Parity synergy pre-game analysis: league 25, week 3
A synergy rating for each roster, built from every game in our store and
the people on these rosters.
This is a live report: it is regenerated as games are played and as
rosters change.
Players are rated on their own store history, not on their current
roster alone. A team's roster can change from game to game, so every number here is about
the people on the roster and the history they carry, never about a team as an entity.
The teams
One card per roster, ordered by rating. Each card carries the sample size behind
the number: pairings, shared points, and shared games.
Set Phasers to Huck (Nick)
+193.2
+37.8 since last game
The roster changed since the last report, so this figure is for a different group of people than the one it is compared against.
Small sample: 1715 points, 34% on <5-game pairings.
29 pairings1715 shared points61 shared games11 of 12 players in our data9 contributing
Best pairing
Liam Paquette with
Simon Krol +$4,586.53
combined excess per shared point over 7 shared points
Weakest pairing
Ariel Untiveros with
Ryan Lindsay -$2,632.71
combined excess per shared point over 17 shared points
Each row: how that team's players earn per point against each opponent,
compared with their usual.
Read a row left to right; every cell carries the number of points
behind it.
earn more than usual
earn less than usual
Read one row left to right: Nick's players vs Heath's: +135 ·Nick's players vs Gabe's: -57 ·Nick's players vs Al's: +64 …
How this was made
The rating measures how much a roster's players earn per point when they are on a
line together, compared with what the same players earn per point across all their points. It
counts a player's whole salary, so it moves with everything they earn: the goals and assists they
make, the passes they complete, and the bonuses every player on a line shares. A player who earns
more per point with a particular teammate than alone pushes their own rating up.
Each player's excess with each teammate is averaged over their teammates, weighted by how
much history each pairing has. The team rating is the mean over its players, so a player counts
once no matter how many teammates they have, and thin pairings count for less. A pair needs at
least 5 shared points to count.
A positive rating means the roster earns more per point with each other than the same
players would alone. A rating near zero means the players neither lift each other nor
drag each other down, which is a real result and not a lack of one: a roster can have a lot of
shared history and still be flat.
A rating on little history is not a rating close to zero. The rosters with the most shared
history sit closest to zero and the thin ones hold the extremes in both directions, which is
why every card carries its sample size.
This is a live report. Each rating covers every game in our store, so it moves as the
season is played. The change since the last game is the difference from the previous report's
figure. When a roster has changed, that comparison spans two different groups of people, and
the card says so.
Correlation with point differential: -0.187
Correlation with amount of shared history: -0.324
The median game margin was 4 points
Games decided by 3 points or fewer: 159 of 342
Shared games behind these rosters' ratings: 29 to 132
Store: 342 games, 13749 points, 342 players
What these numbers do not say
A player's team is a property of a game, not of a person. Teams trade, and players change
teams mid-season, so no number here is a team's record or a head-to-head history between two
teams as entities.
Players with no history in our data may well have played in leagues our store does not
cover. We report that gap rather than guessing at a player's background.
Pairings and shared games are sample sizes. A rating on few pairings rests on less
evidence than a rating on many, and every card says which it is.
A rating is not a record of who has played well. It says how much a group of people
earn per point when they are on a line together, over what the same people earn alone.
Set Phasers to Huck (Nick)
+193.2
+37.8 since last game
The roster changed since the last report, so this figure is for a different group of people than the one it is compared against.
Trades since the last report
in: Camille Normandeauin: Nicole Rideoutin: Joy Robertin: Simon Krolin: Liam Paquetteout: Adrienne Junekout: Sarah Thompsonout: Tiger Yeout: Travis Davidson
Small sample: 1715 points, 34% on <5-game pairings.
29 pairings1715 shared points61 shared games11 of 12 players in our data9 contributing
Best pairing
Liam Paquette with
Simon Krol +$4,586.53
combined excess per shared point over 7 shared points
Weakest pairing
Ariel Untiveros with
Ryan Lindsay -$2,632.71
combined excess per shared point over 17 shared points