Forum | Mathematical Model to compare Ipswich players. by TractorFrog

Forum index | Previous Thread | Next thread

Mathematical Model to compare Ipswich players. 19:37 - Mar 12 with 1242 views

This was the previous, simple version:

Statistical impact of Ipswich players this season. by TractorFrog 25 Feb 10:04
Based on the Axel Tuanzebe thread , here are those stats for every player who has started a game this year. I have calculated PPG in games they started, PPG in games they didn't start (sub appearances count here), then done the first number minus the second. So if the score is greater than 0, we get more points with them, and for a score less than 0, we get more points without them. Note that this exercise is most useful for players who have done about half the games, as if they have only played 1 or 2 games then it can be heavily influenced by a freak result, and the same if they have only missed 1 or 2 games. So here are the results

EDIT: (in brackets is the number of games played out of 26, or the number of games missed: whichever is fewer. To demonstrate that the higher the number, the more useful the data is. So the maximum number here is 13):

1. Liam Delap +0.708 (2)
2. Sam Morsy +0.562 (5)
3. Jens Cajuste +0.413 (10)
4. Nathan Broadhead +0.409 (4)
5. Dara O'Shea +0.362 (3)
6. Conor Townsend +0.360 (1)
7. Cameron Burgess +0.238 (10)
8. Wes Burns +0.179 (12)
9. Axel Tuanzebe +0.179 (12)
10. Omari Hutchinson +0.167 (2)
11. Harry Clarke +0.114 (4)
12. Sammie Szmodics +0.077 (13)
13. Aro Muric +0.042 (8)
14. Ben Johnson +0.020 (9)
15. Christian Walton +0.017 (6)
16. Cheo Ogbene +0.014 (3)

17. Luke Woolfenden -0.042 (8)
18. Conor Chaplin -0.113 (7)
19. Julio Enciso -0.167 (2)
20. Alex Palmer -0.167 (2)
21. Jacob Greaves -0.238 (10)
22. Leif Davis -0.360 (1)
23. Kalvin Phillips -0.440 (12)
24. Jack Clarke -0.633 (6)
25. George Hirst -0.680 (1)
26. Jack Taylor -0.680 (1)
27. Massimo Luongo -0.708 (2)
28. Jaden Philogene -0.708 (2)
29. Ben Godfrey -0.708 (2)

I would say these numbers make the best reading for Sam Morsy, Jens Cajuste and Cameron Burgess, and are not so good for Kalvin Phillips, Jacob Greaves and Jack Clarke.

I have now created a more comprehensive mathematical model for comparing the impact of the Ipswich Town players this season.

The improvements to this model on the last one is that it now takes into account the strength of the opposition, whether it is a home or away game, and adjusts for the higher uncertainty in results when the player has only started one or two games, or only missed one or two games. I haven't yet thought of a good way to include the effect of substitutes but that could maybe come later.

If you want to read about how it works and the process, as well as some of the caveats and what the numbers below represent, there is a link here: https://f1frogblog.wordpress.com/2025/03/12/ipswich-town-fc-mathematical-model/

And the final results are as follows:

1. Sam Morsy (1.295)
2. Nathan Broadhead (1.005)
3. Christian Walton (0.917)
4. Liam Delap (0.866)
5. Omari Hutchinson (0.836)
6. Wes Burns (0.784)
7. Cameron Burgess (0.719)
8. Jens Cajuste (0.634)
9. Axel Tuanzebe (0.389)
10. Conor Townsend (0.368)
11. Sammie Szmodics (0.311)
12. Harry Clarke (0.209)
13. Luke Woolfenden (0.207)
14. Chiedozie Ogbene (0.182)
15. Dara O'Shea (0.164)
16. Ben Johnson (0.131)
17. Aro Muric (0.091)
18. Conor Chaplin (-0.015)
19. Leif Davis (-0.368)
20. Jack Taylor (-0.489)
21. Massimo Luongo (-0.604)
22. Jacob Greaves (-0.719)
23. Ben Godfrey (-0.724)
24. Julio Enciso (-0.784)
25. George Hirst (-0.835)
26. Alex Palmer (-1.058)
27. Kalvin Phillips (-1.168)
28. Jaden Philogene (-1.527)
29. Jack Clarke (-1.892)

	They'd all laugh at me if they knew what I was trying to do. To create a new strain of super-wine in half-an-hour with a fraction of nature's resources and a FOOL for an assistant. 'Bernard Black, he's mad,' they'd say, 'he's insane, he's dangerous.' Well I'll show them! I'll show them all!
Poll:	Who should start in CM with Cajuste?

Mathematical Model to compare Ipswich players. on 21:49 - Mar 12 with 1092 views

matt1977

A little too much time on your hands?

Mathematical Model to compare Ipswich players. on 21:56 - Mar 12 with 1078 views

Nthsuffolkblue

A few of those surprise me (Chaplin especially). Palmer and Enciso look better than their stats but have only been in the team where our form has dropped off.

Cajuste, Burgess, Tuanzebe and Burns appear to come out especially well on that raw data.

Poll:	How do you feel about the re-election of Trump?
Blog:	[Blog] Ghostbusters

Mathematical Model to compare Ipswich players. on 21:11 - Mar 15 with 766 views

TractorFrog

We are now on seven games where Sam Morsy didn't start, and just one point from those seven games. 16 points from 22 vs 1 point from 7 is a considerable difference.

His p-value, in the model including the level of opposition and home/away, drops to 0.194. This means that, if Sam Morsy didn't make any difference to the team at all, there is about a 19.4% chance that the statistics could be this heavily weighted one way or another.

So there is about an 80.6% chance that Ipswich are likely to get better results if Sam Morsy is in the team. Usually statisticians require 95% to consider a result significant, but I think it depends on the context and that 80.6% is the highest in the squad (for positive contribution - Jack Clarke is on 92.9% the other way).

This doesn't take into account who the other players on the pitch are, but this is just directly comparing Ipswich Town with Morsy to Ipswich Town without Morsy.

I think his role as a captain maybe plays a role in motivating the rest of the team to cause this result. Also considering he will be here next year, and Kalvin Phillips won't, I hope Morsy returns to the team for the Bournemouth game.

	They'd all laugh at me if they knew what I was trying to do. To create a new strain of super-wine in half-an-hour with a fraction of nature's resources and a FOOL for an assistant. 'Bernard Black, he's mad,' they'd say, 'he's insane, he's dangerous.' Well I'll show them! I'll show them all!
Poll:	Who should start in CM with Cajuste?

Mathematical Model to compare Ipswich players. on 11:48 - Mar 30 with 458 views

TractorFrog

I tried one where you do it by position, and compare all the GKs, all the RBs etc, still including the league position and home/away. These were the results:

GK
Walton > Muric > Palmer

RCB
Woolfenden > O'Shea

LCB
Burgess > Greaves

RB
Tuanzebe > O'Shea > H.Clarke > Johnson > Godfrey

LB
Townsend > Davis

RCM
Morsy > Phillips

LCM
Cajuste > Phillips > Luongo > Taylor

RW
Johnson > Burns > Ogbene > Hutchinson > Philogene

CAM
Szmodics > Hutchinson > Chaplin > Enciso

LW
Hutchinson > Broadhead > Enciso > Szmodics > Philogene > J.Clarke

ST
Delap > Szmodics > Hirst

	They'd all laugh at me if they knew what I was trying to do. To create a new strain of super-wine in half-an-hour with a fraction of nature's resources and a FOOL for an assistant. 'Bernard Black, he's mad,' they'd say, 'he's insane, he's dangerous.' Well I'll show them! I'll show them all!
Poll:	Who should start in CM with Cajuste?

Mathematical Model to compare Ipswich players. on 15:54 - Mar 30 with 349 views

Nthsuffolkblue

Mathematical Model to compare Ipswich players. on 11:48 - Mar 30 by TractorFrog

This is a good example of why stats alone cannot give the full answer.

I don't think many would rate Walton as our best keeper. I doubt anyone would put Szmodics as a striker ahead of Hirst.

Have you drilled down to separating out data for when Enciso and Szmodics have played each of those roles or is it standardised on whenever they have played? Is the data about the overall team result or down to individual performances within that? If the latter, is it assists and goals only or more? For defenders what is and isn't included?

Ultimately, information that is beyond pure stats is needed to really assess players.

Poll:	How do you feel about the re-election of Trump?
Blog:	[Blog] Ghostbusters

Mathematical Model to compare Ipswich players. on 16:11 - Mar 30 with 333 views

TractorFrog

Mathematical Model to compare Ipswich players. on 15:54 - Mar 30 by Nthsuffolkblue

It’s only on team result, and only on games started. So Szmodics and Hirst each started one game and lost, but as Newcastle (H) is a harder game than Brentford (A), Szmodics rates higher. I have a separate one for goal difference instead that puts Hirst above Szmodics, and the fact they are different demonstrates how stats can never tell the full story. Yes, for this one, I separated the games into ones where they played in each different position, although there are of course different versions of the lineup, it can be approximated into that 4-2-3-1.

For the goalkeeper thing, Walton has a good points per game considering he played against by far the toughest opposition of the three. The Chelsea win makes a big difference here, and again the large goal difference against Man City isn’t accounted for. Most would think Palmer is the best in goal but we get the worst results with him. Maybe this is partly down to being in the same team as other players who do poorly on the stats like Clarke, Philogene, Enciso and Phillips, or maybe it is because he doesn’t play out from the back as much.

I can do one where you also take into account who was playing around them (e.g. Burgess is considered to be better than Greaves, and Woolfenden’s games have all come alongside Greaves whereas O’Shea has had some with Burgess, so Woolfenden’s score goes up compared to O’Shea’s with that method). Ideally you would include all 10 other positions like that but there isn’t enough data to do that.

	They'd all laugh at me if they knew what I was trying to do. To create a new strain of super-wine in half-an-hour with a fraction of nature's resources and a FOOL for an assistant. 'Bernard Black, he's mad,' they'd say, 'he's insane, he's dangerous.' Well I'll show them! I'll show them all!
Poll:	Who should start in CM with Cajuste?

Forum index | Previous Thread | Next thread