Goal Difference!

[Rev Algenon Stickleback H]

I said IF we maintain this GD (would mean a GD of 23) THEN we'd make the play-offs.

Taking the last 8 seasons from http://www.rsssf.com, showns that in the 24 team leagues, a team with the 3rd best goal difference missed the top 6 twice.
True, that shows that 22 times out of 24, the 3rd best team got in, but it's not something you can take for granted.

22 successes v 2 fails = 11-1.

Almost exactly what a correlation of .91 predicts

Yes. A 91% chance. Very likely, but not a given.

[Snowball]

Woah... my normal conversational statements

eg "We'll probably win the league"

are still just statements.

I base mine partly on stats but I'm bound by the same wooliness of conversation. "Probably" by some definitions means "almost certain." It doesn't by mine, certainly not while chatting to a pile of Nov=bbers who misread and misquote anyway.

"Probably", in this context means, "I reckon it will happen."

It's more a "betting cert" than "possibly" and less than certain

Jeez I'm not at Uni here

[Rev Algenon Stickleback H]

But the thread started to make the point that our then current GD suggested we were a better team than our position actually showed. I still maintain that. This is why I say that if we continue our goal-scoring and defensive rates we will finish in the top six DESPITE THE FACT that right now our third-best GD has us in seventh.

It's a nice theory, but have you actually checked to see if there's any data to back it up?

I mean, you are taking the fact that our points tally seems lower than our goal difference suggest it should be, as proof that we are underperforming. Could it also be that it's our goal difference that's the anomoly, not the points?

If you look back at the mid-points of previous seasons and look at teams who have a high GD and a less impressive points tally, do the stats show those teams tending to do better in the second half of the season? It might show their GD normalising instead.

[Snowball]

Yes. A 91% chance. Very likely, but not a given.

I oxf*rd KNOW that. That's what .91 MEANS. It ain't definite. And I've been posting that .91 or 91% predictability all through the thread.

I've even explained my language.

For example, I shall say:

"If we beat Hull and don't get beaten at Cardiff, we will make the play-offs. If we beat Hull, win at Cardiff and beat QPR, we'll finish second."

That is a belief, a conviction. Every normal human being (and the occasional Nobber) KNOWS that I don't mean WILL in an absolute, categorical, nothing-else-is-possible way.

[Rev Algenon Stickleback H]

Yes. A 91% chance. Very likely, but not a given.

I oxf*rd KNOW that. That's what .91 MEANS. It ain't definite. And I've been posting that .91 or 91% predictability all through the thread.

I've even explained my language.

For example, I shall say:

"If we beat Hull and don't get beaten at Cardiff, we will make the play-offs. If we beat Hull, win at Cardiff and beat QPR, we'll finish second."

That is a belief, a conviction. Every normal human being (and the occasional Nobber) KNOWS that I don't mean WILL in an absolute, categorical, nothing-else-is-possible way.

A belief is not statisitical evidence, yet that's how you present it.

[Snowball]

It's a nice theory, but have you actually checked to see if there's any data to back it up? I mean, you are taking the fact that our points tally seems lower than our goal difference suggest it should be, as proof that we are underperforming. Could it also be that it's our goal difference that's the anomoly, not the points?

If you look back at the mid-points of previous seasons and look at teams who have a high GD and a less impressive points tally, do the stats show those teams tending to do better in the second half of the season? It might show their GD normalising instead.

. I

But you could ALSO note that our GD has been growing as the season progresses. I don't mean solely in total terms, I mean in terms of Goal Supremacy per game.

It was zero for the Gylfi games... for the next 14 games it was .43 goals per game, since Elwood arrived it's been .88 goals per game. That AVERAGES .5 goals per game but the average hides a real improvement, a tighter defence AND goal-scoring

As I've said, a trend line on a computer using our position throughout the season has us finishing first or second (but I am NOT saying that will happen)

[Snowball]

A belief is not statisitical evidence, yet that's how you present it.

Jesus! NO.

I PRESENT MY BELIEFS

BASED ON

the statistical evidence.

That's your interpretation.

You might make your predictions based on gut feel, or a feeling that RFC "always come good in the last third of the season" or "always fall away like QPR"

I base mine on watching the team, seeing improvements (or not) and ALSO using stats to overcome my personal biases.

So, for example, I stuck forty quid on Shane today to be CCC top-scorer, NOT because I believe he will, but because he represented decent odds (25-1)... that's mostly gut-feel and a genuine belief that he's going to finish near 20 goals and that it's quite likely that many of those above will have a bad patch.


So what improvements? Griffin solid as a rock, great cover from Cummings, Feds getting better and better, the CB partnership improving every game (and Ivar back as cover), Harte stats giving a lie to the idea he's always skinned, ELWOOD making us a far FAR better midfield (and Karacan loving playing with him), Hunt and Long forming a great partnership, McAnuff more up for it and getting some goals, Kebe still to get back to his top-top form but showing signs that he's close.

[SLAMMED]

[weybridgewanderer]

A belief is not statisitical evidence, yet that's how you present it.

Jesus! NO.

I PRESENT MY BELIEFS

BASED ON

the statistical evidence.

That's your interpretation.
I base mine on watching the team, seeing improvements (or not) and ALSO using stats to overcome my personal biases.

You might make your predictions based on gut feel, or a feeling that RFC "always come good in the last third of the season" or "always fall away like QPR"


So, for example, I stuck forty quid on Shane today to be CCC top-scorer, NOT because I believe he will, but because he represented decent odds (25-1)... that's mostly gut-feel and a genuine belief that he's going to finish near 20 goals and that it's quite likely that many of those above will have a bad patch.


So what improvements? Griffin solid as a rock, great cover from Cummings, Feds getting better and better, the CB partnership improving every game (and Ivar back as cover), Harte stats giving a lie to the idea he's always skinned, ELWOOD making us a far FAR better midfield (and Karacan loving playing with him), Hunt and Long forming a great partnership, McAnuff more up for it and getting some goals, Kebe still to get back to his top-top form but showing signs that he's close.

Feel free to present your statistics to back up your case.

When people present other "evidence", sometimes different statistical info, sometimes historical info, don't bang on and on and on and on and on and on about how much more right your statistic is than what everyone else suggest, and then say its not a belief its only "if"

You keep reverting to "if we keep winning games we will finish high up the league with a strong goal difference"

No shit sherlock!

[Arch]

Statistics shift probabilities, and no more than that. They should always be treated as doing that. So we have a number of indicators about where we belong - points accrued (7th), goals for (is it 8th?), goals against (5th), some others. Along comes GD (3rd), and the strong statistical correlation with position. That shifts the overall probability of a top six finish upward, at least a bit. But suppose, for the sake of argument, the correlation between goals scored and league position is weaker than that for GD, but still positive. Now how do you measure your 0.91 for GD against (say) 0.65 for goals scored? Do you ignore the latter as being weaker or do you say it contributes a modest shift downward in our expectations, counterbalanced by the somewhat stronger shift upward from GD?

[Harpers So Solid Crew]

Just because I wanted to see it again.

[Snowball]

Statistics shift probabilities, and no more than that. They should always be treated as doing that. So we have a number of indicators about where we belong - points accrued (7th), goals for (is it 8th?), goals against (5th), some others. Along comes GD (3rd), and the strong statistical correlation with position. That shifts the overall probability of a top six finish upward, at least a bit. But suppose, for the sake of argument, the correlation between goals scored and league position is weaker than that for GD, but still positive. Now how do you measure your 0.91 for GD against (say) 0.65 for goals scored? Do you ignore the latter as being weaker or do you say it contributes a modest shift downward in our expectations, counterbalanced by the somewhat stronger shift upward from GD?

Some logic errors in their Arch

Obviously it is generally true that the more goals you score the better you do (EXCEPT...)
Obviously it is generally true that the less goals you concede the better you do (EXCEPT...)

I've posted in the past that both of these have poorer correlations with final position than GD

But really that's not surprising. Some teams score loads of goals but concede loads, maybe even more than they score.
Some teams are very defensive, can shut out the opposition but don't win enough because they don't score enough

That's why GD is a better predictor.

BUT, you can't factor in goals scored and goals conceded and GD because the latter already includes the two former.

[Arch]

Statistics shift probabilities, and no more than that. They should always be treated as doing that. So we have a number of indicators about where we belong - points accrued (7th), goals for (is it 8th?), goals against (5th), some others. Along comes GD (3rd), and the strong statistical correlation with position. That shifts the overall probability of a top six finish upward, at least a bit. But suppose, for the sake of argument, the correlation between goals scored and league position is weaker than that for GD, but still positive. Now how do you measure your 0.91 for GD against (say) 0.65 for goals scored? Do you ignore the latter as being weaker or do you say it contributes a modest shift downward in our expectations, counterbalanced by the somewhat stronger shift upward from GD?

Some logic errors in their Arch

Obviously it is generally true that the more goals you score the better you do (EXCEPT...)
Obviously it is generally true that the less goals you concede the better you do (EXCEPT...)

I've posted in the past that both of these have poorer correlations with final position than GD

But really that's not surprising. Some teams score loads of goals but concede loads, maybe even more than they score.
Some teams are very defensive, can shut out the opposition but don't win enough because they don't score enough

That's why GD is a better predictor.

BUT, you can't factor in goals scored and goals conceded and GD because the latter already includes the two former.

Not sure what you're counting as "logic errors". What I wrote is compatible with your reply up as far as the last line, which I can't figure out because I don't know how you're using "factor in". The "EXCEPTS" are redundant. You could use the same schema for GD because it too has exceptions. What I wrote concedes to you that the correlation is better for GD than for goals scored or goals conceded (though if you think you've shown that, you're wrong). So the question I ended up with was (to simplify) if you have two indicators of success, one 0.91 reliable saying we'll be third (if etc...) and one 0.65 reliable saying we'll be eighth (if etc...), do you simply throw out the second or do you use it to moderate your confidence in the first indicator? It's not a rhetorical question, and thus not a conclusion of anything let alone errors of logic.

I think you meant by your last sentence that the indicators are not independent, and that the unreliability of the goals scored indicator is strongly tied to the fact that it doesn't take into account something relevant that GD does, viz goals conceded. A fair enough point, if so, but note that nothing I wrote restricts my remarks to goals scored or conceded as alternatives (I also mentioned points accrued, which is a pretty damn good indicator and there may be others like time of possession). So, no, no errors of logic. I won't bore you with my credentials there.

[From Despair To Where?]

I'm convinced Snowball was the editor of Pravda in a previous life.

Just substitute tractor production figures for Ian Hartes' pace and you have Comrade Snowball's glorious Five Game Plan. The figures cannot be questioned. It's 5 years in the Statistical Gulag for you, Comrade.

[Snowball]

I'm convinced Snowball was the editor of Pravda in a previous life.

Just substitute tractor production figures for Ian Hartes' pace and you have Comrade Snowball's glorious Five Game Plan. The figures cannot be questioned. It's 5 years in the Statistical Gulag for you, Comrade.

Ian Harte is stunningly slow. The only slower player I have ever known was John Relish of Newport County and he managed 400 league appearances as fullback.

But Harte's record, the defence's record is extremely good, and it's THAT I base my support on.

22 Goals conceded in 23 games since Harte arrived
21 Goals conceded in 22 games in which Harte started

0.95 Goals per Game

1 26 17 0.65 Queens Park Rangers
2 21 24 0.92 Millwall
2 24 22 0.92 Nottingham Forest
4 27 25 0.93 Swansea City
3 22 21 0.95 Reading, Harte Starting
5 26 27 1.04 Reading, All Games

However Reading's defensive record is, IMO better than this. We had one aberrant 4-game spell where there was a goal-glut. 4-3, 1-1, 3-3, 1-3

That means in the other 22 games we have let in 17 goals. .77 per game (including before Harte arrived)

That means in the other 17 games we have let in 11 goals. .64 goals per game (Harte's starts excluding the 4 game goal-fest)

I don't UNDERSTAND it. I keep expecting wingers to have a field day but the facts are the facts are the facts.
22 Games Harte has played. NINE clean sheets, Nine games conceding 1 goal.

That is a good defence and Harte must be worth his place. Now throw in 4 goals. It's a no-brainer. He is playing on merit.

[Snowball]

As a matter of correctness, on "True tables" (ppg) we are eighth according to Statto.

As for allegations that the team is inconsistent, Statto's rating for RFC
has been between 800 and 818 for the last 39 games, with a slow rise
in form until a recent steeper rise. Currently we have equal top rating
with QPR and show a "rising" rating, meaning we might well do better
than the raw stats would predict.

Currently they reckon it could come down to Goal Difference!

1 79 818 QPR
2 74 812 Swansea
3 74 814 Cardiff
4 73 811 Norwich
5 72 808 Watford
6 71 818 Reading
7 71 801 Leeds


Of course, their ratings will not yet take into account signings in the window

[weybridgewanderer]

Of course, their ratings will not yet take into account signings in the window

Of course, neither does your goal difference correlation.

[RobRoyal]

As a matter of correctness, on "True tables" (ppg) we are eighth according to Statto.

As for allegations that the team is inconsistent, Statto's rating for RFC
has been between 800 and 818 for the last 39 games, with a slow rise
in form until a recent steeper rise. Currently we have equal top rating
with QPR and show a "rising" rating, meaning we might well do better
than the raw stats would predict.

Currently they reckon it could come down to Goal Difference!

1 79 818 QPR
2 74 812 Swansea
3 74 814 Cardiff
4 73 811 Norwich
5 72 808 Watford
6 71 818 Reading
7 71 801 Leeds


Of course, their ratings will not yet take into account signings in the window

No Forest?

They're looking a good bet for 2nd IMO.

[Agent Balti]

Of course, their ratings will not yet take into account signings in the window

Of course, neither does your goal difference correlation.

Or what any other club does...and that's what always irks me which all this stat heaven, it ignores the variables of what the other clubs do. I know Snowball will now say that's moot, but it obviously isn't as our stats are driven by our results against them. But it's the others versus the others that equally matter...hence why we have a league table.

[Snowball]

Of course, their ratings will not yet take into account signings in the window

Of course, neither does your goal difference correlation.

Which I mentioned.

Which is why I have continually said, "IF we maintain our relative GD position" with emphasis on the IF