampling from? Well the population doesn’t actually exist, but it would need to be the same for Leroy and Shane to make their stats comparable (as your 7 games vs 7 games obviously aren). So we make a population up, which exists purely for us to apply a variation to Shane’s and Leroy’s statistics, so we can estimate what it would be if Leroy had played in Shane’s games and vice versa, or what Leroy’s statistics would be if he played another seven games, or Shane’s, or whatever. Most of your criticisms are that your ‘populations’ don’t compare. Using your definition of population, that is blatantly obvious. Lita’s 7 games are different to Long’s 7 games. If you start accepting that those 7 games are in fact samples from the same population, and embrace the variation this produces (or even just recognise there is some variation to your statistic when you use it for any form of prediction) then your statistics will be more purposeful.

Statistics just isn’t as black and white as you are presenting it. As it happens, and I repeat for emphasis, I do think Shane is better than Leroy and am happy to see statistics supporting this, but there is nothing definite about what those statistics prove or predict.

You are wrong. Based on the fact as you see them we could argue that the table itself is invalid!! Reading v Sheffield Wednesday was played on a different day to Wolves v Sheffield Wednesday, with different teams, each team on a different run, different weather, different referees, possibly different times of day. THEREFORE it is "totally unfair" to compare the matches.

But if that applies to games v Wednesday it applies to ALL games and thus we should not have points and goal difference but just have a load of professionals sit around a table and argue for which is the best team.


Ditto the season's top-scorer. Not fair. If Doyle ends up on 30 goals it's "meaningless" because he played with better players, and didn't have to beat Federici or Hanneman


We do the best we can with the information we have. I believe, longer run, that players starting games have a better chance of getting better and scoring more goals (in the long run). YES I am aware of those who (incorrectly) say Shane has never done well when starting. That is patently untrue. HOWEVER, it's likely he would get better with a longer run just as virtually all players do.

Meanwhile, all I can do is look at ALL the Reading League games played by Long (starts and as sub) and ALL the Reading League games played by Lita and compare them for minutes played, shots, goals scored etc.

You are also incorrect to say that comparison should be made when they are in THE SAME GAME, because then the position is further complicated because their performances are no longer independent. For example Lita may choose not to pass to Long who is in a tap-in position, blow it himself... so Lita fails but also causes Long to fail.

On Tuesday, S Hunt did well to look up and find Long. SHunt gets an assist and Long gets a goal, and Reading get three points.

SHunt might have tried to beat his man instead and we might have drawn, or worse, lost.

Lita from all the games I've watched, would have selfishly tried to grab the glory for himself