The European football thread - Page 369

theasfl · 11-06-2006, 12:00 AM

Quote:

Originally Posted by Sdpm100

Dola, if you want to do it throughout the season, then ask and I can post the Excel worksheet that has the formulas already in. It will be a lot easier and less labour intensive than by hand.

I already have way too many excel files open and a lack of time to do it myself, but Id certainly be glad if you kept up with it

bababui · 11-06-2006, 12:26 AM

Quote:

Originally Posted by GUTIERREZ

I'm suprised they found a spot for an englishman. Arsenal should relocate to France.

And Liverpool to Spain!

Raderick · 11-06-2006, 12:56 AM

Quote:

Originally Posted by bababui

And Liverpool to Spain!

And Manchester United to Tampa!

Sdpm100 · 11-06-2006, 03:03 AM

Quote:

Originally Posted by theasfl

I already have way too many excel files open and a lack of time to do it myself, but Id certainly be glad if you kept up with it

So you didn't want the formula either then?

bababui · 11-06-2006, 03:11 AM

Quote:

Originally Posted by Sdpm100

So you didn't want the formula either then?

id do it but I dont know how to operate Excel..sad as that is.

Sdpm100 · 11-06-2006, 03:19 AM

Quote:

Originally Posted by bababui

id do it but I dont know how to operate Excel..sad as that is.

I really don't mind if nobody wants it, it's probably a load of junk anyway.

I just thought he asked for the formula, but maybe he didn't want it after all. So I got it wrong yet again.

bababui · 11-06-2006, 03:21 AM

Quote:

Originally Posted by Sdpm100

I really don't mind if nobody wants it, it's probably a load of junk anyway.

I just thought he asked for the formula, but maybe he didn't want it after all. So I got it wrong yet again.

Could you teach me, please?

Sdpm100 · 11-06-2006, 03:23 AM

Quote:

Originally Posted by bababui

Could you teach me, please?

The formulas and methods are in post 7358.

bababui · 11-06-2006, 03:28 AM

Quote:

1. Hockey games end in ties. A team could never win and never lose a game (All ties) and end up as a .500 team (0-0-82). This doesn't skew the overall numbers much (since such a team gets as many points as a team tht goes 41-41-0), but certain teams can be very good at tying up games while others either win/lose...which leads to...

Doesnt the fact that soccer gives 3 points for a win scrw the results?

Sdpm100 · 11-06-2006, 03:45 AM

I'll go through an example by using Manchester United (just to please Raderick

).

Here's their line:

Code:

			P	F	A	Pts	Pct.	Pyt. %	Exp. Pts.
Manchester United	11	26	5	28	0.848	0.943	28

For pythagorean record, like in baseball, you only need the goals each team has scored and conceded.
So Manchester United have scored 26 and conceded just 5.

The first percentage (the 0.848) is a points percentage. So to get that you divide the total number of points by the (number of games * maximum points per game).

So you have 28/(11*3) = 28/33 = 0.848 (to 3 d.p.)

Then you work out the pythagorean points percentage which uses the well-known from baseball equation.

Code:

         Goals Scored²
-------------------------------
Goals Scored² + Goals Conceded²

I used 1.7 as the exponent instead by looking at results from this season so far (this is the average winning margin). All it does is it keeps the range of the expected percentages closer together than using 2. Conversely if you used an even smaller exponent the range would get smaller than I have.
I think 1.7 works well.

Anyway, using Manchester United's data, we have -

Code:

    26^1.7         254.363
---------------  = --------  = 0.943 % (to 3 d.p.) 
26^1.7 + 5^1.7     269.789

Finally we then look at the expected points.
Now there are two ways to do this.
The first is the quick way, but it isn't quite as accurate.
You use the maximum number of points per game, which is 3.

But because there aren't 3 points up for grabs in every game, it will inflate the results. But if you're not using Excel then this is the quick way, because you have to work out the totals for the whole league in points and games.

So the equation for either method is -
pythagorean % * maximum (or average) points per game * number of games team has played

So for Manchester United
E(pts) = 0.943 * 3 * 11 = 31 points (31.1 rounded to the nearest whole number, for obvious reasons)

Now this is different to my total, because I've worked out the average points per game for the league is 2.745.
So the more accurate method is -
E(pts) = 0.943 * 2.745 * 11 = 28 points (28.47 rounded to the nearest whole number)

Hope that helps.

Sdpm100 · 11-06-2006, 03:50 AM

Quote:

Originally Posted by bababui

Doesnt the fact that soccer gives 3 points for a win scrw the results?

Yes, it will inflate the results (or in the NHL deflate the results), because less than 3 points are awarded in every game.

That's why, like that guy did, I worked out the average points per game for the whole league. I didn't even think of that idea until I saw his work, so it was good to find.

bababui · 11-06-2006, 03:55 AM

SDPM..I dont quite understand what that formula is telling us/predicting?

Sdpm100 · 11-06-2006, 03:56 AM

Quote:

Originally Posted by bababui

SDPM..I dont quite understand what that formula is telling us/predicting?

It's probably a load of rubbish that's why. Just forget I wasted everyone's time by posting it. Sorry.

bababui · 11-06-2006, 03:59 AM

Quote:

Originally Posted by Sdpm100

It's probably a load of rubbish that's why. Just forget I wasted everyone's time by posting it. Sorry.

Thats not what I mean!! What is it measuring/predicting?? Im just trying to get a better understanding of it.

Sdpm100 · 11-06-2006, 04:03 AM

Quote:

Originally Posted by bababui

Thats not what I mean!! What is it measuring/predicting?? Im just trying to get a better understanding of it.

It's only used to look at whether teams are over- or under-achieving based solely on what they score and concede. Just like the pythagorean records in baseball.
You could also ignore it as a piece of junk.

I mentioned it over the page, but we could suggest from the full results that Arsenal and Everton are under-achieving based on their goal records (-4 real points to expected points), whilst Manchester City and Sheffield United are over-achieving based on for and against (+5 real points to expected points).

Sdpm100 · 11-06-2006, 04:05 AM

I guess you could also use at as a projection of points. If they continue on the same pythagorean percentage, how many points will they accrue by the end of the season?

EDIT - I've just checked for Manchester United, and they should pick up another 70 points based on pythagorean %, to give them a total of 98.

bababui · 11-06-2006, 04:06 AM

Quote:

Originally Posted by Sdpm100

It's only used to look at whether teams are over- or under-achieving based solely on what they score and concede. Just like the pythagorean records in baseball.
You could also ignore it as a piece of junk.

I mentioned it over the page, but we could suggest from the full results that Arsenal and Everton are under-achieving based on their goal records (-4 real points to expected points), whilst Manchester City and Sheffield United are over-achieving based on for and against (+5 real points to expected points).

Good deal.

Would the formula be more effective in its predictions if we waited till the halfway point in the season and all teams have played one another?

Sdpm100 · 11-06-2006, 04:11 AM

Quote:

Originally Posted by bababui

Good deal.

Would the formula be more effective in its predictions if we waited till the halfway point in the season and all teams have played one another?

Yes it would, but mainly because it would be more likely that the average of goals by each team (both for and against) would be more settled, and wouldn't be so variable based on a big win or loss.

What I mean is if you win 4-0 and you've only scored 8 goals all season, that is 50% added on in one game.
Where if you've scored 52 goals, then adding 4 to that is an average goals scored per game increase of 7.7%.

bababui · 11-06-2006, 04:14 AM

Quote:

Originally Posted by Sdpm100

Yes it would, but mainly because it would be more likely that the average of goals by each team (both for and against) would be more settled, and wouldn't be so variable based on a big win or loss.

What I mean is if you win 4-0 and you've only scored 8 goals all season, that is 50% added on in one game.
Where if you've scored 52 goals, then adding 4 to that is an average goals scored per game increase of 7.7%.

Are you working on any original statistical work? It seems like you have found your calling.

Sdpm100 · 11-06-2006, 04:17 AM

Quote:

Originally Posted by bababui

Are you working on any original statistical work? It seems like you have found your calling.

No I'm not.
And this isn't that original either is it? It borrows from everyone else.

11-06-2006, 03:45 AM	#7370
Sdpm100 Hall Of Famer Join Date: May 2005 Location: Cheltenham, England Posts: 7,644	I'll go through an example by using Manchester United (just to please Raderick ). Here's their line: Code: P F A Pts Pct. Pyt. % Exp. Pts. Manchester United 11 26 5 28 0.848 0.943 28 For pythagorean record, like in baseball, you only need the goals each team has scored and conceded. So Manchester United have scored 26 and conceded just 5. The first percentage (the 0.848) is a points percentage. So to get that you divide the total number of points by the (number of games * maximum points per game). So you have 28/(113) = 28/33 = 0.848 (to 3 d.p.) Then you work out the pythagorean points percentage which uses the well-known from baseball equation. Code: Goals Scored² ------------------------------- Goals Scored² + Goals Conceded² I used 1.7 as the exponent instead by looking at results from this season so far (this is the average winning margin). All it does is it keeps the range of the expected percentages closer together than using 2. Conversely if you used an even smaller exponent the range would get smaller than I have. I think 1.7 works well. Anyway, using Manchester United's data, we have - Code: 26^1.7 254.363 --------------- = -------- = 0.943 % (to 3 d.p.) 26^1.7 + 5^1.7 269.789 Finally we then look at the expected points. Now there are two ways to do this. The first is the quick way, but it isn't quite as accurate. You use the maximum number of points per game, which is 3. But because there aren't 3 points up for grabs in every game, it will inflate the results. But if you're not using Excel then this is the quick way, because you have to work out the totals for the whole league in points and games. So the equation for either method is - pythagorean % maximum (or average) points per game * number of games team has played So for Manchester United E(pts) = 0.943 * 3 * 11 = 31 points (31.1 rounded to the nearest whole number, for obvious reasons) Now this is different to my total, because I've worked out the average points per game for the league is 2.745. So the more accurate method is - E(pts) = 0.943 * 2.745 * 11 = 28 points (28.47 rounded to the nearest whole number) Hope that helps. Last edited by Sdpm100; 11-06-2006 at 03:55 AM.

11-06-2006, 04:05 AM	#7376
Sdpm100 Hall Of Famer Join Date: May 2005 Location: Cheltenham, England Posts: 7,644	I guess you could also use at as a projection of points. If they continue on the same pythagorean percentage, how many points will they accrue by the end of the season? EDIT - I've just checked for Manchester United, and they should pick up another 70 points based on pythagorean %, to give them a total of 98. Last edited by Sdpm100; 11-06-2006 at 04:08 AM.