Ratings through the latest IMDb dataFollow @screenballtv@screenballtvscreenballtv@gmail.com
ScreenBall

Glossary

Every term on the site, in plain words.

Rating
The IMDb user rating, 1 to 10. A season's rating is the mean of its episodes, weighted so that an episode with 40 votes counts less than one with 4,000. A newly aired episode's rating counts once it has 750 votes, or three days after it aired, whichever comes first; until then it shows as pending.
Lane
What a show should rate before any people are considered: its network, genre, half-hour or hour, season length, limited or ongoing, and franchise history. The lane is context. Nobody is credited for it.
Credit
A person's share of how far a season or episode beat or missed expectation, in rating points. Showrunners and billed cast are credited for the season. Writers and directors are credited for how far each episode differs from its own season. A show's credits add up to what it did beyond its lane. The models use credit as it is, over average; the site shows people as RAR (below).
Expected part and surprise share
Credit has two pieces. The expected part is what the person's record led the model to expect. The surprise share is their part of what nobody expected, split by seat and tilted toward people whose record on other shows agrees with it.
Billing
Cast are billed #1 to #8. Each slot carries a measured weight: #1 is 1.82, #2 1.10, #3 0.67, down to 0.055 at #8. A lead moves a season about 33 times as much as an eighth-billed regular. Actor records are kept per unit of billing weight and shown “as a lead”.
Record
A person's track record: the average of their credit with a 7-year half-life, so recent work counts more.
Kept factor
How much of a record the model keeps. It is 1 for long records across several shows and smaller for thin ones. A thin record is pulled toward zero, which is average. Seasons with few IMDb votes count as less than a full season.
Projection
Record × Kept, plus the replacement piece for one season or episode in the role, so it is in RAR. The piece is the same for everyone in a role, so the boards rank people exactly as the model does. The forecasts use record × Kept alone.
RAR (ratings above replacement)
How much a person added beyond what a replacement-level hire would have: their credit plus their replacement piece, in rating points. Every number about people on the site is RAR: leaders, the record book, show and person pages, projections. An average professional earns positive RAR just by working; someone below replacement is negative. RAR explained
Replacement level
What a season or episode made by freely available people would do: about 0.20 rating points below the zero of credit for a season and 0.04 for an episode (2026). Each season's gap is split among its showrunners and cast by their standard seat, and each episode's among its writers and directors, so working counts: an average professional on 200 episodes is worth more than nobody. The line averages two estimates, seasons made by people with no record and shows that ended after one or two seasons, smoothed over past years only. It is for display: the forecasts and the records use credit over average.
Legacy credit
When someone leaves a show, later seasons still owe part of their level to that person. Those rows are marked legacy. They can shrink to zero, never change sign, and never collect later good news.
Preseason forecast
Made on Dec 31 before a season airs, from data through that day only, then frozen. It is the only forecast that is graded.
Latest forecast
The same model with today's records. It moves as news and results arrive, and is never graded.
Baseline
The simple alternative every forecast is compared with. For a returning show: last season's rating plus the typical season-to-season change. For a new show: the lane alone.
Tier
Where a new show's forecast ranks among all past first-season forecasts, in five bands. Shown with the range that past shows in the same band actually landed in.