Embed
Email

Kendall

Document Sample

Shared by: ajizai
Categories
Tags
Stats
views:
0
posted:
12/4/2011
language:
English
pages:
28
Scheduling English Football over Holiday

Periods



Please note: This version of my presentation has

had the pictures and background removed. This

is to reduce the (large) file size and also to ensure

there are no copyright issues with the pictures.





Graham Kendall

Contents

• Problem Description and Definition

• Solution Methodology

• Results

• Future Work and Possible Alan Shearer breaks Milburn's

record, scoring his 201st

Research Directions Newcastle goal (against

Portsmouth)

• Conclusions

• Questions

Problem Description

Football supporters and motoring

organisations have called for changes

in the scheduling of fixtures [over the

Christmas holiday period] to prevent

supporters from having to drive home

George Best won 37 caps for

from matches well into the early hours Northern Ireland, including this

of the morning. match against England in May

1971

The Times (Sport, Football), 30th December 2005, pp

68

League Structure

 Four Divisions

 Barclays Premiership (20 teams)

 Coca-Cola Championship (24 teams)

 Coca-Cola League One (24 teams)

 Coca-Cola League Two (24 teams)



 Number of Matches

 380 (Premiership), 552 for other divisions.

Total 2036

Michael Owen returns to the line up at

 Matches to be Scheduled

Newcastle after injury (29 th April

 Each league plays a double round robin 2006)

tournament. Therefore, there are no matches

played between different divisions, but…..

Pairing System

 Teams are paired which means they cannot

normally play at home on the same day



 Typically this is due to teams which are

geographically close (e.g. Manchester

United/Manchester City,

Sunderland/Newcastle,

Liverpool/Everton/Tranmere)



 Note that the pairing system operates across

Michael Owen returns to the line up at

division boundaries

Newcastle after injury (29 th April

2006)

 We are allowed a given number of pairing

violations

Season BD NYD

Pairing System

2005-2006 12 14

 Teams are paired which means they cannot

normally play at home on the same day 2004-2005 10 10

 Typically this is due to teams which are

2003-2004 8 14

geographically close (e.g. Manchester

United/Manchester City,

Sunderland/Newcastle, 2002-2003 10 8

Liverpool/Everton/Tranmere)

Number of actual clashes from

 Note that the pairing system operates across previous seasons

division boundaries

Key: BD = Boxing Day

 We are allowed a given number of pairing NYD = New Years Day

violations

Geographic Constraints

 There are other geographical constraints

which must be adhered to



 The number of London based clubs that can

play at home on the same day = 6



 The number of London based Premier clubs Chelsea win the Premiership for the

second consecutive season (April

that can play at home on the same day = 3

2006)

 The number of Manchester based clubs that

can play at home on a the same day = 4

Date Specific Constraints

 On New Years Day and Boxing Day each team

is required to play



 If a team plays at home on Boxing Day they

must play away on New Years Day



 The same teams cannot play each other on Wayne Rooney makes it 4-0 - giving

Manchester United the biggest League

New Years Day and Boxing Day

Cup Final victory in the competition's

history (2006)

 Paired teams cannot play each other

 In fact this is violated in the published

fixtures, but we treat it as a hard constraint



 The aim is to minimise the

distance travelled

Date Specific Constraints

 On New Years Day and Boxing Day each team

is required to play



 If a team plays at home on Boxing it must play • Analysis of previous

away on New Years Day

seasons fixtures and

meetings with the Football

 The same teams cannot play each other on

League have shown that

New Years Day and Boxing Day this is the case

 Paired teams cannot play each other • Not like TSP or TTP as we

 In fact this is violated in the published cannot minimise the

fixtures, but we treat it as a hard constraint overall distance travelled

(throughout the entire

season)

 The aim is to minimise the

distance travelled

Premier League Teams (20 teams)

Champion League Teams (24 teams)

League 1 Teams (24 teams)

League 2 Teams (24 teams)

Solution Methodology

 Two-stage algorithm

Depth-first search

Local search









Bobby Charlton: Then and Now

Solution Methodology

 Two-stage algorithm • Produce fixture lists for each

Depth-first search division for these two days



Local search • Pick a team at random and

generate all possible fixtures

(root node)

• Expansion is done by

producing all possible

(feasible) fixtures (e.g. a

team cannot play twice in

any one branch of the search

tree)

Solution Methodology Depth-first search

 Two-stage algorithm • Do not expand if distance

Depth-first search between teams is > 200

miles (expect Plymouth)

Local search

• Do not expand if total travel

distance > best solution

found so far

• Do not allow paired teams to

play one another

• Do not allow paired teams to

both be playing at home

Solution Methodology Depth-first search

 Two-stage algorithm • Follow same procedure to

Depth-first search generate for New Years Day,

but…

Local search

• Relax condition that paired

teams cannot both be at

home (else no feasible

schedules are found)

• Teams cannot play each

other again (from Boxing

Day)

• If a team played at home on

Boxing Day, they must play

away on New Years Day

(and thus vice versa)

Solution Methodology Depth-first search

 Two-stage algorithm • At the end of depth-first

Depth-first search search we have two

complete fixtures for each

Local search division

• They are improvements over

the published fixtures (wrt to

overall distance) but they

violate some of the

constraints

Solution Methodology

 Two-stage algorithm • The aim is to reduce the

Depth-first search number of constraint

violations to acceptable

Local search levels. That is, less than, or

equal to, the published

fixtures

Solution Methodology Local search

 Two-stage algorithm

Swap Operation

Depth-first search

• Our local search makes use

Local search of the following swap

procedure



… … … …



Chelsea Portsmouth Arsenal Chelsea







Portsmouth Fulham









f f′

Solution Methodology Local search

 Two-stage algorithm

Swap Operation

Depth-first search

• Our local search makes use

Local search of the following swap

procedure



… … … …



Portsmouth Chelsea Arsenal Portsmouth







Chelsea Fulham









f f′

Solution Methodology Local search

Iterate (5000)

 Two-stage algorithm 1. If pair clash in f then

Depth-first search swap

Local search 2. As 1, but for f′

3. Check London clubs in f,

and swap if necessary

4. Check GMR clubs in f,

and swap if necessary

5. Check London Prem

clubs in f and swap if

necessary

6. As 3 but for f′

7. As 4 but for f′

8. As 5 but for f′

Results

Published Generated

Fixtures Fixtures



Total 4295

3049

Boxing Day 1 Distance



Season 2005- Clashes 12 10

2006 Total 6336

New Years 3868

Distance

Day 2

Clashes 14 10

Total 4563

3725

Boxing Day Distance



Season 2004- Clashes 10 10

2005

Nottingham Forest Football

Total 5122

New Years Distance

3059 Ground

Day

Clashes 10 10

Total 3872

2859

Boxing Day Distance



Season 2003- Clashes 8 8

2004 Total 4296

New Years 3500

Distance

Day 3

Clashes 14 14

Total 3827

2839

Boxing Day Distance



Season 2002- Clashes 10 10

2003

Results

Season Total Total Distance Number of Number of Pair

Distance Distance Saving Pair Clashes Clashes for

for for for Published Generated

Published Generated Fixtures Fixtures

Fixtures Fixtures



2005-2006 10631 6917 3714 26 20

(34.94%)

2004-2005 9685 6784 2901 20 20

(29.95%)

2003-2004 8168 6359 1809 22 22

(22.15%)

2002-2003 7791 6040 1751 18 18

(22.47%)



Bobby Moore in

1966

Planned Further Work

 Try to find the optimal solutions



 Tighten/change the bounds on which we

branch to improve the search time



 Add heuristics to the branching so that less

expansions are necessary



 Produce a 4-fixture schedule









Peter Osgood (1947-2006)

Future Directions

 Produce schedules for a complete season



 Hold additional meetings with the Football

League

 We have met with them once and they

provided lots of information for 2005-2006



 Investigate other leagues and other sports Gary Lineker scoring against

Paraguay in 1986

Summary

 New sports scheduling problem has been

introduced



 Lots of data available The new Wembley Stadium



 Many possible research directions



 Above all, a very interesting domian

Questions/Discussion









“They think it’s all over. It is now!”



Related docs
Other docs by ajizai
Fall 2010
Views: 0  |  Downloads: 0
Math 111
Views: 0  |  Downloads: 0
Training_listing_275360_7
Views: 1  |  Downloads: 0
C4-051739
Views: 0  |  Downloads: 0
DEFINITIONS
Views: 0  |  Downloads: 0
Unit POPULATIONS
Views: 0  |  Downloads: 0
albhed
Views: 0  |  Downloads: 0
price_list
Views: 9  |  Downloads: 0
By registering with docstoc.com you agree to our
privacy policy

You are almost ready to download!

You are almost ready to download!