26
Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis Griffin MyFC Team Selector

Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Embed Size (px)

Citation preview

Page 1: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Feasibility studyFunctional requirementsProcesses and Calculations

Revision 3

Dave Twisleton-Ward (Davoloid)With special thanks to footy19 and Dr Lewis Griffin

MyFC Team Selector

Page 2: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

What are we looking for?

• What is the best formation for this game– Opposition– Upcoming games (easier/harder)

• Who is best qualified to fill this position– Injuries– Form– Skills

Page 3: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Wisdom of Crowds

• Average choice should be correct more often than single expert

• Decisions can be made more quicklyconditional on– Diversity of members – Freedom of vote– Access to information – Working for the same common goal– No bias caused by influence of experts

Page 4: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Final Goal

• To produce a team lineup and formation that Liam and the Coaching staff can use

• Requires: – Access to information about team

• Problem of database & interface design • Not this document

– A method for calculating the choices of many members

Page 5: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

My Previous examples

• Complicated to explain• Complex calculation• Tied into increasingly complex interface• Brute force matching not even discussed past

initial thought

• Great flexibility and granularity (I think)• Could be relevant for future developments

Page 6: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Realistically

• Most people want:– Formation– Players

Anything more complicated becomes difficult to synchronise with training regimes and implementation on matchday.

Page 7: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Basic principle

1 2 3 4 5 6 7 8 9 10 11

Chris McPheeNeil BarrettMark RickettsLiam ColemanGeorge PurcellLuke MooreJohn AkindeSam MottMichael BostwickRonnie BullRob French

Lance CroninPeter HawkinsSacha OpinelGary MacDonaldJames SmithPaul McCarthy (captain)Danny SlatterStacy LongRaphael NadeChukki EribenneCharley Hearn (on loan from Grays Athletic)

Who are the best 11 players for a particular match?

Page 8: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Aggregating selections

• We each have a different opinion on who is best for each box. • Calculation of most common choice: Box 1…Box11• Produces a player for each position• This is a simple calculation

1 2 3 4 5 6 7 8 9 10 11Cronin Opinel Hawkins Ricketts Charles Bostwick McPhee Barrett Purcell Long Moore

This is the starting 11, now where do they go?

Page 9: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

FormationsWe need to agree where each slot is for each formation.

1

2 3 4

5 7 8 96

10 11

1

2 3 4

5 8

7

96

10 11

4-5-13-5-2

1

2 3 4

5 8

7

96

10

11

4-4-2

Page 10: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Selection IDAny selection can be described as:

Formation choice + 11 Players in Slots = Selection ID (SID)

F 1 2 3 4 5 6 7 8 9 10 113-5-2 Cronin Opinel Hawkins Ricketts Charles Bostwick McPhee Barrett Purcell Long Moore

Each player has a unique 2-digit number: PID

Each formation has a unique 2-digit number: FID

F 1 2 3 4 5 6 7 8 9 10 113-5-2 Cronin Opinel Hawkins Ricketts Charles Bostwick McPhee Barrett Purcell Long Moore

02 01 03 02 15 24 25 12 14 17 08 18

So we have a discrete 24 digit SID for each different formation:

In this example: 020103021524251214170818

Page 11: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Using SIDE.g. 020103021524251214170818

Gives us up to 99 different formations, and 99 players

•99 Formations IDs gives us the ability to distinguish between 4-4-2 defensive and 4-4-2 attacking, for example.

•99 Player ID’s allows us to keep historical records, probably up to 5 years?

We reserve 00 for when a formation is blank, or a player is not selected or becomes unavailable.

There may be algorithms that can “compress” this value to save storage / memory during processing.

Page 12: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Benefits of SID

• Simple method of storing and retrieving selections• Readable by humans• Selection process can now be done client side, sending

only a text string to the team selection server.• SID is the same whatever interface is used

– E.g. Flash, HTML drop down box• Submission via email or text message?• Increases ease of changing interface,• Allows analysis and export of data.

Page 13: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Determining best SID

• As members make their selections the SID is generated and added to a table.

• If someone else makes the same SID (which we would expect) the count is increased.

• We end up with a table of SIDs along with how many people have selected each one.

• We should expect that many people will choose the same formation and lineup.

Page 14: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Table of SIDsSID Count02 01 03 02 04 11 05 06 08 07 09 10 25002 01 03 02 04 05 06 07 09 08 10 11 21402 01 03 02 04 11 05 06 07 08 09 10 19001 01 02 03 04 05 06 07 08 09 10 11 9002 01 03 02 04 11 05 06 08 07 09 10 7002 01 03 02 04 05 06 07 09 08 10 11 3302 01 02 03 04 05 06 07 08 09 10 11 2202 01 02 03 04 05 06 07 09 08 10 11 2102 01 03 02 04 05 06 07 08 09 10 11 1902 01 04 03 02 05 06 07 08 09 10 11 1702 19 02 03 04 11 05 06 07 08 09 10 1502 19 04 03 02 05 06 07 08 09 10 11 1302 19 04 03 02 05 06 07 08 10 09 11 1103 01 02 03 04 05 06 07 08 09 10 11 403 01 02 04 03 11 05 06 07 08 09 10 403 01 02 02 04 05 06 07 08 09 10 11 403 01 03 02 04 05 06 07 09 08 10 11 406 01 03 02 04 05 06 07 09 08 10 11 303 19 03 02 04 11 05 06 08 07 09 10 304 01 02 03 04 05 06 07 08 09 10 11 207 01 03 02 04 05 06 07 09 08 10 11 203 01 02 04 03 11 05 06 07 08 09 10 103 01 02 02 04 05 06 07 08 09 10 11 104 01 02 04 03 11 05 06 07 08 09 10 104 01 02 03 04 05 06 07 08 09 10 11 104 01 02 04 03 11 05 06 07 08 09 10 102 01 03 02 04 05 06 07 08 09 10 11 102 01 03 02 04 11 05 06 07 08 09 10 104 01 04 03 02 05 06 07 08 09 10 11 105 19 02 03 04 05 06 07 08 09 10 11 1

We might expect a normal distribution, with less deviation (blue line) as familiarity with players and access to information improves.

Page 15: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

A problem

But all we’ve done now is picked the most common selection (modal average).

Is this satisfactory?

Maybe…

• The variation between the top 3 may be so small that we can present 3 options to Liam

What about the rest of the selections?

We can take them into account with further calculations.

Page 16: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

The “real” best FormationBefore we determine this we have to work out the difference between formations.

The slots are in a physically different position, if only slightly. The tactical context of that position has also changed.

We ask the question: How confident would we be that a player in 4-4-2, position 3 would be the right choice for 3-5-2, position 3?

So let’s give it a confidence, c, where 0 would be no match, 1 would be perfect match.

We would probably give slot 1 (the keeper) a confidence of 1 as this position never changes.

Page 17: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Not convinced?Look how much players move when you change formation in the

current team selector:

Page 18: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Difference between Formations

1

2 3 4

5 7 8 96

10 11

1

2 3 4

5 8

7

96

10 11

4-4-2

We know that there’s a good match, but what is required from a player in each slot is slightly different for each formation.

3-5-2Slot 3 is reasonably close, so we’ll give it a confidence of 0.8

Slot 7 is not as close, so we have a lower confidence, say 0.5

The distance between two slots is 1-c

We add up the distance for all 11 slots for this pair of FIDs.

Page 19: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Map of formations

4-4-2

5-4-14-5-1

4-3-3

3-5-2

We do this for each pair of formation IDs. This would be in consultation with Liam as it has to reflect the training regime and positioning.

Once done, no need to go back.

We now know the distance between each pair of FIDs, and can create a map. (see PublicWhip.org.uk)

If we need more tactical variation, e.g. 4-4-2 with offensive wingbacks?

Simply create a new formation ID based on the previous 4-4-2. (this is why we need 99 formation ID’s)

Only a slight adjustment to the distances for each pairing is needed, and can be verified with Liam.

Page 20: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Mapping the playersWe now work out the least distant FID taking into account how many people have voted for each.

(We may find that the most common FID was actually good enough.)

Now we’ve found the “best” formation ID, we can perform the basic “best 11” calculation we began with:

•We map the players from the losing formation IDs using the confidence values we’ve established earlier.

•We then total up the votes for each player in each slot as well as the mapped votes.

•We don’t count players that have become unavailable, but we still keep the original SID data for reference

Page 21: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Example mapping

In the following example, 4-4-2 has been established as the best formation.

We map the losing 3-5-2 selections to 4-4-2 so

that we can count them and work out the starting 11.

We do this for all SIDs with a losing FID. (e.g. 4-5-1, 4-3-3…)

Page 22: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Example mappingWhat we’ve said here is that 200 people voted for this SID. Because we’re sure Cronin will be fine in goal in 4-4-2, our confidence was 1, so we count this as 200 votes for Cronin in slot 1.

Barrett, on the other hand is in a very different position in slot 7 so we count only 140 votes for him in this slot under 4-4-2.

F 1 2 3 4 5 6 7 8 9 10 113-5-2 Cronin Opinel Hawkins Ricketts Charles Bostwick McPhee Barrett Purcell Long Moore

02 01 03 02 15 24 25 12 14 17 08 18

Conf 1 0.8 0.8 0.8 0.5 0.8 0.8 0.7 0.7 0.9 0.9

# SIDs 200 200 200 200 200 200 200 200 200 200 200

Votes 200 160 160 160 100 160 160 140 140 180 180

Page 23: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

End result

• We might find that the second or third most popular SID might be the best.

• But remember that there would probably be very little difference between the top 3 selections.

• Now we have taken into account everybody’s opinion, which is more likely to be correct. (according to Wisdom of Crowds theory).

• Because the calculations are so basic, this would not take long.

Page 24: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Further uses

• Because the selection ID is just a numerical string, these can be added to a members profile (for their own viewing or for audit purposes).

• Or if this was kept in a database, member numbers just need to be added to each SID.

• We might want to do this to examine trends: – Who is more likely to make the best choice, people

who attend the matches or people who watch on TV?– Is this user consistently making “malicious” selections

to sabotage the vote?

Page 25: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Summary• Selection ID = Numerical string• SID = Formation ID+Slot1+Slot2…Slot11• Physical location of each Formation ID

determined with Liam• Confidence between slots in different formations

also determined with Liam• Simple storage and calculation• Uses all contributed selections• Interface agnostic

Page 26: Feasibility study Functional requirements Processes and Calculations Revision 3 Dave Twisleton-Ward (Davoloid) With special thanks to footy19 and Dr Lewis

Development Process

• Iterative – start basic, build tactical formations (e.g. 4-4-2 offensive) as confidence in system improves

• Compare results after “tweaks” to confidence maps

• Usability testing with mockups – member opinions

• Need for good test data from previous matches!!!