

The Mystery of Points Per Million

To kick off a series on FPL misconceptions (I hope will be useful for 20/21), first I'll cover "points per million" & the problems with how it's used by the community



Firstly, what is points per million (ppm)
?
A player's ppm is just the points they have accrued
divided by their price
. On the FPL website it is called "Value".
For example, in 19/20
KDB cost £9.5m & scored 251 pts. Therefore his ppm was 251/9.5 = 26.4 (1dp) 


.

A player's ppm is just the points they have accrued


For example, in 19/20





The idea is that this will give you a picture of a player's value for money, allowing you to compare players across different price ranges
.
Then you will know whether spending £9.5m on KDB is "worth it", or whether you would be better off spreading the funds elsewhere

.

Then you will know whether spending £9.5m on KDB is "worth it", or whether you would be better off spreading the funds elsewhere



By this logic
, the best value players from last season (by end price) were:
32.7
Pope
31.3
Lundstram
30.8
D Henderson
28.9
Patricio
28.7
Ryan
Hang on! There are 4 GKs here
. Does that mean GK is the most important? Obviously not. So what's really going on? 

32.7

31.3

30.8

28.9

28.7

Hang on! There are 4 GKs here



Well, there are multiple flawed ideas at play here
.
Firstly, ppm is a completely meaningless number in terms of how the game actually works.
You have to pick exactly 5 MIDs & have a total budget of £100m. What if every MID cost £2.0m less & your budget was £90m?





You have to pick exactly 5 MIDs & have a total budget of £100m. What if every MID cost £2.0m less & your budget was £90m?




Then the game would be exactly the same
! Say you spent £Xm on non-MIDs & £(100-X)m on MIDs. Then in the new game you can still spend £Xm on the same non-MIDs & £(100-X-10)m on the same MIDs as they each cost £2m less (so £10m less overall).
Nothing changes at all
...

Nothing changes at all


...except ppm. You guessed it
, even though the game is exactly the same, suddenly the ppms for MIDs are drastically different
.
Now the best value players are:
44.4
Cantwell
41.7
Dendoncker
39.1
Douglas Luiz
38.9
Rice
38.2
Grealish
Ppm does not make sense
.



44.4

41.7

39.1

38.9

38.2

Ppm does not make sense




We've done the same to everyone & it's still the same game


12.2

11.9

11.3

11.2

10.8

There is a more even spread which looks completely different to the 4 GKs

We can even make prices 0 or negative & keep it the same game but ppm does not make sense.


Is there a better way to think about player value
? In this thread you can see an example of figuring out players' "true values" by subtracting the minimum prices of players in that position (as you are required to fill every position anyway)
:
https://twitter.com/FPLRatchet/status/1290257728674377730?s=19



https://twitter.com/FPLRatchet/status/1290257728674377730?s=19


This is a reasonable start. A popular next step is to work out a minimum number of pts to associate with each position's cheapest player
. Then subtract that from a player's points to get Effective Points (EP)
, i.e. how many more they get you than the cheapest player. 




For example,
Rashford was £8.5m & scored 177 pts. Greenwood was £4.5m (the cheapest) & scored 103 pts
.
So if you had Rashford then you paid £4.0m more to get 74pts more
. The value was 74/4 = 18.5
.
Is this fixed then? Let's look at the cheapest in each position
...


So if you had Rashford then you paid £4.0m more to get 74pts more


Is this fixed then? Let's look at the cheapest in each position


These were the highest scoring
:
GK
20
Martin
11
McGovern
DEF
144
Lundstram
84
Rico
64
Kelly
51
B Williams
26
Lamptey
MID
111
Cantwell
100
Dendoncker
90
Douglas Luiz
73
Hayden
71
Saka
FWD
103
Greenwood
61
Connolly
39
Martinelli




20

11


144

84

64

51

26


111

100

90

73

71


103

61

39




I've included enough to fill a team because that's essentially what's happening
. We start with a base team for £64m (8×4.5m + 7×4m), & see how many more points we get by adding in different players
. So in order to calculate EP (& EPpm), we'll take an average by position
.



The averages are:
GK 15.5
DEF 73.8
MID 89.0
FWD 67.7
Now we can rate more expensive players
by their "EPpm" calculated by:
EPpm = (pts - average)/(price - min)
(min = 4 or 4.5 based on position)
Note EPpm is not affected by changing prices how we did before!




Now we can rate more expensive players

EPpm = (pts - average)/(price - min)
(min = 4 or 4.5 based on position)
Note EPpm is not affected by changing prices how we did before!


Let's see the top EPpms (by start price)
:
GK
309
Pope
289
Henderson
DEF
136
Baldock
118
Egan
108
Dunk
98
Basham
92
Soyuncu
MID
82
Traore
40
Grealish
37
Martial
32.4
De Bruyne
32 Mount/Shelvey/Rice/Noble
FWD
129
Ayew
87
Ings
42
Maupay


309

289


136

118

108

98

92


82

40

37

32.4

32 Mount/Shelvey/Rice/Noble

129

87

42


Now we create our wonder team, + bench/C/VC:
Pope 170 (v +22)
Lundstram 144
Baldock 142
Egan 133
KDB 251 © (+251)
Martial 200
Grealish 149
Mount 137
Ings 198
Ayew 132
Maupay 131
-
Martin (0)
Traore (55)
Dunk (38)
Soyuncu (8)
We've amassed 2161pts
, surely winning... oh 

Pope 170 (v +22)
Lundstram 144
Baldock 142
Egan 133
KDB 251 © (+251)
Martial 200
Grealish 149
Mount 137
Ings 198
Ayew 132
Maupay 131
-
Martin (0)
Traore (55)
Dunk (38)
Soyuncu (8)
We've amassed 2161pts




Above courtesy of @EricDFreeman & @FPL_Defender
:
https://twitter.com/EricDFreeman/status/1287518781166714880?s=19
https://twitter.com/FPL_Defender/status/1287690118028046337?s=19
What happened? Well, we only spent £81.5m
. Despite the "fix", cheap players still dominate EPpm
. This exposes a 2nd flaw:
Points do not increase linearly with price.

https://twitter.com/EricDFreeman/status/1287518781166714880?s=19
https://twitter.com/FPL_Defender/status/1287690118028046337?s=19
What happened? Well, we only spent £81.5m





The fundamental idea
in ppm is points increase linearly with price. No matter when, if you ever do points
price & expect a constant K, then you are drawing a line on the points vs price chart with gradient K
.
Top players from the last 2 seasons show this flaw below
:



Top players from the last 2 seasons show this flaw below


Price per million is hard to fix because it relies on this flaw, and that's before we even get on to captaincy & bench players
.
So how can you know which players represent the best value
?
I would argue that it might be helpful to think about it differently
:

So how can you know which players represent the best value


I would argue that it might be helpful to think about it differently


The best value 1-2 players per position are usually easy to spot anyway
- you don't need a complex metric to put Pope, TAA, Lundstram, KDB, Martial, Ings, Jimenez in your team
.
To fill the rest, I recommend first creating a gut feel team. You have to start somewhere!


To fill the rest, I recommend first creating a gut feel team. You have to start somewhere!

Then take groups of 2-3 players and try replacing them with different combinations
. Comparing these will truly tell you which potential team is best, instead of estimating with some fancy metric 
.
For example, if you want to know whether Aubameyang is worth £11m

:



For example, if you want to know whether Aubameyang is worth £11m



Simply compare a combination of Auba + 1-2 others with another combo that spreads your money differently 
.
Worried that a couple of better MIDs will give better value
? Just test it! Get a conclusive answer rather than a guess 
.
This idea is explained well here
:


Worried that a couple of better MIDs will give better value



This idea is explained well here


https://twitter.com/FPL__Raptor/status/1286615499376340998?s=19
thread by @FPL__Raptor 
If you don't know where to begin (but you do use FPL twitter
), by all means check out differential & value threads 
- just don't use ppm alone as it really does not make sense.
Thanks for reading!
Kiwi out



If you don't know where to begin (but you do use FPL twitter



Thanks for reading!

Kiwi out



PS:
With this included, this is the longest thread you can tweet in one go (25 tweets)
!
Some accounts who (based on a quick search) might find this interesting
: @FactsFpl @Prem_Tipster @FPLChurch @allaboutfpl @FFH_HQ @FPLTIPZ @fplreview
Feel free to discuss below
...
With this included, this is the longest thread you can tweet in one go (25 tweets)

Some accounts who (based on a quick search) might find this interesting

Feel free to discuss below


This has had such an amazing response - thank you to everyone in the #FPLCommunity for your kind words
, likes
, RTs
and mostly just bothering to read it at all 

!
I've been blown away by the support
for me, for this, and for each other
. Thank you everyone!






I've been blown away by the support


