Power Ratings

[Benny]

I hear a lot about the use of power ratings in the NFL. How could one apply the use of power ratings to the AFL?

Any suggestions?

Benny

[thebookie]

What are Power Ratings, Benny?

[Mr J]

Power ratings are assigning a number to a team, the subracting the number of the opposing team to get the spread. You add adjust the number each week depending on home advantage, injuries, form etc.

[sportznut]

Yep, me and Moeee do this sort of thing already. I'm sure others do too. These were my last set of ratings from before each of the Grand Finals this year:

AFL
126 Port Adel
126 Brisbane
120 St Kilda
117 Geelong
111 Sydney
108 West Coast
105 Essendon
102 Melbourne
96 Fremantle
96 Kangaroos
93 Carlton
90 Collingwood
90 Adelaide
81 West B'dogs
81 Hawthorn
72 Richmond

NRL
36 Bulldogs
34 Syd Roosters
30 Penrith
28 NQ Cowboys
28 Melbourne
28 St Geo/Ill
24 Brisbane
22 Canberra
22 West Tigers
20 Sharks
20 Newcastle
18 Parramatta
18 Manly
14 NZ Warriors
12 Souths

Super12
42 Brumbies
35 Crusaders
31 Stormers
31 Chiefs
31 Blues
28 Bulls
28 Waratahs
24 Sharks
24 Reds
21 Highlanders
21 Hurricanes
10 Cats

To the teams' ratings I then add a bonus for home ground advantage depending on where the game is being played:
AFL: 0,6,12,or 18 pts
NRL: 0,2,4, or 6 pts
S12: 0,3,7 or 10 pts

There may also be some points added or subtracted depending on injuries or suspensions, or one team having a decided head to head advantage over the other.

The final points difference between the two teams is the predicted margin in the game. Well, that's the way it's supposed to work anyway.


<font size=-1>[ This Message was edited by: sportznut on 2004-11-09 11:26 ]</font>

[moeee]

Gee Sportznut,you make me feel homesick.
C'mon 2005 AFL season.
Sportznut,if I can hurry up and lose all my money on the greyhounds,I will then have time to not only refine my footy ratings,but in fact attempt to rate every player.
I will be using Champion Data Ratings as a major base.
Are you aware of these ratings Sportznut?

[sportznut]

I've heard of them but I don't think I'd ever go to the lengths of rating each player. How do they work?


<font size=-1>[ This Message was edited by: sportznut on 2004-11-09 11:25 ]</font>

[sportsmad]

On 2004-11-09 11:23, sportznut wrote:
I've heard of them but I don't think I'd ever go to the lengths of rating each player. How do they work?



This link might enlighten a few on the subject

http://www.swin.edu.au/maths/main/publications/bond98Teamstats.html




Also this link if allowed to post, although a PDF file


http://www.championdata.com.au/files/rankings.pdf

_________________


<font size=-1>[ This Message was edited by: sportsmad on 2004-11-09 11:35 ]</font>

[Benny]

How do i go about compling the ratings? What information do i use?

Benny

[sportznut]

If you're talking about the sort of ratings that I use, then I'm afraid it's all just personal opinion with a little help from statistics and ladder positions.

[rabbitz]

On 2004-11-09 11:17, moeee wrote:
Gee Sportznut,you make me feel homesick.
C'mon 2005 AFL season.

MOEEE WE ARE HAVING A TIPPING COMP NEXT YEAR
WHAT ABOUT A MULTI COMP
YOU INTERESTED????
Sportznut,if I can hurry up and lose all my money on the greyhounds,I will then have time to not only refine my footy ratings,but in fact attempt to rate every player.
I will be using Champion Data Ratings as a major base.
Are you aware of these ratings Sportznut?

[Mr J]

Yeh power ratings come from your own analysis. They're just something to start with, and you'll probally find most people have very similiar ratings. I think the most important part of using them is knowing how much to adjust them when needed, not the original rating.

[moeee]

Spotznut.The Champion ratings are mathematically calculated ratings players receive for their efforts depending on the disposals during the game.
What might seem insignificant is for example Buckley doesn't play.How do you know how much to deduct from the teams rating.
S'pose you can just guess,but that's why I will use the player ratings.
Not perfectly clear at the minute as to how,but as the greyhounds are eating me alive,I had better get my arse into gear soon,before the fat lady starts singing.

[moeee]

Ask me next year Rabbitz!
And well done Sportsmad.
Can you keep posting please.Especially if that's the sort of material you are aware of!

[Mr ed]

I am looking at a tennis system using sort of the same principle. Taking the world ranking of a player and then dividing that number by the roung no. they went out in in their most recent tournament + the number of matches won against opponent. e.g Joe has a ranking of 60 went out in round three last tounament and has beaten Jim 3 out of 5 times his rating would be 60/3+3 = 10. Jim has a rating of 50 went out first round last tournament and has won 2/5 against joe his rating would be 50/ 1+2 =16.66 the player with the lowest rating would be the selection. Also room for adding extra points to the division side if one player is suited by the court, one player had a long tough match in the previuos round etc. Haven't bet yet but am starting to investigate previous results with positive signs. Any thoughts.

[Mr J]

Ed, you'll have to convert the final ratings into percentages though. Ie, how much more likely is the 10 going to win than the 16.66'??

Rank might be judging skill, last round finished judging form and h2h might judge matchup. Problem with h2h is you have to know why one player has beaten the other. Was it experience vs inexperience? Surface? Style of play? Injury? Home crowd etc.

It might be possible to come up with tennis rankings using rank and recent form, and then make adjustments for h2h, matchup, surface etc. Think it's better just trying to create power ratings than using a simple formula. Try it anyway (on paper or something), see how it goes.

[Benny]

I'm working on an NBL system of power ratings.

Taking the ladder position last rnd and dividing by the wining margin and adding the no of games the team one agianst their opponant.

E.G. Syd Kings ladder position is 2 and the won their last game by 10 pts

Syd Kings 2/10+6 = a power rating of 6.2
Melb Tigers 10/15+4 = a power rating of 4.6

I'm still fiddling around with it.

Benny

[thebookie]

The Tigers should have a negative rating-they are a disgrace!!!

[moeee]

For your information Spotznut,here were my ratings prior to the AFL Grand Final beside yours.

AFL
126 Port Adel 120
126 Brisbane 128
120 St Kilda 108
117 Geelong 106
111 Sydney 97
108 West Coast 99
105 Essendon 94
102 Melbourne 92
96 Fremantle 88
96 Kangaroos 86
93 Carlton 73
90 Collingwood 86
90 Adelaide 90
81 West B'dogs 66
81 Hawthorn 58
72 Richmond 55

You seem to be a lot more generous overall,but particularly to the lower ranked teams.
Do you think you can put the home dogs system to better use next year to maximise your profits.
Or are you still concerned that it will eventually fall apart?

[sportznut]

Moeee,

I find if you have the top teams rated too far above the bottom teams, you won't come up with many upsets. You'll almost always end up with the favourites rated well on top. That's why I tend to rate them fairly conservatively. Usually I start the year with about 6 or 7 goals between top and bottom teams and by season's end that's normally about 9 or 10. I give bonuses of up to 3 goals for home ground advantage and up to 2 goals for head to head performance.

If the Homedog system carries on as it has done this year, it would be FANTASTIC, but chances are it could start to level out a bit. At the moment, I'm betting both the System selections and my selections equally but I might have to look at betting a bit more on the System selections in future.

Anyway Mo, here's how the system went on AFL this year:

34 bets, 24 wins
R1 Kangaroos (+6.5) vs Adelaide WON
R1 W.Bulldogs (+9.5) vs West Coast WON
R2 Adelaide (+11.5) vs Brisbane LOST
R3 Essendon (+7.5) vs West Coast WON
R4 Melbourne (+19.5) vs Port Adelaide WON
R4 West Coast (+5.5) vs Brisbane WON
R4 Adelaide (+5.5) vs St Kilda LOST
R5 Richmond (+10.5) vs Adelaide LOST
R5 Carlton (+12.5) vs West Coast WON
R6 St Kilda (+4.5) vs Brisbane WON
R6 Geelong (+4.5) vs Adelaide WON
R7 Kangaroos (+28.5) vs Brisbane LOST
R8 Hawthorn* (+8.5) vs Fremantle WON
R8 Geelong (+5.5) vs Melbourne WON
R8 Kangaroos (+15.5) vs Port Adelaide WON
R10 Carlton (+46.5) vs St Kilda LOST
R11 Sydney (+30.5) vs St Kilda WON
R12 Essendon (+8.5) vs Brisbane LOST
R13 Geelong (+19.5) vs Brisbane WON
R14 Carlton (+17.5) vs Sydney WON
R14 Adelaide (+7.5) vs Melbourne WON
R15 Hawthorn* (+7.5) vs West Coast WON
R16 Collingwood (+22.5) vs Brisbane LOST
R16 Hawthorn (+34.5) vs Port Adel LOST
R18 Adelaide (+8.5) vs Kangaroos WON
R18 Essendon (+11.5) vs Port Adel WON
R18 Sydney (+21.5) vs Brisbane WON
R19 Richmond (+33.5) vs Geelong WON
R20 Hawthorn (+41.5) vs Brisbane WON
R21 Collingwood (+15.5) vs Port Adel LOST
R21 Adelaide (+16.5) vs Geelong WON
R22 Richmond (+22.5) vs Sydney LOST
SF St Kilda (+7.5) vs Sydney WON
PF Geelong* (+30.5) vs Brisbane WON

By the way, the Homedog system has had 23 wins from 38 bets on the Rugby League, and as for Rugby Union, well how about 17 wins from 18 bets!!! If I thought for a minute that sort of strike rate would continue, I'd stop betting on anything else!!!


<font size=-1>[ This Message was edited by: sportznut on 2004-11-14 09:40 ]</font>

[sportznut]

Make that 24 from 39 for Rugby League after this morning.

[sportznut]

Oh, and those R.Union stats don't include Wales vs S.Africa last week or Ireland vs S.Africa this morning, so I've actually gone 19 from 20 with it this year!!!

<font size=-1>[ This Message was edited by: sportznut on 2004-11-14 09:44 ]</font>

[moeee]

Just a quick glance at your AFL results reveals a possible filter when the homedog is given over 20 points head start.
What I intend doing Sportznut,is allocate the teams rating to each player's Champion Data rating in proportion and then when the teams playing list for the next round comes out,I can tally them up individually and hopefully the couple of points difference this system comes up with could make me a winner more often.
Also I intend going back through lasr years result margins and comparing to my predicted margins and averaging could give me a better idea of how much and when to apply the home ground advantage.
For those interested,Stephen Clarke wrote something to the effect that the Home Advantage is due to 3 factors,not necessarilly equally.
Travelling causes fatigue and unfamiliarity.
The crowd influences players and umpires.
The ground played on is unfamiliar.

[Benny]

I have found a web site that shows me how to calculate power ratings based on points for and against.

Kings have a power rating of 103
Bullets have a power rating of 98

The result should be that the kings should win by 5 pts or more.

Benny

[sportznut]

On 2004-11-14 14:51, moeee wrote:
Just a quick glance at your AFL results reveals a possible filter when the homedog is given over 20 points head start.


Yes, I know about that. I noticed it during the year, and I thought about halving my bet or possibly deleting them altogether, but I decided to stick with them till the end of the season. I'll have a think about it for next year though.

[sportznut]

Chalk another one up for the system on Rugby Union with Scotland staying within 19.5 points of Australia.

[moeee]

Was playing around at the Swinburne website.
It is revealed that from 1997 till 2002,The homedogs at $1 on each,would have got you your money back overall.
So it seems 2004 was one out of the box!
Still,if you continue next year,I wish you all the best.
I tipped 119 winners last year for the home and aways.
I think I can do better than that!