Football Musings

[andy28]

I see the 0-0 were back in force today in MLS

[wearthefoxhat]

I see the 0-0 were back in force today in MLS

Add in the odd 1-6, 1-4 result!

The only 0-0 that caught me out was Nashville SC v DC United. TG's were predicted @ 3.15 (2.55 v 0.60)

Philadelphia away @ 2.95 was wrongly priced by the market.

The overall stats so far, are quite interesting.

Stats.png

[andy28]

End of the season, report will be must try harder but I did learn a lot and that is the main thing. Bring on next season

[firlandsfarm]

End of the season, report will be must try harder but I did learn a lot and that is the main thing. Bring on next season

only 81 days to go!

[wearthefoxhat]

End of the season, report will be must try harder but I did learn a lot and that is the main thing. Bring on next season

The main takeaway for me is that anything relating to mathematical prediction modelling, can be achieved, mainly thanks to the GPT.

Kie Miller's poisson distribution Y-T video, helped give me starting point and with a little perserverance, I was able to produce something that logically looks the part and works well.

I've also created a neat horse racing excel sheet that looks at ratings/value, and, currently in the process of a creating a cracking greyhound one too.

All of these are value based, conventional ways of betting and it's surprising how often the market is out of step and exploitable.

[wearthefoxhat]

As we near a new Premier League season, I thought I'd test out a more conventional approach until the form settles down.

I'm re-constructing last years Championship season by converting key metrics and xG into goal comparisons, home/away. Essentially, if they were to match up in the next round of games, then the score-line is already established to aid prediction.

After 12 rounds of games (end of October), what is shown is the Home/Away goals that "should" have occurred if the key metrics were satisfied as per their probability of happening. This is in-line with how I saw the early part of the season with Middlesbro' performing well, but their league position was around 6th. Leeds were already showing up well, tracked by the usual suspects.

Also, it's easy to spot teams like Watford & Blackburn whose home form outweighed their away form and Plymouth's away form that, as we know, couldn't recover from through out the season. A team rating also helps gives some order.

Round 12.png

Running alongside it, is the data table for home and away games. which, when updated, converts performance into points, then into goals prediction, ready for the next round of games.

Champ Table.png

The trick is to find the games where a team may have won, but the stats indicate otherwise. The market then might overbet certain teams and create value on others.

[TupleVision]

As we near a new Premier League season, I thought I'd test out a more conventional approach until the form settles down.

I'm re-constructing last years Championship season by converting key metrics and xG into goal comparisons, home/away. Essentially, if they were to match up in the next round of games, then the score-line is already established to aid prediction.

After 12 rounds of games (end of October), what is shown is the Home/Away goals that "should" have occurred if the key metrics were satisfied as per their probability of happening. This is in-line with how I saw the early part of the season with Middlesbro' performing well, but their league position was around 6th. Leeds were already showing up well, tracked by the usual suspects.

Also, it's easy to spot teams like Watford & Blackburn whose home form outweighed their away form and Plymouth's away form that, as we know, couldn't recover from through out the season. A team rating also helps gives some order.


Round 12.png


Running alongside it, is the data table for home and away games. which, when updated, converts performance into points, then into goals prediction, ready for the next round of games.



Champ Table.png


The trick is to find the games where a team may have won, but the stats indicate otherwise. The market then might overbet certain teams and create value on others.

Are you using xGa too or just xG?

I'm getting solid results blending both and computing an attack/defence score.

[wearthefoxhat]

As we near a new Premier League season, I thought I'd test out a more conventional approach until the form settles down.

I'm re-constructing last years Championship season by converting key metrics and xG into goal comparisons, home/away. Essentially, if they were to match up in the next round of games, then the score-line is already established to aid prediction.

After 12 rounds of games (end of October), what is shown is the Home/Away goals that "should" have occurred if the key metrics were satisfied as per their probability of happening. This is in-line with how I saw the early part of the season with Middlesbro' performing well, but their league position was around 6th. Leeds were already showing up well, tracked by the usual suspects.

Also, it's easy to spot teams like Watford & Blackburn whose home form outweighed their away form and Plymouth's away form that, as we know, couldn't recover from through out the season. A team rating also helps gives some order.


Round 12.png


Running alongside it, is the data table for home and away games. which, when updated, converts performance into points, then into goals prediction, ready for the next round of games.



Champ Table.png


The trick is to find the games where a team may have won, but the stats indicate otherwise. The market then might overbet certain teams and create value on others.

Are you using xGa too or just xG?

I'm getting solid results blending both and computing an attack/defence score.

Well done.

With this approach, I'm scoring each round's home/away result and their respective attack/defence stats, to produce a "GoaI Impact Score" that can then be compared to the next game match ups.

The scoring can then be analysed over previous rounds to see an upturn/downturn in form.

In previous posts, I used the Poisson Excel sheets that are good for predictive models. Both approaches could be used once there's sufficient data.

[wearthefoxhat]

Continuing on with the re-construction, after round 16, not too much has changed. Leeds remain strong, alongside Middlesbro' whom were handing out 5-1 & 6-2 beatings to confirm the previous postings stats. Using my own metrics, so far, no side has "bested" Leeds on the stats, although they had lost 0-1 to Burnley at Home and 1-0 to Millwall away, both "should" have been draws.

Sheff Utd at home were the best at this stage, alongside Middlesbro' away. It was looking grim for Portsmouth & Plymouth and no real signs of the trouble for Luton, as yet.

After Rd 16.png

---


I've added in the True Odds calculator that Paul Steele covered on Page 21 of his Profitable Football Betting book. Essentially, it calculates the true odds of the Home team, based a league table's HW/HL/AW/AL/Games Played. The same calculation can cover the Away team, then what remains, can make up the draw odds. I bolted it in alongside the Poisson Distribution excel sheet, so it's all in one place and useful for comparison.

Round 17 True Odds.png

----


The Poisson sheet looks at the Attack/Defence stats, based on goals scored/conceded, so it produces a different stand point to then compare with true odds calculator above.

Comparisons.png

In this game, the predicted odds are short (understandably) as Plymouth were in poor form, especially away. The final score was 6-1. Can't recall the actual odds of the game pre-off, if it was anything around 1.40/1.50 for Norwich, then it would have been a banker home win for sure.

------

[wearthefoxhat]

Up to Round 20 (re-construction of the last years Championship season) 10th December 2024.

Round 20 Chmp.png

Leeds continue to dominate, although Blackburn "bested" them 1-0 (Metrics score was 3-2) since the last post. Burnley are starting to draw games, their home defence is solid, and it seems away from home they match up well. Sheff Utd and Sunderland look similar to Burnley, although they are both finding a way to win.

At the other end of the table, Cardiff are starting to struggle, Plymouth's away form is still awful, with Portsmouth's away record looking poor too. Luton's away form is starting to look ragged.

The way I have the colour coding set up, helps quickly visualise the strengths and weaknesses of home/away performance. Also, I can do further form tables by using the team rating data and see if a side is progressive/regressive in the last 4/6 games...etc.

As I have everything bolted inside the old Poisson Distribution spreadsheet, I can set up the games early and produce PD odds and true odds, along with score-line predictions.

I can see over the new season, I'll need to specialise in one or two leagues as the data input is quite a bit.

[wearthefoxhat]

Quick post on the Premier League Round 1 results, for comparative purposes. Based on key in-game metrics and a few other bits and bobs, this is how I see the "true" results should have been in round 1.

Round 1 Prem.png

Main differences.


On balance, Newcastle should have nicked the game against 10 man Aston Villa. (sent off 66th minute)

Newc 2Half.png

----

Man Utd should of gained at least 1 point and the game have more goals than just 1.

The Arsenal defence rated well, including the keeper.

Arsenal Def.png

-----