Goal scoring problems?

[Sarah Star]

Coppell's policy was "Clean sheet. Score goals. Win game"

[Snowball]

Simple "Goals Scored" is NOT a very good predictor of league position.

In the data from the top six for the last six years the goals scored
produced a correlation of only .44 with league position.

For example, one club came second, automatic promotion
with the LOWEST goals scored of any club in the top six,
the sixth highest scorer (6th in the whole six years) came 5th.
The 11th highest scorer finished 6th.

A prediction of .44 is not very good. If you like, simplistically,
goals scored is about 44% accurate in predicting league position.

Goal DIFFERENCE, is far more accurate and has a correlation of .7
or is 70% accurate

[Snowball]

Coppell's policy was "Clean sheet. Score goals. Win game"

In fact, "Clean sheet, score GOAL" suffices.

[Murts-is-Lej]

In the data from the top six for the last six years the goals scored
produced a correlation of only .44 with league position.
<snip>
Goal DIFFERENCE, is far more accurate and has a correlation of .7
or is 70% accurate

Any chance of doing the same calculation with goals conceded? I'd guess the answer is somewhere between .44 and .7, i.e. a good defence is better than a free-scoring attack... Even more so in the bottom 6?

[Snowball]

.70 correlation Goal Difference
.53 correlation Goals Conceded
.44 correlation Goals Scored


So if you want to predict, take GD

Failing that, the better defences finish higher.

The WORST predictor is goals scored.

[Murts-is-Lej]

.70 correlation Goal Difference
.53 correlation Goals Conceded
.44 correlation Goals Scored


So if you want to predict, take GD

Failing that, the better defences finish higher.

The WORST predictor is goals scored.

Trouble is GD is a relative measure, so not particularly useful for analysing a team.
Does confirm my impression, though, that a better defence is more valuable than a good attack - Gylfi for Harte+Khiz+cash anyone?

[Snowball]

In the data from the top six for the last six years the goals scored
produced a correlation of only .44 with league position.
<snip>
Goal DIFFERENCE, is far more accurate and has a correlation of .7
or is 70% accurate

Any chance of doing the same calculation with goals conceded? I'd guess the answer is somewhere between .44 and .7, i.e. a good defence is better than a free-scoring attack... Even more so in the bottom 6?

What I was doing was a bit misleading (looking just at the top six and combining seasons)

If you look at the final table for last season you get

72.6% Prediction GOALS SCORED (Worst Predictor)

78.7% Prediction BEST DEFENCES (Second Best)

90.8% Prediction GOAL DIFFERENCE. (BEST Predictor)


A correlation of .91 is really very, very accurate when you think of all the variables
that impact a league in a season, freak wins, freak defeats etc.

[Snowball]

This early in the season you might expect that the
correlation between goal-difference and league
position might be lower. I certainly expected that
as we haven't had time for GDs to settle down
and smooth out odd results like the 1-6 at Millwall
or 0-4 at Sheffield.

Surprisingly, the correlation is .89, 89% predictability already!

Pos-GD

01 21
02 8
03 4
04 9
05 4
06 7
07 5
08 2
09 1
10 -2
11 2
12 1
13 -2
14 3
15 -3
16 -6
17 -1
18 -7
19 -7
20 -6
21 -2
22 -10
23 -10
24 -11

[Ian Royal]

Snowball... if you don't want to talk about this topic, why don't you just oxf*rd off?

You clearly don't get the whole point. Stop spouting your irritating shit and spamming a thread you clearly have no interest in. If it dies, fine I couldn't give a monkey's, it means no one is interested in the point of the thread.

To spell it out... This is a look at whether we are likely to struggle to score goals over the season. And whether we will necessarily miss a prolific goalscorer upfront. Not where we will finish. Your goal difference has no relevance to this.

[Snowball]

Snowball... if you don't want to talk about this topic, why don't you just oxf*rd off?

You clearly don't get the whole point. Stop spouting your irritating shit and spamming a thread you clearly have no interest in. If it dies, fine I couldn't give a monkey's, it means no one is interested in the point of the thread. To spell it out... This is a look at whether we are likely to struggle to score goals over the season. And whether we will necessarily miss a prolific goalscorer upfront. Not where we will finish. Your goal difference has no relevance to this.

Very funny coming from a guy who spouts cr@p without going to (many) games.

To spell it out... GD IS relevant to "goal-scoring" because you DO NOT HAVE TO HAVE LOTS OF GOALS FOR SUCCESS.

Last season Reading scored 68 goals. Swansea scored 40.

To help you, that means Reading scored 28 goals MORE than Swansea

Swansea were 6 points above Reading. OH!

Do you know WHY, Ian? It's because they played a very tight, maximum-possession 4-5-1
and only let in 37 goals. That was the second-best goals-against record after Newcastle.

Do you SEE? Mac is playing a style which keeps our goals against DOWN.

That means less chances for the strikers. The club stats show that Shane has had just 11 chances this season.

Play 4-4-2 and probably you'll have more chances for the strikers, but you'll give the opposition more, too.

You see, I don't think there IS a goal-scoring problem (and therefore, in a free country, I think I have the right
to discuss that issue). True it'd be nice to see 1-2 more goals (as long as goals against don't increase) but we
are seventh after a poor start, so enough goals are going in (at the moment.)

[Platypuss]

Why not focus on eg. Marcus Williams' (non-)appearances?

Shhhhhh!!!!!!!!!!!!

[PEARCEY]

Potential HoF thread. Ramp it up IR/Snowball.

[Ian Royal]

Potential HoF thread. Ramp it up IR/Snowball.

nah mate, I didn't take a break to come back to argue with the type of "challenging customers" i get all day at work. Gone are the days of lengthy wars of words... I don't have the energy for it anymore, it's just a bit depressing.

[cmonurz]

.44 correlation Goals Scored

Can you run me through that calculation?

Here's last season's Championship table.

http://stats.football365.com/2010/ENG/D1/glsfor.html

Now I'm no maths dork, but only three teams have a differential of more than 5 between their league position and goals ranking, and the majority only 2 or 3 places different. How is that only a 44% correlation?

[Snowball]

.44 correlation Goals Scored

Can you run me through that calculation?

Here's last season's Championship table.

http://stats.football365.com/2010/ENG/D1/glsfor.html

Now I'm no maths dork, but only three teams have a differential of more than 5 between their league position and goals ranking, and the majority only 2 or 3 places different. How is that only a 44% correlation?

You're right, cmonurz, I posted the stats for the table last year (all 24 clubs) and had correlations of .71 Goals .79 (Defence) and .91 (GD)


The stats I looked at was a block of JUST the top six teams for the last six years.

I posted to say that this data was somewhat misleading and posted the figures for all-clubs, one season

[Snowball]

Here's last season's Championship table.

http://stats.football365.com/2010/ENG/D1/glsfor.html

Now I'm no maths dork, but only three teams have a differential of more than 5 between their league position and goals ranking, and the majority only 2 or 3 places different. How is that only a 44% correlation?

Here's the post from earlier

What I was doing was a bit misleading (looking just at the top six and combining seasons)

If you look at the final table for last season you get

72.6% Prediction GOALS SCORED (Worst Predictor)

78.7% Prediction BEST DEFENCES (Second Best)

90.8% Prediction GOAL DIFFERENCE. (BEST Predictor)


A correlation of .91 is really very, very accurate when you think of all the variables
that impact a league in a season, freak wins, freak defeats etc.

[Arch]

wng

One of many Coppellisms I'm a big fan of.

I use many of them. I found myself saying just this week "we don't want the highlight of our season to be a Sunday in October".

[Hoop Blah]

If our defence doesn't tighten up we'll need to find more goals.

If we suddenly become hard to score against then perhaps we won't miss that consistent goal threat from our centre forward (whichever of our options are selected, they're all much of a muchness anyway). The two are linked though. The less we score the more pressure the defence come uder to win us points, and physical pressure during the game if we can't keep the ball in the opposition half, and so we're likely to leak more goals.

As ever, a balanced team with decent players will come good. Our squad is still a bit unbalanced in terms of the right quality in a couple of areas. It is a lot better than it was before the end of the transfer window though.

[Maguire]

For a professional statistician, as you claim to be

Wut

Methinks he's winding you up

[Gordons Cumming]

Simple "Goals Scored" is NOT a very good predictor of league position.

Correct. In fact Middlesbrough finished bottom of the First Division in 1928 and scored a massive 81 goals in the process.
They conceded only 88! That season there was only 2 points seperating the bottom 9 clubs. What a season!
Boro had 37 points only 16 points from the champions (Everton) that season.(2 points for a win)

HTH