Two weeks down and Hull defenders are still the meta. If they keep this up, we might see the next Leicester City. Manchester City and Liverpool dominate the top 20 this week, with 8 City players and 5 from Liverpool.
Players still available:
Abdul Fatawu: F, rank 21, 49% rostered
Ross Barkley (?!): M, rank 37, 20% rostered
Malick Yalcouye: M, rank 43, 3% rostered
Anan Khalaili: D, rank 57, 22% rostered
Coming soon:
Custom league scoring
Mobile version
Updated roles
How the week 2 rankings compared to Fantrax
Note: I improve this model every week, so these metrics are just a snapshot of the current version. They will be outdated next week.
Fantrax projected 0 points on roughly 60% of the pool. Those rows are excluded on both sides, so this is a like for like comparison. This leaves 206 players to compare.
This is a subset of metrics that I rate my model on
Haul Rates
This is the chance of an individual player scoring 6+, 10+, 15+ or 20+ points on any given week. Fantrax doesn’t predict this so this is for my rankings only.
Each dot is a predicted starter that actually started (177/199, 89% of starters) because predicted non-starters don’t have haul rates. Reading across the graph is the chance each player was given of clearing that haul rate. Up is what he actually scored. The red line is the bar itself, so green cleared it and gray did not. Each panel is split into four columns of equal width, and every column shows what the board said against what happened.
Green thickens to the right in all four panels. A higher number meant a likelier haul. At 10+, the least likely column returned 14% and the third returned 61%.
Three cells read “got 100%”. They are 1 of 1, 1 of 1, and 2 of 2. A single green dot alone in a column is not a track record, and the counts are printed so you can see that.
Are the haul rates reliable?
The ranking works at every haul rate. Sort players by these numbers and the ones at the top really do score more. Start everyone rated above 60% for 6+ and it pays off 75% of the time. Above 40% for 10+, 60% of the time. But at 15+ the best cutoff worth using still fails 66% of the time, and at 20+ there was no one better than a 42% chance, because nobody deserves one.
Trust it to tell you who is likelier. Do not trust it to tell you who is likely.