Goals scored (1 Viewer)

Forums New posts

Can we score 100+ goals in a season?

  • Yes

    Votes: 18 20.0%

  • No

    Votes: 72 80.0%
  • Total voters
    90
JSL

JSL

Well-Known Member
We are on 72 goals scored with 11 games to go. 28 goals needed to break 100 goals. Yes or No?
 
fatso

fatso

Well-Known Member
A more realistic poll would be "will we get another penalty before the end of the season"
Or "will Haji Wright go 90 minutes without being caught offside" or "will Thomas score from a corner ever again"
 
Skybluedownunder

Skybluedownunder

Well-Known Member
Other than Sheff Wed there are no easy games, don’t see us getting close to nearly 3 goals a game as an average, tough fixtures


Sent from my iPhone using Tapatalk
 
mrfr

mrfr

Well-Known Member
Using the same model I built to project final league positions, this is how it sees our final goal tally. Only a 1.3% chance of scoring 100 goals.

Code: 
Current:   72 goals in 35 games  (2.06 per game)
Remaining: 11 fixtures

Projected final total:
  Central estimate:   89 goals  (median)
  Mean:               89.5
  Range:              ±4.2 goals (1 standard deviation)

Distribution:
  80–84 goals   10.8%  ████
  85–89 goals   40.3%  ████████████████
  90–94 goals   36.5%  ██████████████
  95–99 goals   10.7%  ████
  100+ goals     1.3%


Per fixture expected goals:
  Sat 7 Mar   (A) Bristol City    1.49
  Wed 11 Mar  (H) Preston         1.56
  Sat 14 Mar  (H) Southampton     1.63
  Sat 21 Mar  (A) Swansea         1.47
  Fri 3 Apr   (H) Derby           1.63
  Mon 6 Apr   (A) Hull            1.56
  Sat 11 Apr  (H) Sheff Wed       1.94  ← highest
  Sat 18 Apr  (A) Blackburn       1.52
  Tue 21 Apr  (H) Portsmouth      1.62
  Sat 25 Apr  (H) Wrexham         1.62
  Sat 2 May   (A) Watford         1.44  ← lowest
 
Last edited:
skybluecam

skybluecam

Well-Known Member
mrfr said:
Using the same model I built to project final league positions, this is how it sees our final goal tally. Only a 1.3% chance of scoring 100 goals.

Code: 
Current:   72 goals in 35 games  (2.06 per game)
Remaining: 11 fixtures

Projected final total:
  Central estimate:   89 goals  (median)
  Mean:               89.5
  Range:              ±4.2 goals (1 standard deviation)

Distribution:
  80–84 goals   10.8%  ████
  85–89 goals   40.3%  ████████████████
  90–94 goals   36.5%  ██████████████
  95–99 goals   10.7%  ████
  100+ goals     1.3%


Per fixture expected goals:
  Sat 7 Mar   (A) Bristol City    1.49
  Wed 11 Mar  (H) Preston         1.56
  Sat 14 Mar  (H) Southampton     1.63
  Sat 21 Mar  (A) Swansea         1.47
  Fri 3 Apr   (H) Derby           1.63
  Mon 6 Apr   (A) Hull            1.56
  Sat 11 Apr  (H) Sheff Wed       1.94  ← highest
  Sat 18 Apr  (A) Blackburn       1.52
  Tue 21 Apr  (H) Portsmouth      1.62
  Sat 25 Apr  (H) Wrexham         1.62
  Sat 2 May   (A) Watford         1.44  ← lowest
Click to expand...
How are you modelling predicted expected goals per fixture?
 
shmmeee

shmmeee

Well-Known Member
skybluecam said:
How are you modelling predicted expected goals per fixture?
Click to expand...

Guessing it’s the average of the home xG and away xGA or visa versa looking at Sheff Wed. They’re away xGA is 1.91 our home xG is 1.99 so 1.94 is in the middle. That would always undershoot us though as the home xG leaders.
 
mrfr

mrfr

Well-Known Member
skybluecam said:
How are you modelling predicted expected goals per fixture?
Click to expand...

Step 1: Build a blended attack rating for Coventry and a blended defensive rating for each opponent.

Rather than relying solely on xG stats or actual goals, we blend the two equally. For Cov in attack, we average season xGF per game (1.74 from Opta) and actual goals per game (2.06 across 35 matches) to get 1.90.

We do the same for each opponent’s defence by averaging their xGA per game with their actual goals conceded per game.

The rationale behind blending is that pure xG can underestimate a team genuinely outperforming its expected output over a large sample while pure actual goals can be noisy and include fortunate finishes. Even after 35 games, both signals still carry valuable information.

Step 2: Combine attack and defence with a venue adjustment.

For a home fixture, we multiply Coventry’s blended attack by 1.10 (a home advantage boost) and the opponent’s blended defensive figure by 0.90 (teams defend slightly worse away). We then average the two. For an away fixture, we invert those multipliers.
 
shmmeee

shmmeee

Well-Known Member
mrfr said:
Step 1: Build a blended attack rating for Coventry and a blended defensive rating for each opponent.

Rather than relying solely on xG stats or actual goals, we blend the two equally. For Cov in attack, we average season xGF per game (1.74 from Opta) and actual goals per game (2.06 across 35 matches) to get 1.90.

We do the same for each opponent’s defence by averaging their xGA per game with their actual goals conceded per game.

The rationale behind blending is that pure xG can underestimate a team genuinely outperforming its expected output over a large sample while pure actual goals can be noisy and include fortunate finishes. Even after 35 games, both signals still carry valuable information.

Step 2: Combine attack and defence with a venue adjustment.

For a home fixture, we multiply Coventry’s blended attack by 1.10 (a home advantage boost) and the opponent’s blended defensive figure by 0.90 (teams defend slightly worse away). We then average the two. For an away fixture, we invert those multipliers.
Click to expand...

Why not just use home and away xG?
 
mrfr

mrfr

Well-Known Member
shmmeee said:
Why not just use home and away xG?
Click to expand...
Because xG measures the quality of chances created, not the quality of finishing, and its finishing which determines goals, not chances. Over the course of the season we have outperformed xG by 0.3 goals per game which over the course of 35 games becomes a pretty meaningful difference.
 
shmmeee

shmmeee

Well-Known Member
mrfr said:
Because xG measures the quality of chances created, not the quality of finishing, and its finishing which determines goals, not chances. Over the course of the season we have outperformed xG by 0.3 goals per game which over the course of 35 games becomes a pretty meaningful difference.
Click to expand...

But you still use xG, you just modify it with an arbitrary number.
 
skybluecam

skybluecam

Well-Known Member
mrfr said:
Step 1: Build a blended attack rating for Coventry and a blended defensive rating for each opponent.

Rather than relying solely on xG stats or actual goals, we blend the two equally. For Cov in attack, we average season xGF per game (1.74 from Opta) and actual goals per game (2.06 across 35 matches) to get 1.90.

We do the same for each opponent’s defence by averaging their xGA per game with their actual goals conceded per game.

The rationale behind blending is that pure xG can underestimate a team genuinely outperforming its expected output over a large sample while pure actual goals can be noisy and include fortunate finishes. Even after 35 games, both signals still carry valuable information.

Step 2: Combine attack and defence with a venue adjustment.

For a home fixture, we multiply Coventry’s blended attack by 1.10 (a home advantage boost) and the opponent’s blended defensive figure by 0.90 (teams defend slightly worse away). We then average the two. For an away fixture, we invert those multipliers.
Click to expand...
Interesting - I've seen 2/3 xG 1/3 actual goals used as a weighting before. What's Sheff Wed's predicted goals against us by your model?
 
shmmeee

shmmeee

Well-Known Member
mrfr said:
It’s not arbitrary, it’s actual goals, that’s a less arbitrary measure than xG, surely? But sure, it’s entirely opinionated, we could just try reading tea leaves.
Click to expand...

“

Step 2: Combine attack and defence with a venue adjustment.

For a home fixture, we multiply Coventry’s blended attack by 1.10 (a home advantage boost) and the opponent’s blended defensive figure by 0.90 (teams defend slightly worse away). We then average the two. For an away fixture, we invert those multipliers.”

Where have those numbers come from?
 
mrfr

mrfr

Well-Known Member
shmmeee said:
“

Step 2: Combine attack and defence with a venue adjustment.

For a home fixture, we multiply Coventry’s blended attack by 1.10 (a home advantage boost) and the opponent’s blended defensive figure by 0.90 (teams defend slightly worse away). We then average the two. For an away fixture, we invert those multipliers.”

Where have those numbers come from?
Click to expand...
That is an attempt to bake in some home advantage because if you don’t do that then any team with an uneven number of home/away fixtures remaining will have their xG skewed in one direction or another compared to their PPG. The 10% boost/penalty is an arbitrary number though, you’re right about that. we can though look at data across the season to try and quantify an average impact, and looking at that data says I likely underestimate that impact somewhat:

On a PPG basis the venue effect is much bigger. The league average home PPG is 1.531 vs away 1.224, a ratio of ×1.25. to mimic this we’d apply a .25 home advantage and away disentangle as opposed to the .1 I use.

On a goals basis the effect is more modest — 1.393 home goals per game vs 1.201 away, a ratio of ×1.16. That’s closer to my assumption but still above it marginally.

🤷🏻does that make sense? I think so, at least more sense than not factoring in any kind of home advantage.
 
mrfr

mrfr

Well-Known Member
skybluecam said:
Interesting - I've seen 2/3 xG 1/3 actual goals used as a weighting before. What's Sheff Wed's predicted goals against us by your model?
Click to expand...
0.79, with a 45% chance of them not scoring and an 81% chance of them scoring 0 or 1. Still might be a bit generous 😂
 
shmmeee

shmmeee

Well-Known Member
mrfr said:
That is an attempt to bake in some home advantage because if you don’t do that then any team with an uneven number of home/away fixtures remaining will have their xG skewed in one direction or another compared to their PPG. The 10% boost/penalty is an arbitrary number though, you’re right about that. we can though look at data across the season to try and quantify an average impact, and looking at that data says I likely underestimate that impact somewhat:

On a PPG basis the venue effect is much bigger. The league average home PPG is 1.531 vs away 1.224, a ratio of ×1.25. to mimic this we’d apply a .25 home advantage and away disentangle as opposed to the .1 I use.

On a goals basis the effect is more modest — 1.393 home goals per game vs 1.201 away, a ratio of ×1.16. That’s closer to my assumption but still above it marginally.

🤷🏻does that make sense? I think so, at least more sense than not factoring in any kind of home advantage.
Click to expand...

This is why I asked why not just use the home and away XG/xGA numbers. I’d expect some teams to set up very differently home and away. I still think this undershoots is a bit but I’m biased.
 
PUSB-We_are_going_up

PUSB-We_are_going_up

Well-Known Member
2.54 goals a game needed (28g in 11 games) with Sheffield Wednesday still to play at home as well that could be a few goals if we have our shooting boots on, it could also be the game we confirm promotion considering there are only 4 games left after
 
You must log in or register to reply here.

Users who are viewing this thread

Total: 2 (members: 0, guests: 2)
Top