How we measure a career
This is our methodology for measuring sustained career excellence. It is a considered model, not objective truth — and every input is on screen so you can disagree with it precisely.
Three brilliant projects vs a hundred good ones
- Adjusted quality
- 7.84
- Career depth
- Solid (34)
- Confidence
- Low
- Adjusted quality
- 8.60
- Career depth
- Elite (100)
- Confidence
- Very high
Creative B leads by 17.1 points despite the lower raw average. Three data points get pulled toward the population mean of 7.4; a hundred do not. Depth and longevity add more evidence on top. Push Creative B's raw average down toward 6.5 and watch volume stop helping.
Why raw averages mislead
An average treats one credit and one hundred credits as equally informative. It rewards the person who worked three times and stopped, and punishes anyone who kept working through a fallow year. We still show the raw average — we just refuse to rank on it alone.
Why sample size matters
Small filmographies are regressed toward a population mean of 7.4 with a prior worth 8 weighted credits. A four-credit career keeps most of its prior; an eighty-credit career keeps almost none. This is the Bayesian-style shrinkage step.
Why career depth matters — with limits
Depth grows logarithmically. Going from 3 to 20 meaningful credits moves it a lot, 20 to 50 still matters, 50 to 100 matters less, and 100 to 150 barely registers. Depth is capped at 16% of the score, so a hundred mediocre credits can never outrank twenty exceptional ones.
Why consistency matters
8.4 / 8.6 / 8.7 / 8.5 / 8.8 and 9.8 / 4.2 / 9.4 / 5.1 / 9.6 average out the same and describe completely different careers. We measure the weighted standard deviation and reward the floor, not just the ceiling.
Why meaningful credits matter
Credits are weighted by significance, and anything at or above 0.4 counts as meaningful. A 90-second cameo in a masterpiece contributes almost nothing. Episodic television and nominal executive-producer credits are discounted so no single series or title farm can manufacture a career.
How Lift works
Lift = performance score − blended project score. A 9.1 performance in a 6.3 film is +2.8: they carried it. A 6.7 performance in a 9.0 film is −2.3: great movie, weaker performance. Lift is an actor-only metric — we will not fake an equivalent for directors, writers or producers until we have one that means something.
Credit weighting
Actor
A 90-second cameo in a masterpiece barely moves a career score. Leads carry full weight; minor parts and cameos are heavily discounted.
Director
A primary directing credit counts in full. Episodic television is down-weighted so one series cannot manufacture dozens of equivalent 'projects'.
Writer
Created-by and sole screenplay credits count in full. Story credits, shared credits and additional-writing passes are discounted.
Producer
The hardest category to measure. Hands-on producer credits count fully; nominal executive-producer credits are heavily discounted so volume alone never wins.
How rating sources are normalised
WUZIT! uses two rating sources: TMDB user ratings and WUZIT! member ratings. Each is shown as itself, on a 0–10 scale, labelled with its own name and vote count. They are never blended into a combined Project Score, and no other provider's rating is estimated, inferred or substituted anywhere in the product.
No restricted site is scraped and no rating is ever invented. If a source has no value for a title, WUZIT! shows “—” instead of a placeholder number.
Career score composition
Animation and voice acting
Animation is a production format, not a genre and not a niche. Every project carries a content format (live action, animation, hybrid) and, where animated, one or more animation types — 2D, CGI, stop motion, anime, adult and children's animation. Every feature in the product works the same way for animated titles: project score, prediction, creative confidence, trust match, box office, rankings and saved searches.
Performances are typed too: live action, voice, motion capture, self and other. These records are kept apart. A voice performance and an on-camera performance are different crafts, so a career carries several acting scores, each with its own sample size and confidence rather than one blended acting number.
Voice work is broken down further by context — animation, adult animation, children's animation, anime and dubbing — and by genre, with the same shrinkage rules used elsewhere. Original-cast and dubbed / localised performances are tracked separately so a dubbing record never inflates an original-voice one, and the language of each performance is retained.
This changes prediction and creative confidence. On an animated project, an actor with a substantial voice record is scored on that record, not their live-action career score; directors, writers and producers are likewise scored on their animation work where enough of it exists. The breakdown says which body of work each number came from.
The underlying model is person → profession / career → credit type → performance type → project context → genre → score, each level carrying its own sample size and confidence. No one is reduced to a single acting number.
How the prediction score works
An unreleased title has no ratings, so it gets no project score. Instead it gets a prediction score: a weighted read of every known team member's historical record, measured in the genre this project actually is, not their career average.
- Each person contributes their genre-relevant score, shrunk toward the population mean when they have few relevant credits. Recency matters — recent work counts for more than a decade-old peak.
- Contributions are weighted by seat: director heaviest, then writer, then leads and producers. These weights are adjustable priors, not truths — the accuracy page exists to correct them.
- Small bonuses apply for several elite creatives appearing together, and for teams who have made good work together before. Both are capped, because a strong list of names is not a film.
- Unknown seats never count as average. A missing director does not drag the score toward 5 — it lowers confidence instead, and is listed as a risk.
Prediction confidence is a product of three things: how much evidence the known people carry, how complete the team information is, and how certain the release date is. A high score with low confidence is a rumour, and is labelled as one.
Predictions are backtested: every released title is re-predicted with all credits from its release year onward hidden, then compared with what it actually scored. The engine runs systematically conservative — track records cannot foresee a masterpiece.
Watchlist, follow and interest signals are preferences, not ratings. They personalise what you are shown before release and are deliberately excluded from every score.