In my regular posts, I take a look at 2 samples: randomly selected 20,000 FPL teams and the top 10,000 FPL teams as of the start of a gameweek.

Because of the law of large numbers, we can make inferences about the overall FPL league based on the statistics for the random sample. Interval estimates for most statistics for the whole FPL game are also available. If you’re familiar with the concept of confidence intervals, you can point at a specific number characterising the random sample to see 99% confidence intervals for the respective number characterising all the 3 million FPL managers.

#### Contents of Regular Posts

#### 1. Statistics on Transfers and Team Selection

##### Introduction

It’s just a welcome statement, but it also includes a link to a google spreadsheet with full statistics. Complete lists are very long, so I don’t include them in my posts and limit tables to 5 players. *Hint: numbers in the spreadsheet are rounded, but if you click on any given cell, you can see the exact number in the bar at the top of the spreadsheet.*

##### Table ‘Transfers and Team Value’

- Number of managers. Number of managers in my samples: 20,000 FPL teams in the random sample, 10,000 teams in the top 10K.
- Average transfers made. Number of transfers made by an average team in the sample; wildcard transfers are excluded.
- Point hits taken. Percentage of teams in the sample which took point hits this week.
- Average points deducted (incl. no hits). Average points deducted for point hits from a team in the sample; the calculation includes teams without any point hits.
- Wildcards played. Percentage of teams in the sample which played their wildcards this week.
- Wildcard status. Shows the percentage of teams with their wildcards ‘available’, ‘played’ or ‘active’. The ‘active’ status is rare, but it’s possible if a wildcard was activated straight after the deadline (before I pull the data for this particular team).
- Average team value. Average team value for the sample, money in the bank is not included.
- Average money in the bank. This is clear.

##### Table ‘Transfers Made’

This table shows what percentage of teams in a sample made 0,1,2,3, or more transfers. Wildcards are accounted for separately.

##### Table ‘Point Hits Taken’

This table shows what percentage of teams in a sample took point hits and what the size of those point hits was.

##### Tables ‘Transfers In’ & ‘Transfers Out’

These tables show 5 players with the most transfers in and out and the number of different players bought and sold by fantasy managers in the samples. To create these tables, transfer pages on the FPL site are analysed. Wildcard transfers are included.

The complete lists are available in the google spreadsheets linked in the intoduction.

##### Table ‘Formations’

This table shows what formations were chosen at the start of a gameweek.

##### Table ‘Captaincy’

This table shows what captains were chosen at the start of a gameweek. I don’t include vice-captains because these data are usually pointless.

##### Tables ‘Starting goalkeepers’, ‘starting defenders’, ‘starting midfielders’, ‘starting forwards’, ‘benched players’

These tables show 5 most popular players for each position and the total number of different players picked by fantasy managers in the samples.

Complete lists are available in the google spreadsheets linked in the intoduction.

##### ‘Ownership and captaincy distribution for most captained players’

This slideshow shows ownership and captaincy distribution for the 10 most captained players in the random sample. Each slide depicts how a certain player’s ownership and captaincy are distributed over the overall FPL table. The random sample is divided into 20 subsamples based on the overall rank; ownership and captaincy rates are calculated for each subsample.

Robin van Persie is still the second most captained player, but this slideshow shows where all his owners and captainers are ranked.

#### 2. Intermediate Results

All results are **always** posted **with projected bonus points**. Bonus points are calculated according to the BPS. I usually wait until the ESP numbers appear – after this moment the BPS numbers are usually no longer reshuffled.

##### Introduction

It’s a welcome statement. I usually write a short summary of gameweek results here. Please note, that the FPL towers round their official average down.

##### Table ‘Summary’

- Number of managers. Number of managers in my samples: 20,000 FPL teams in the random sample, 10,000 teams in the top 10K.
- GW average score. Average points scored by the teams in the sample. All the scores are added and divided by the number of managers in the sample. Point hits taken are ignored. Please note, that the FPL towers round their official average down.
- Average points deducted for point hits. Average points deducted for point hits from a team in the sample; the calculation includes teams without any point hits. These are the same numbers that are included in the earlier post on transfers and team selection. To calculate an average gameweek score accounting for point hits taken, one would need to subtract this number from the GW average score above.
- Players played per team (out of 12). This number shows how many players in the starting 11 have already played some minutes this week and therefore can’t be automatically substituted at the end of the gameweek. Captains are counted twice as they effectively play twice. Hence, it is ‘out of 12’: 11 starting players + 1 captain once again.
- Captains played. This is a separate indicator for captains. It shows how many captains have already played, and how many are left to play or be substituted for vice-captains at the end of the gameweek.

##### ‘Gameweek Rank Projections’

This image shows the projected gameweek ranks for all the scores of the teams in my random sample. These projections have proved to be rather accurate. The number of gameweek transfers determines whether your actual gameweek rank will be closer to the upper bound of the interval or to the lower one. If you made no transfers, it will be close to the upper end.

Gameweek rank is usually a good indicator whether your GW score is good or bad. However, it doesn’t take into account point hits.

##### ‘Average points distribution’

This graph shows the distribution of average points over the overall FPL league. The random sample is divided into 20 subsamples based on the overall rank at the start of the gameweek; average score is calculated for each subsample. The grey line depicts the average score for each subsample. The blue line depicts cumulative scores: the 1st subsample is the first 5%, the 2nd subsample is the first 10%, the final subsample is the first 100%, i.e. the whole random sample.

Average points are rarely distributed evenly, therefore the overall average doesn’t make much sense for most fantasy managers. This graph a good indicator whether you can expect to get a green arrow or a red arrow. As a rule of thumb, to get a green arrow, one needs to score more points than the average score in the subsample where his overall rank falls into (but also keep in mind point hits taken). For the top 10K average, use the data from the summary table.

##### Table ‘Average points per team by source of points’

The interpretation of this table is rather straightforward. It shows the sources of the average score. In what proportions do the points for minutes played, goals, assists, bonus points, etc, add up to the average gameweek score?

##### Table ’10 players contributing the most to the average’

This table shows the 10 players who made the biggest impact on the gameweek average score. It is supposed to help us understand, how come the average is so high/low, i.e. by how many points certain players increase the average score.

##### Table ‘Average points per team by line in formation’

This table shows how many points on average come from a particular line in formation.

‘Points for captain’ shows average points scored by captains (doubled).

‘Points on the bench’ shows how many points have been scored by players currently on the bench (some of them may be subbed on at the end of the week).

##### Table ‘Average points by position’

This table shows average points scored by a player in a particular line in formation. These scores equal to the points scored by a particular line in formation divided by the number of players in this line. As written at the bottom of the table, it only accounts for players who have played positive minutes.

##### Table ‘Gameweek Hall of Fame’

This table provides the links to the best results of the week. ‘Maximum points’ links the team with the best result in the sample. ‘Best defense line’ links the team which has the most points scored by a goalkeeper and defenders. ‘Best bench’ links the team with the most points currently sitting on the bench (some of them may be subbed in later). ‘Bonus points magnet’ links the team with the most bonus points (captain’s BPs are doubled, bonus points on the bench are excluded).

##### Table ‘Gameweek Hall of Shame’

This is the opposite of the table above. ‘Cards magnet’ links the team with the most points deducted for red and yellow cards (bench players included).

#### 3. Final Results

The contents of the posts with final results are almost identical to the posts with intermediate results. They are also **always** posted **with projected bonus points**. Bonus points are calculated according to the BPS. I usually wait until the ESP numbers appear – this is the moment after which the BPS numbers are usually no longer reshuffled.

The only differences from the posts with intermediate results:

- final results are
**always**posted**allowing for automatic substitutions and vice-captains**(I do autosubs and VCs in my Excel file via a special macro)

- there are two additional statistics:

- Summary table. The ‘of Which: Vice-Captains‘ line shows in how many teams vice-captains have played instead of captains.
- Table ‘Average points per team by line in formation’. The ‘Points off the Bench‘ line shows how many points were scored by the players who were automatically subbed in (only autosubs, vice-captains not included here).