• Welcome to the ShrimperZone forums.
    You are currently viewing our boards as a guest which only gives you limited access.

    Existing Users:.
    Please log-in using your existing username and password. If you have any problems, please see below.

    New Users:
    Join our free community now and gain access to post topics, communicate privately with other members, respond to polls, upload content and access many other special features. Registration is fast, simple and free. Click here to join.

    Fans from other clubs
    We welcome and appreciate supporters from other clubs who wish to engage in sensible discussion. Please feel free to join as above but understand that this is a moderated site and those who cannot play nicely will be quickly removed.

    Assistance Required
    For help with the registration process or accessing your account, please send a note using the Contact us link in the footer, please include your account name. We can then provide you with a new password and verification to get you on the site.

Pubey

Guest
For stats nerds.

I decided to do a simulation of the remaining games.

My simulation is simple and limited, for now.

I use the season form of each team (probability of a win, draw or loss) to simulate the outcome of all remaining fixtures.

I then ran the simulation 10,000 times to work out a prediction of what might happen to Southend.

First place = 3.7%
2nd or 3rd place = 24.5%
Playoffs = 60.1%
Miss out completely = 11.8%

(Never in 10,000 simulations were we relegated - are we actually mathematically safe?)

Important details/assumptions:
- Each predicted result is independent of all other results - therefore doesn't account for recent form
- All teams are 'assumed equal' - The probabilities of the result are based on the home team and on their past performance across the league. It's not based on the actual match, who they're facing and if it's home and away. I might try this but it'll take a lot more time.
- When the model predicts the ranked league table, and where teams have level points, the team who are currently higher in the table are assumed to rank highest. This is probably a fair assumption and avoids having to simulate goals scored.

Enjoy!
 
Last edited:

Pubey

Guest
You have too much time on your hands, but looks like play-offs then!

It took me half an hour to put together while Only Connect was on. 15 mins was spent redoing it after I realised I needed to re-do my results grid in league position order rather than alphabetical order!

If anyone wants the model (Excel) PM me and I'll dropbox it.
 

DoDTS

The PL League Boss⭐
Joined
Apr 28, 2006
Messages
8,996
Location
PL Headquarters Hullbridge

First place = 3.7%
2nd or 3rd place = 24.5%
Playoffs = 60.1%
Miss out completely = 11.8%

(Never in 10,000 simulations were we relegated - are we actually mathematically safe?)

Hartlepool can't catch us and if Tranmere lose at Cambridge tonight or we win at Burton then we are mathematically safe!

Unless we had points deducted of course.....................
 

Jam_Man

Life President
Joined
Apr 7, 2005
Messages
25,545
Location
Southend
Lol, how on earth did you manage this!?

On face value the percentages look about right though dont they, Most of the season we have been in the play offs so thats the likely outcome unless we get a sudden kick in form.
 

Pubey

Guest
Lol, how on earth did you manage this!?

On face value the percentages look about right though dont they, Most of the season we have been in the play offs so thats the likely outcome unless we get a sudden kick in form.

I think face value looks ok, it is simplistic though.

The basic method is very simple.

1. Take the current results of the teams to work out their probability distribution of win/draw/loss.

Southend's is 17/11/8 so 47.2%/30.6%/22.2%

2. For each remaining fixture, use the probabilities for the home team.

3. Using a uniform random number draw from 0-1 (simply rand() in Excel) you can sample from that probability distribution

This is just a little nested IF statement that says if rand() < 47.2% then W. if rand() > (1-22.2%) then L, else D.
You use 1 rand() for the internal if statement.

Then if you press F9 on Excel the worksheet will re-calculate, which will basically sample a result based on those probabilities

4. Do this for all remaining fixtures

5. Work out total points for each team, and use rank.eq() function to rank teams (that's why the list of teams needs to be in order of current league position!)

6. Write a little VBA macro that will loop from 1 to 10,000, each time copy and pasting the resulting league table into a long list (offset based on loop value).

SIMPLES.

Like I said, happy to send it across to anyone who's interested. There's potential to do more with it but that would get a lot more complex. Basically wanting to predict outcomes based on the combined form of the two teams.
 

Prawn

Where's The Fooball!⭐
Joined
Sep 6, 2009
Messages
1,500
Location
Perth, Australia
For stats nerds.

I decided to do a simulation of the remaining games.

My simulation is simple and limited, for now.

I use the season form of each team (probability of a win, draw or loss) to simulate the outcome of all remaining fixtures.

I then ran the simulation 10,000 times to work out a prediction of what might happen to Southend.

First place = 3.7%
2nd or 3rd place = 24.5%
Playoffs = 60.1%
Miss out completely = 11.8%

(Never in 10,000 simulations were we relegated - are we actually mathematically safe?)

Important details/assumptions:
- Each predicted result is independent of all other results - therefore doesn't account for recent form
- All teams are 'assumed equal' - The probabilities of the result are based on the home team and on their past performance across the league. It's not based on the actual match, who they're facing and if it's home and away. I might try this but it'll take a lot more time.
- When the model predicts the ranked league table, and where teams have level points, the team who are currently higher in the table are assumed to rank highest. This is probably a fair assumption and avoids having to simulate goals scored.

Enjoy!

Are you able to do another mathematical simulation to see if there is intelligent life in a certain north-eastern Essex location?
 

londonblue

Topgun Pilot
Joined
Feb 18, 2004
Messages
16,156
For stats nerds.

I decided to do a simulation of the remaining games.

My simulation is simple and limited, for now.

I use the season form of each team (probability of a win, draw or loss) to simulate the outcome of all remaining fixtures.

I then ran the simulation 10,000 times to work out a prediction of what might happen to Southend.

First place = 3.7%
2nd or 3rd place = 24.5%
Playoffs = 60.1%
Miss out completely = 11.8%

(Never in 10,000 simulations were we relegated - are we actually mathematically safe?)

Important details/assumptions:
- Each predicted result is independent of all other results - therefore doesn't account for recent form
- All teams are 'assumed equal' - The probabilities of the result are based on the home team and on their past performance across the league. It's not based on the actual match, who they're facing and if it's home and away. I might try this but it'll take a lot more time.
- When the model predicts the ranked league table, and where teams have level points, the team who are currently higher in the table are assumed to rank highest. This is probably a fair assumption and avoids having to simulate goals scored.

Enjoy!

Don't worry, if it wasn't we wouldn't be able to understand!
 

JD

Resident Jinx
Joined
Aug 5, 2006
Messages
930
Location
Brentwood
For stats nerds.

I decided to do a simulation of the remaining games.

My simulation is simple and limited, for now.

I use the season form of each team (probability of a win, draw or loss) to simulate the outcome of all remaining fixtures.

I then ran the simulation 10,000 times to work out a prediction of what might happen to Southend.

First place = 3.7%
2nd or 3rd place = 24.5%
Playoffs = 60.1%
Miss out completely = 11.8%

(Never in 10,000 simulations were we relegated - are we actually mathematically safe?)

Important details/assumptions:
- Each predicted result is independent of all other results - therefore doesn't account for recent form
- All teams are 'assumed equal' - The probabilities of the result are based on the home team and on their past performance across the league. It's not based on the actual match, who they're facing and if it's home and away. I might try this but it'll take a lot more time.
- When the model predicts the ranked league table, and where teams have level points, the team who are currently higher in the table are assumed to rank highest. This is probably a fair assumption and avoids having to simulate goals scored.

Enjoy!

How can we trust someone who can't add up?:winking::smile:
 

Pubey

Guest
Don't worry, if it wasn't we wouldn't be able to understand!

That's part of the issue. There are plenty of websites that do similar for the top leagues, but there's no way of getting to grips with their actual method - because they're all secretive and potentially either selling their results, or using it for betting purposes!

There are some academic groups and journal papers looking at this but as you say, it'd still lose some validity because these are much more complex and involved.

My biggest issue is finding a decent website where the data for each season can be easily parsed and downloaded. Once it's all in excel then it's relatively easy, but manually entering 528(?) results isn't going to happen.
 

Harry Bullocks

President⭐
Joined
Aug 7, 2009
Messages
7,053
Location
Exile in Hampshire
We are talking about Southend United here :scared:......probabilities and possibilities.....the eternal roller coaster ride, I know having followed The Shrimpers for the last 40 years....anyway, top work Pubey
 

Pubey

Guest
Updated after tonights games. Plus bonus graph!

ka4qdt.png
 
Top