• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

What are the best football ratings ?

cosmicsports

cosmicsports

Colt
There are many rating systems for the football teams.
The most commonly known are I think the elo ratings which you can find here:

For clubs:

Football Club Elo Ratings

www.clubelo.com www.clubelo.com

For national teams:


But what are the best ratings ?
I understand this is maybe a difficult question because there are no indices of betterness published.
Such indices exist if one performs the task to compute information sums, by comparing ratings and results - he must however have in his possession all of the match histories to do that.

So I 'm asking what do you think from experience,

For the elos the criticism against them is they change very slowly.
For example in the premier they have Arsenal now at 1988, but Forest who are following close by in the table at 1771 and Man. City who are visibly struggling at 1945.
If Arsenal play Forest tomorrow at home, with the 217 points difference as the elos suggest, the match should be a dead easy one for Arsenal, but we know it is n't.
 
Last edited:
On the club.elo website, looking at tonights Chelsea v West Ham game, the column that interests me the most, with regards recent form, is the +/- column. Chelsea were flying along, winning domestic matches and their games in europe too, then, for whatever reason, they dropped about 40 elo points in 5 games. It's possible. it's the market that is slower to react, which then presents value opportunities.

The main Elo is not bad as a rating, it does adjust fairly quickly if a good team gets beaten by the poor teams consistently.

The odds calculations seems to be referred to in the pre-off markets by a lot of the market makers, especially on Betfair. (and maybe Pinnacle too). It's already predicted the odds for the next league game on the 14th Feb! (before any last minute re-adjustments)

Chelsea

1738580990503.png

West Ham

1738581603578.png
 
DuckandDive said:
On the club.elo website, looking at tonights Chelsea v West Ham game, the column that interests me the most, with regards recent form, is the +/- column. Chelsea were flying along, winning domestic matches and their games in europe too, then, for whatever reason, they dropped about 40 elo points in 5 games. It's possible. it's the market that is slower to react, which then presents value opportunities.

The main Elo is not bad as a rating, it does adjust fairly quickly if a good team gets beaten by the poor teams consistently.

The odds calculations seems to be referred to in the pre-off markets by a lot of the market makers, especially on Betfair. (and maybe Pinnacle too). It's already predicted the odds for the next league game on the 14th Feb! (before any last minute re-adjustments)

Chelsea

View attachment 153788

West Ham

View attachment 153789
Click to expand...

Yeah, but look what it says here:

medium.com

Implementing the ELO ratings for soccer teams in python.

In this article I will present pythonic implementation of several variations of ELO ratings and examine their performance in a simple…
medium.com medium.com

He says "the elo ratings are quite inferior to those of the bookies".
While of course the bookies consider more factors than ratings and hence go a bit better, it is a negative opinion from this mr. "ML soccer betting" fellow.

As for Nottingham Forest the elos have them on a par with beleagured Spurs.
I don't know at this moment what to say - I 'm thinking.
I remember back in 2014 FIFA published a rating chart for the world cup teams of that year that differed from elo.
So I wrote an email to them asking about the value of their ratings and they replied to me "we don't care about value or comparisons !".

Regarding Chelsea against West Ham today I already opposed them.
 
Last edited:
Some maths going in that article!

I reckon it's possible to create our own Elo (with the help of A.I). It's possible, by fine tuning certain variables, there is a data set that can be used, that maybe, others have yet to discover.

DeepSense kicks off with some starting points.

-------


Creating your own Elo ratings for football (soccer) involves using historical match data to calculate team ratings based on their performance. The Elo system adjusts ratings after each match, considering the expected outcome versus the actual result. Here’s the best data to use and how to approach it:

Key Data to Use​

  1. Match Results (Most Important):
    • Date of the match.
    • Home team and away team.
    • Final score (goals scored by each team).
    • Whether the match was played at a neutral venue (if applicable).
  2. Home/Away Advantage:
    • Home teams typically perform better, so you may want to incorporate a home-field advantage factor (e.g., +100 Elo points for the home team).
  3. Match Importance:
    • Include the context of the match (e.g., friendly, league, cup, or international tournament). More important matches (e.g., World Cup) could have a higher weight in the rating system.
  4. Goal Difference:
    • Some Elo systems incorporate goal difference to adjust the magnitude of rating changes. For example, a 3-0 win might result in a larger rating change than a 1-0 win.
  5. Team Strength Over Time:
    • Use data over multiple seasons to capture team strength trends. Teams improve or decline over time, and your Elo system should reflect this.
  6. International vs. Club Football:
    • Decide whether you want to rate clubs, national teams, or both. If both, you may need separate Elo systems or a way to integrate them.

Steps to Create Your Elo Ratings​

  1. Assign Initial Ratings:
    • Start with an initial Elo rating for each team (e.g., 1500 for all teams or use historical data to set initial values).
  2. Calculate Expected Outcome:
    • Use the Elo formula to calculate the expected score for each team:
      EA=11+10(RB−RA)/400EA=1+10(RB−RA)/4001EB=11+10(RA−RB)/400EB=1+10(RA−RB)/4001
      Where:
      • EAEA = expected score for Team A.
      • RARA = Elo rating of Team A.
      • RBRB = Elo rating of Team B.
  3. Update Ratings After Each Match:
    • Adjust the ratings based on the actual result:
      RA′=RA+K×(SA−EA)RA′=RA+K×(SA−EA)RB′=RB+K×(SB−EB)RB′=RB+K×(SB−EB)
      Where:
      • RA′RA′ = new rating for Team A.
      • SASA = actual result (1 for win, 0.5 for draw, 0 for loss).
      • KK = weight factor (typically between 20 and 40, depending on the importance of the match).
  4. Incorporate Home Advantage:
    • Add a fixed Elo boost (e.g., +100) to the home team’s rating before calculating the expected outcome.
  5. Adjust for Goal Difference (Optional):
    • Modify the KK factor based on the margin of victory. For example:
      Kadjusted=K×ln⁡(∣GD∣+1)Kadjusted=K×ln(∣GD∣+1)
      Where GDGD is the goal difference.
  6. Iterate Through Historical Data:
    • Process all matches in chronological order, updating ratings after each match.

Where to Get Data​

  • Free Sources:
  • Paid Sources:
    • Opta: Detailed match and player statistics.
    • StatsBomb: Advanced football data.

Tips for Success​

  • Start with a simple model and refine it over time.
  • Experiment with different KK values and home advantage adjustments.
  • Validate your ratings by comparing them to established systems like FIFA rankings or club Elo ratings (e.g., ClubElo.com).
  • Consider adding weights for recent performance to account for team form.
By using this data and approach, you can create a robust Elo rating system tailored to your needs.






New chat
 
DuckandDive said:
Some maths going in that article!

I reckon it's possible to create our own Elo (with the help of A.I). It's possible, by fine tuning certain variables, there is a data set that can be used, that maybe, others have yet to discover.

DeepSense kicks off with some starting points.

-------


Creating your own Elo ratings for football (soccer) involves using historical match data to calculate team ratings based on their performance. The Elo system adjusts ratings after each match, considering the expected outcome versus the actual result. Here’s the best data to use and how to approach it:

Key Data to Use​

  1. Match Results (Most Important):
    • Date of the match.
    • Home team and away team.
    • Final score (goals scored by each team).
    • Whether the match was played at a neutral venue (if applicable).
  2. Home/Away Advantage:
    • Home teams typically perform better, so you may want to incorporate a home-field advantage factor (e.g., +100 Elo points for the home team).
  3. Match Importance:
    • Include the context of the match (e.g., friendly, league, cup, or international tournament). More important matches (e.g., World Cup) could have a higher weight in the rating system.
  4. Goal Difference:
    • Some Elo systems incorporate goal difference to adjust the magnitude of rating changes. For example, a 3-0 win might result in a larger rating change than a 1-0 win.
  5. Team Strength Over Time:
    • Use data over multiple seasons to capture team strength trends. Teams improve or decline over time, and your Elo system should reflect this.
  6. International vs. Club Football:
    • Decide whether you want to rate clubs, national teams, or both. If both, you may need separate Elo systems or a way to integrate them.

Steps to Create Your Elo Ratings​

  1. Assign Initial Ratings:
    • Start with an initial Elo rating for each team (e.g., 1500 for all teams or use historical data to set initial values).
  2. Calculate Expected Outcome:
    • Use the Elo formula to calculate the expected score for each team:
      EA=11+10(RB−RA)/400EA=1+10(RB−RA)/4001EB=11+10(RA−RB)/400EB=1+10(RA−RB)/4001
      Where:
      • EAEA = expected score for Team A.
      • RARA = Elo rating of Team A.
      • RBRB = Elo rating of Team B.
  3. Update Ratings After Each Match:
    • Adjust the ratings based on the actual result:
      RA′=RA+K×(SA−EA)RA′=RA+K×(SA−EA)RB′=RB+K×(SB−EB)RB′=RB+K×(SB−EB)
      Where:
      • RA′RA′ = new rating for Team A.
      • SASA = actual result (1 for win, 0.5 for draw, 0 for loss).
      • KK = weight factor (typically between 20 and 40, depending on the importance of the match).
  4. Incorporate Home Advantage:
    • Add a fixed Elo boost (e.g., +100) to the home team’s rating before calculating the expected outcome.
  5. Adjust for Goal Difference (Optional):
    • Modify the KK factor based on the margin of victory. For example:
      Kadjusted=K×ln⁡(∣GD∣+1)Kadjusted=K×ln(∣GD∣+1)
      Where GDGD is the goal difference.
  6. Iterate Through Historical Data:
    • Process all matches in chronological order, updating ratings after each match.

Where to Get Data​

  • Free Sources:
  • Paid Sources:
    • Opta: Detailed match and player statistics.
    • StatsBomb: Advanced football data.

Tips for Success​

  • Start with a simple model and refine it over time.
  • Experiment with different KK values and home advantage adjustments.
  • Validate your ratings by comparing them to established systems like FIFA rankings or club Elo ratings (e.g., ClubElo.com).
  • Consider adding weights for recent performance to account for team form.
By using this data and approach, you can create a robust Elo rating system tailored to your needs.






New chat
Click to expand...

Good info that.
As it happens I did try to produce my own "elos" once, that is use the same logic but depart from the standard formula.
Got stuck because it needed lots of attention and daily updates and I 'm not a team of journalists !
It's still not enough of though. You got to find where the market has got it wrong. Where the market has got it right it's just a useless endeavour.
 
I think ELO's are over used and lack value, nowadays.

To be honest, football is too dynamic to slap a rating on a team and hope it sticks. They don't factor in who's playing on the day and who isn't, whether the team has lost confidence in the manager, etc, or if they are a tired team after too many games in close succession.

A better approach would be to formulate an algorithm that adjusts to the live game dynamics. It could include SOT, corner count, team momentum and so on.

If you get to know certain teams, you will be able to spot if they are going to be scoring today or not, judging from how they play within the first 5-10mins of the game. My team, Leeds Utd, is a prime example.

Regards
 
cosmicsports said:
Good info that.
As it happens I did try to produce my own "elos" once, that is use the same logic but depart from the standard formula.
Got stuck because it needed lots of attention and daily updates and I 'm not a team of journalists !
It's still not enough of though. You got to find where the market has got it wrong. Where the market has got it right it's just a useless endeavour.
Click to expand...

Agree it's a lot of fussing over something that may have imperfections, but the market is also made up of imperfections too.

WIth A.I. in the mix, if I were you, dig out your old research notes and re-apply it. GPT or DeepSense mighthelp you discover another way in.

The bit on the end covers various ways in.

Tips for Success​

  • Start with a simple model and refine it over time.
  • Experiment with different KK values and home advantage adjustments.
  • Validate your ratings by comparing them to established systems like FIFA rankings or club Elo ratings (e.g., ClubElo.com).
  • Consider adding weights for recent performance to account for team form.
 
Leodis said:
I think ELO's are over used and lack value, nowadays.

To be honest, football is too dynamic to slap a rating on a team and hope it sticks. They don't factor in who's playing on the day and who isn't, whether the team has lost confidence in the manager, etc, or if they are a tired team after too many games in close succession.

A better approach would be to formulate an algorithm that adjusts to the live game dynamics. It could include SOT, corner count, team momentum and so on.

If you get to know certain teams, you will be able to spot if they are going to be scoring today or not, judging from how they play within the first 5-10mins of the game. My team, Leeds Utd, is a prime example.

Regards
Click to expand...

Leeds are my team as well!

Elo would only be one cog in the wheel for sure. A combination of pre-off preparation/expectation, with a way of monitoring the game-flow in-play, would be ideal. If your model predicts goals/no goals and the attack/momentum stats supports the pre-off set up, then maybe at half-time, a market could be chosen that's out of sync and has some value.

The flip side from Leeds are Burnley who love a 0-0 score-line, so your model would adjust accordingly. (conceded 9 goals in 30 games).
 
DuckandDive said:
Leeds are my team as well!

Elo would only be one cog in the wheel for sure. A combination of pre-off preparation/expectation, with a way of monitoring the game-flow in-play, would be ideal. If your model predicts goals/no goals and the attack/momentum stats supports the pre-off set up, then maybe at half-time, a market could be chosen that's out of sync and has some value.

The flip side from Leeds are Burnley who love a 0-0 score-line, so your model would adjust accordingly. (conceded 9 goals in 30 games).
Click to expand...
It should be simple just identify a team that is historically strong like my team Manchester United and when you see that every player they buy is a mercenary coming for 5+ times their actual value and are all completely unworthy to wear the shirt , and even if they were remotely the standard needed they don’t give a shit anyway and are only there for the £300k a week and the onlyfans girls. Then just bet against them every time they play because they are completely shit and even the bookies can’t catch up with just how shit they are.
 
I have read contributions from someone on another forum and while it's true he consistently makes a profit but for me the approach is extremely boring, for instance he will identify a match in some lower league game in a country like norway or chile, then he points to a fact that over the last 30 games team A win a average of 10 corners per game and team B concede an average of 10 corners per game, ( i'm exaggerating the numbers somewhat ) but it gives you an idea of how his mind works and just to remind he does make it pay but i fail to see the fun in that type of approach, my simple approach recently is to back 365 super boosts.

ps...I should just admit that i'm from sheffield so make of that what you will.
 
Outlander said:
It should be simple just identify a team that is historically strong like my team Manchester United and when you see that every player they buy is a mercenary coming for 5+ times their actual value and are all completely unworthy to wear the shirt , and even if they were remotely the standard needed they don’t give a shit anyway and are only there for the £300k a week and the onlyfans girls. Then just bet against them every time they play because they are completely shit and even the bookies can’t catch up with just how shit they are.
Click to expand...

:hi:

You could have an angle there.

Each player has a transfer market value, by using the transfer market website, when they devalue, and the manager keeps selecting them to play, their rating will drop. Add all the players together for each game, you then have a whole rating per game. It's sounds like the "moneyball" approach but a lot of work, could be fun though.
 
tacker said:
I have read contributions from someone on another forum and while it's true he consistently makes a profit but for me the approach is extremely boring, for instance he will identify a match in some lower league game in a country like norway or chile, then he points to a fact that over the last 30 games team A win a average of 10 corners per game and team B concede an average of 10 corners per game, ( i'm exaggerating the numbers somewhat ) but it gives you an idea of how his mind works and just to remind he does make it pay but i fail to see the fun in that type of approach, my simple approach recently is to back 365 super boosts.

ps...I should just admit that i'm from sheffield so make of that what you will.
Click to expand...

It would be hard for a bookies to keep tabs on all games around the globe, so their approach by honing in on certain markets and exploiting the data, could keep them ahead of the game. One issue, might be, not being able to get your bet on a game played in the outer hebrides, especially if you consistently show a profit.

Don't get me wrong, there are some syndicates that are well ahead of the game with analysing and watching games..etc. I reckon it's still worth attempting something by looking at combinations of things, and with A.I. in the mix, the gap could be closed very quickly.

Nothing wrong with Sheffield.
 
DuckandDive said:
It would be hard for a bookies to keep tabs on all games around the globe, so their approach by honing in on certain markets and exploiting the data, could keep them ahead of the game. One issue, might be, not being able to get your bet on a game played in the outer hebrides, especially if you consistently show a profit.

Don't get me wrong, there are some syndicates that are well ahead of the game with analysing and watching games..etc. I reckon it's still worth attempting something by looking at combinations of things, and with A.I. in the mix, the gap could be closed very quickly.

Nothing wrong with Sheffield.
Click to expand...
We had a 15 yr old stabbed to death yesterday, homelessness/ drugs everywhere and when you walk around the city centre it's very difficult to believe we're still in england, supermarkets are all going self-service in more ways than one but offset by the joys that SUFC & SWFC bring.
 
I was talking about Forest before.
Last year they narrowly escaped relegation. Well, maybe not so narowlly but they did finish 17th.
So what are they now, to the ratings and to the bookies ? A new top side or an autumn firework ?
In the past there are examples of both.
So the elos are not terribly wrong but they need some refinement.
After all there a number of constant coefficients in the standard formula and I wonder if those have been researched and when.

Here is an ancient method of mine:
Divide the league teams into five groups, A-B-C-D-E.

A = Title contenders
B = Europe place contenders
C = High side of middle table
D = Low side of middle table
E = relegation zone

Then worked out statistics for those groups and in the very early season my estimations were naturally somewhat arbitrary.
It was n't bad but we 're talking about the football of the eighties (when backpasses were allowed !).
This stinks a little when the opponents are from different countries. What exactly would AC Milan do if they were playing in the premier, or Chelsea if they were playing in the campionato ?

Ancient football was different from modern football and the advantage of home venue used to be a big one.
I remember this vividly. AEK Athens beat Derby County 2-0 in Athens for the UEFA cup.
As I was in London I was thinking of taking the train north to watch the replay.
Then I read in the papers "the shadow of a heavy defeat looms large over AEK Athens after the insufficient 2-0".
Today you d' call them crazy.
As it happened the Greek team won in the Baseball ground as well by 3-2, but those defeatist journalists were guided by the prevalent statistics of those years. So could I blame them ?
Today things are different. A 1-0 home win would be reckonned as decisive advantage.
 
Last edited:
cosmicsports said:
I was talking about Forest before.
Click to expand...

Threads sometimes go off-piste, that's a good thing though.

cosmicsports said:
Here is an ancient method of mine:
Divide the league teams into five groups, A-B-C-D-E.

A = Title contenders
B = Europe place contenders
C = High side of middle table
D = Low side of middle table
E = relegation zon
Click to expand...

That's look okay. Some of the leagues tend to split into mini-leagues, within the same league, which helps spot which teams are motivated to get a result in the later stages of the season.

cosmicsports said:
This stinks a little when the opponents are from different countries. What exactly would AC Milan do if they were playing in the premier, or Chelsea if they were playing in the campionato ?
Click to expand...

That's where the club.elo website comes in handy. I often wondered how Celtic/Rangers would fare if the scottish premier was amalgamated with the English leagues. When looking at their respective Elo's, Celtic/Rangers compare with the likes of Wolves and West Ham, whilst the next teams from 3rd onwards, compare to Derby/Plymouth of the Championship.

I tend to like Elo for that measurement, but I reckon the nitty gritty stuff is about predicting expected goals, and then run them through a Poisson Distribution washing machine to compare the odds and other markets that are impacted by that data.
 
tacker said:
We had a 15 yr old stabbed to death yesterday, homelessness/ drugs everywhere and when you walk around the city centre it's very difficult to believe we're still in england, supermarkets are all going self-service in more ways than one but offset by the joys that SUFC & SWFC bring.
Click to expand...

I like the big derby match ups, which often produce surprise results.

The UK overall has got a lot of on-going issues to deal with. You hear a lot kicking off in London, not just gangs, but random stuff that innocent people get caught up in.
 
DuckandDive said:
Threads sometimes go off-piste, that's a good thing though.



That's look okay. Some of the leagues tend to split into mini-leagues, within the same league, which helps spot which teams are motivated to get a result in the later stages of the season.



That's where the club.elo website comes in handy. I often wondered how Celtic/Rangers would fare if the scottish premier was amalgamated with the English leagues. When looking at their respective Elo's, Celtic/Rangers compare with the likes of Wolves and West Ham, whilst the next teams from 3rd onwards, compare to Derby/Plymouth of the Championship.

I tend to like Elo for that measurement, but I reckon the nitty gritty stuff is about predicting expected goals, and then run them through a Poisson Distribution washing machine to compare the odds and other markets that are impacted by that data.
Click to expand...

Let it be understood that we are subject to the following restrictions:

1) We can't really do things that require a team of dedicated personel to manage (we are not CEA-CERN !).
2) We can't allow the bookies to sniff what we are doing.

But now I think I will try to revive the A-B-C-D-E method, as the basis.
 
cosmicsports said:
1) We can't really do things that require a team of dedicated personel to manage (we are not CEA-CERN !).
2) We can't allow the bookies to sniff what we are doing.
Click to expand...

1) That can sometimes put someone off trying to find some sort of niche within the football markets. Trying to replicate how Starlizard in Camden do things, ain't gonna happen, however, there's plenty of different approaches we can set up within our own models.

2) In this day and age, the bookies are a law unto themselves. Until the rules change, ie: They have to accept a bet to take out £xxx.xx, (Australian Model), then restrictions will apply to the more successful punters. They won't care what's written in the forums as they have their own algorithms and customer management programmes to lean on. The Betfair exchange is the way to go for now, although liquidity is lower than it was 5+ years ago.

I still reckon the market makers on Betfair follow club.elo and Pinnacle to gauge the market, if so, it can be exploited under certain conditions.
 
DuckandDive said:
1) That can sometimes put someone off trying to find some sort of niche within the football markets. Trying to replicate how Starlizard in Camden do things, ain't gonna happen, however, there's plenty of different approaches we can set up within our own models.

2) In this day and age, the bookies are a law unto themselves. Until the rules change, ie: They have to accept a bet to take out £xxx.xx, (Australian Model), then restrictions will apply to the more successful punters. They won't care what's written in the forums as they have their own algorithms and customer management programmes to lean on. The Betfair exchange is the way to go for now, although liquidity is lower than it was 5+ years ago.

I still reckon the market makers on Betfair follow club.elo and Pinnacle to gauge the market, if so, it can be exploited under certain conditions.
Click to expand...

The market situation is poor.
Sighs !
The politicians are a compendium of third reich and North Korea, when it comes to it.
Maybe your government, my government are the best since the golden age of Pericles in all other things but re. the betting market situation they are a compendium of third reich and North Korea.
This is pretty fixed now.
In Greece they shut down our race course, through ultra-taxation,
But let us stay in our fantasy world - it does n't hurt.
 
You must log in or register to reply here.
Back
Top