Relegation watch

[Ken Tram]

Probability of Relegation

0% Huddersfield
39% Derby (+13%)
76% Rotherham (-7%)
85% Sheffield (-5%)
100% Wycombe
(Wycombe > 99.99%)

Before today's games

0.3% Huddersfield
26.1% Derby County
83.3% Rotherham United
90.3% Sheffield Wednesday
100.0% Wycombe Wanderers

Note

This is not taking into account trends, but is based upon the season's home and away results of these teams and their opponents, in order to get a feel for the spread of results that a team can tend to have, considering the variability of match outcomes.

This analysis considers takes each team's home or away results into account to gain a probability of getting 0, 1 or 3 points against whichever opponent they've got, and whether they're playing home it away.

So, there are lots of factors that are not being taken into account. But because football results are so variable, not infrequently being down to momentary luck; there is perhaps not much to be gained from incorporating lots of factors, because the analysis is only indicative in the first place. Using home and away tables does give a sense of the spread of results that there team has achieved.

The idea is to get a feel for the probabilities - in a way that is hard when juggling all of the different opponents and their rankings.

In terms of trends, and other factors that seem important - anyone can just tweak the figures up or down a bit. But, these baseline figures can still give us something to think about, and to either reinforce or thinking about what might happen, or to make us challenge whatever we are thinking, and to double check our thoughts.

I ran the matches 10,000 times; creating 10,000 league tables; and then counted how many times each team would have been relegated.

I was able to ignore goal difference because the scenarios in which it would decide promotion seemed to require relatively large swings in goal difference. But, we can lop off a percentage point or two to cover that. (I would have had to incorporate scorelines if it had been a greater factor, but it would probably have made it much more complicated, for not very much gain.)

I just wanted to try out a systematic way to take into account all of the different permutations! It did not need to be super accurate to achieve that. And then when I got the 26.1% figure for Derby yesterday (bring the equivalent of flipping two tails in a row), I thought that I'd share it!

Obviously, the bookmakers use the best methods to calculate odds, so that's the place to go for super accurate predictions, although I guess that betting odds are adjusted to maximise profit, which is not the same as being proportional to the outcomes that bookmakers actually expect!
 

[JuanFloEvraTheCocu'sNesta]

We won't beat Sheff Wed's. We just won't. 

[Ken Tram]

23 minutes ago, 86 Hair Islands said:

Good work mate! Whatever parameters you used it's hard to crab the results. I've just run a very basic calculation and came up with the following:

Derby - 56%
Rotherham - 61%
Wednesday - 88%
Wycombe - 100%

No modelling as such indeed quite notional as I've essentially run a book on the outcome of the remaining games with my objective head on  and my numbers are a mile off way off the figures from the professional odds compilers.

What does your modelling suggest now if it's not too big a job to re-run the numbers?

 

 

Thanks for the feedback. I appreciated it.

Good stuff. I guess that when I did it last night, I was trying to juggle two or three matches per team.

And I also wanted to overcome my need to believe(!) and to convince myself that Derby would be fine.

My figures do give Derby a lower chance of relegation then yours.

That might be because Derby's recent trend has a much larger effect upon our chances of beating Sheffield.

With one game to go, my analysis is pretty meaningless because lots of fans will probably have a more realistic idea of our chances of beating Sheffield.

And, of course, if Rotherham  win the next two games, then Derby have to beat Sheffield by at least 8. And the other permutations of Rotherham getting a win and a draw.

Edited May 1, 2021 by Ken Tram

[Jimbo Ram]

2 minutes ago, JuanFloEvraTheCocu'sNesta said:

We won't beat Sheff Wed's. We just won't. 

A draw could be enough. I have thought for the last 4 or 5 weeks that we wouldn’t get another point and that if we needed something from the final game we were doomed. I do hope I am wrong...

[SecretDave]

I think our best hope is that Rotherham lose one of their matches and we somehow get a draw. Can't see it happening though. 

[Comrade 86]

1 minute ago, Ken Tram said:

Probability of Relegation

0% Huddersfield
39% Derby (+13%)
76% Rotherham (-7%)
85% Sheffield (-5%)
100% Wycombe
(Wycombe > 99.99%)

Before today's games

0.3% Huddersfield
26.1% Derby County
83.3% Rotherham United
90.3% Sheffield Wednesday
100.0% Wycombe Wanderers

Note

This is not taking into account trends, but is based upon the season's home and away results of these teams and their opponents, in order to get a feel for the spread of results that a team can tend to have, considering the variability of match outcomes.

This analysis considers takes each team's home or away results into account to gain a probability of getting 0, 1 or 3 points against whichever opponent they've got, and whether they're playing home it away.

So, there are lots of factors that are not being taken into account. But because football results are so variable, not infrequently being down to momentary luck; there is perhaps not much to be gained from incorporating lots of factors, because the analysis is only indicative in the first place. Using home and away tables does give a sense of the spread of results that there team has achieved.

The idea is to get a feel for the probabilities - in a way that is hard when juggling all of the different opponents and their rankings.

In terms of trends, and other factors that seem important - anyone can just tweak the figures up or down a bit. But, these baseline figures can still give us something to think about, and to either reinforce or thinking about what might happen, or to make us challenge whatever we are thinking, and to double check our thoughts.

I ran the matches 10,000 times; creating 10,000 league tables; and then counted how many times each team would have been relegated.

I was able to ignore goal difference because the scenarios in which it would decide promotion seemed to require relatively large swings in goal difference. But, we can lop off a percentage point or two to cover that. (I would have had to incorporate scorelines if it had been a greater factor, but it would probably have made it much more complicated, for not very much gain.)

I just wanted to try out a systematic way to take into account all of the different permutations! It did not need to be super accurate to achieve that. And then when I got the 26.1% figure for Derby yesterday (bring the equivalent of flipping two tails in a row), I thought that I'd share it!

Obviously, the bookmakers use the best methods to calculate odds, so that's the place to go for super accurate predictions, although I guess that betting odds are adjusted to maximise profit, which is not the same as being proportional to the outcomes that bookmakers actually expect!
 

Christ that was quick! ? It's fascinating how diverse the numbers are, yours and mine I mean. The most interesting figure for me is the Rotherham number. Predictably your way more sophisticated analysis is much closer to my basic math and perceived trends (read gamblers hunch) as my unsubstantiated workings imply Rotherham are only marginally less likely to stay up than we are.

Can I politely ask (in layman's terms) how you go about your assessment? That is, what are the inputs and how do you decide how to 'weight' them? Apologies for silly questions but I'm fascinated by this. I don't want to hijack the thread nor take up lots of your time so DM me if you can be bothered and no worries if not!

[Rample]

1 hour ago, G-Ram said:

Its mental to think we could finish on the same number of points as wycombe when they win on the last day and we get beat 

If anyone had said that at any point up until March I wouldn't have believed it.

Credit to them for battling. Bet their fans are at least proud of their players. Quality definitely a factor, attitude not in question.

[Comrade 86]

11 minutes ago, Ken Tram said:

With one game to go, my analysis is pretty meaningless because lots of fans will probably have a more realistic idea of our chances of beating Sheffield.

I disagree and I think you are being way too modest here. Dispassionate and methodical analysis trumps even 'considered opinion' every time (well mostly!), especially over a period of time. Very telling that your figures are broadly speaking on a par with industry professionals too whereas mine really are not! Thank you in any case. A welcome diversion on a grim afternoon.

Edited May 1, 2021 by 86 Hair Islands

[Ken Tram]

9 minutes ago, Jimbo Ram said:

A draw could be enough. I have thought for the last 4 or 5 weeks that we wouldn’t get another point and that if we needed something from the final game we were doomed. I do hope I am wrong...

Benjamin Bloom has been saying for weeks that 45 points would be needed, and maybe one or two less.

That's why, like you, just getting one or two points seemed like all we would need.

Maybe that got into the players' heads too.

If I knew that I needed one point out of six games, I might have thought that it was a done deal, and not come to the batches with the right mindset. Maybe?

[Ken Tram]

22 minutes ago, JuanFloEvraTheCocu'sNesta said:

We won't beat Sheff Wed's. We just won't. 

You might be right. (At least you won't mind if you get your prediction wrong!) Maybe put some cash on Wednesday winning or drawing? Then, at least you have funded a sesh if Derby lose.

But, the whole season, I have not thought that we should count Sheffield as being a team below us, because at each point in the season, their form has exceeded their position because of the points deduction.

I didn't go as far as you in thinking that we would lose to Sheffield, but I didn't think that Sheffield would end up in the bottom three. Obviously, I have changed my mind a bit in recent months - but my initial thoughts seemed to be similar to the one that you have now.

Edited May 1, 2021 by Ken Tram

[JuanFloEvraTheCocu'sNesta]

5 minutes ago, Ken Tram said:

You might be right. (At least you won't mind if you get your prediction wrong!) Maybe put some cash on Wednesday winning or drawing? Then, at least you have funded a sesh if Derby lose.

But, the whole season, I have not thought that we should count Sheffield as being a team below us, because at each point in the season, their form has exceeded their position because of the points deduction.

I didn't go as far as you in thinking that we would lose to Sheffield, but I didn't think that Sheffield would end up in the bottom three. Obviously, I have changed my mind a bit in recent months - but my initial thoughts seemed to be similar to the one that you have now.

We could be playing a National League side and I'd back us to lose. We are so pathetically fragile that any semi pro team with the desire to put some effort in would turn us over in our current state. 

Our squad and management team simply do not deserve to be in their respective positions. 

Edited May 1, 2021 by JuanFloEvraTheCocu'sNesta

[RammingStone66]

4 minutes ago, JuanFloEvraTheCocu'sNesta said:

We could be playing a National League side and I'd back us to lose. We are so pathetically fragile that and semi pro team with the desire to put dome effort in would tutn us over in our current state. 

Our squad and management team simply do not deserve to be in their respective positions. 

Well we did go 0-0 with barrow and lose to Chorley this season didn't we? ? your perspective seems pretty nailed on. Any team with drive or determination and we are screwed

[Gone]

Early goal for Wednesday next week and you can do all the statistical modelling you desire, we are down. If I was Darren Moore I’d say get as many corners or set pieces around their area as you can in the 1st 15 mins and you’ll win the game. 

[TomTom92]

Still fuming now.

[JuanFloEvraTheCocu'sNesta]

2 minutes ago, TomTom92 said:

Still fuming now.

Even if we stay up we should all be very angry. So many people at the club from top to bottom asleep at the wheel right now it's a ducking joke. 

[Jimbo Ram]

28 minutes ago, Ken Tram said:

Benjamin Bloom has been saying for weeks that 45 points would be needed, and maybe one or two less.

That's why, like you, just getting one or two points seemed like all we would need.

Maybe that got into the players' heads too.

If I knew that I needed one point out of six games, I might have thought that it was a done deal, and not come to the batches with the right mindset. Maybe?

Is he the actor?

[Jimbo Ram]

5 minutes ago, TexasRam said:

Early goal for Wednesday next week and you can do all the statistical modelling you desire, we are down. If I was Darren Moore I’d say get as many corners or set pieces around their area as you can in the 1st 15 mins and you’ll win the game. 

Late Jordon Rhodes header from a set piece to win it written all over it...

[Mckram]

Id love to see the % chance of us going down 13/14 games ago.

[Jimbo Ram]

Just now, Mckram said:

Id love to see the % chance of us going down 13/14 games ago.

I sensed difficult times ahead when Bielik was injured...

[Ken Tram]

57 minutes ago, 86 Hair Islands said:

Christ that was quick! ? It's fascinating how diverse the numbers are, yours and mine I mean. The most interesting figure for me is the Rotherham number. Predictably your way more sophisticated analysis is much closer to my basic math and perceived trends (read gamblers hunch) as my unsubstantiated workings imply Rotherham are only marginally less likely to stay up than we are.

Can I politely ask (in layman's terms) how you go about your assessment? That is, what are the inputs and how do you decide how to 'weight' them? Apologies for silly questions but I'm fascinated by this. I don't want to hijack the thread nor take up lots of your time so DM me if you can be bothered and no worries if not!

Sure. Firstly, I did not take into account trends! I just wanted to find a way to get a sense of the permutations of all the matches. Also, as we know, just because a team has lost five games in a row, it does not mean that they won't beat the team that's top of the league in the next match! So, the easiest thing that I could think of doing was to look at the season's home and away tables. (Incidentally, Wycombe have won 6 out of 22 games at home.) That's not zero. And even if such a team were on an upward or downward trend, it would be wrong to simply assume that they will lose! So, taking the league and away form did give a way to incorporate the wide spread of results that can happen.

And I bet that I did something fundamentally flawed - but my intention was to get a feel for the permutations on my phone in the middle of the night!

And I wanted a way to factor in the fact that teams were playing opponents from different ends of the table!

Anyway, what I did was as follows:

Middlesbrough's home form is: win 11, draw 4, lose 7.

Wycombe's away form is: win 3, draw 5, lose 14.

That's 44 games altogether. 

So, in order to get a simplistic indication of the chance of a result, taking into account the opponent's position in the league, was jus to combine the figures above, as follows:

Wycombe winning can be seen as having a probability of 10/44 (3+7),

drawing as 9/44 (4+5), and

losing being 25/44 (14+11).

And then I pick a random number between 1 and 44. If it is 1-10, Wycombe get 3 points; between 11-19, they get 1 point; and between 20-44, then they get 0 points.

I did this for the 10 games affecting the bottom 5 clubs covering the last two weeks, and found the final league position. I then repeated it 1,000 times, and counted how many times a team finished in the bottom 3. And then repeated that ten times.

Luckily, I did not have to worry about goal difference, because in all of the scenarios where relegation would depend upon goal difference, the required swing was too small to bother adding that complexity. I got round this by giving Derby an extra 0.1 points, Sheffield 0.2, and Rotherham 0.3; because the various permutations resulting in pairs of these teams ending up with the same number of points, and where it would also affect relegation (as opposed to league position), could be addressed by those decimal places. I was lucky that the was such a fudge.

This was a bit of a fudge, but, because I was only trying to get an indication to interpret the matches, it didn't seem to matter.

Edited May 1, 2021 by Ken Tram