The TollyPrema Score is a single number between 0 and 5 that represents the collective critical consensus on a Telugu film. It is not a straight average. It is a weighted average — and the weights are directly influenced by you, the community, through your votes on individual reviews.
Here is exactly how it works.
Step 1 — Normalising Critic Scores
Telugu film critics do not all use the same rating scale. Some give stars out of 5, some out of 10, some use letter grades, and some just write a verbal verdict. Before any aggregation can happen, every score must be converted to a common 0–5 scale.
Numeric ratings are scaled proportionally: round((rating / outOf) × 5, 1 decimal). So a 3.5/5 stays 3.5. A 7/10 becomes 3.5. A 3/4 becomes 3.75.
Verbal verdicts are mapped through a fixed lookup table curated from how Tollywood critics actually write:
| Verdict | Score | Verdict | Score |
|---|---|---|---|
| Blockbuster / Masterpiece | 4.5 | One Time Watch / Average | 2.8 |
| Outstanding / Excellent | 4.3 | Okay / Mixed Bag | 2.5 |
| Good / Clean Entertainer | 3.7 | Disappointing / Below Average | 1.8 |
| Worth a Watch / Enjoyable | 3.5 | Disaster / Flop / Avoid | 1.0 |
Some reviews contain no numeric score and no recognisable verbal verdict — only a qualitative write-up. For these, we use AI sentiment analysis to infer an approximate label. We track the source separately with a score_source flag (explicit vs sentiment) so users can always see how a score was derived.
Step 2 — The Weighted Average
Once all scores are on the same scale, we compute a weighted mean where each reviewer's score is multiplied by their accuracy multiplier:
TollyPrema Score = Σ(normalized_score × accuracy) / Σ(accuracy)
A reviewer with an accuracy multiplier of 2.0 has twice the influence on the final score as a reviewer at 1.0. A reviewer at 0.1 contributes at one-twentieth the weight. Their review still counts — it just counts less. New reviewers start at a neutral multiplier of ~1.05.
The Tier System.Reviewers are assigned to Tier 1 (established, high-credibility critics) or Tier 2 (newer or lower-volume critics). We compute the score using Tier 1 reviewers only. If a film has no Tier 1 reviews yet, we fall back to Tier 2. This prevents a film's score from being silently diluted by mixing critics of very different standing.
Step 3 — Your Vote Moves the Weight
On every review on TollyPrema, you will see a thumbs up and thumbs down button. These are not reactions to the review's writing quality — they are a signal about whether the critic's assessment matched what you experienced after watching the film.
Did the critic call it a 4/5 and you agreed? Thumbs up. Did they call it a 2/5 and you thought it was far better? Thumbs down.
Over time, as votes accumulate across all of a critic's reviews, we recompute their accuracy multiplier using a Bayesian-smoothed formula:
accuracy = 0.10 + ((agrees + 5) / (total_votes + 10)) × 1.90
The “+5” and “+10” are pseudo-counts — a statistical technique that prevents wild swings on small sample sizes. With zero votes, the formula evaluates to exactly 1.05 (neutral). As real votes accumulate, the multiplier converges: consistent community agreement pushes it toward 2.0; consistent disagreement pulls it toward 0.10.
This is a long-game system. A reviewer's multiplier is computed across all their votes across all movies — not per film. Their track record determines their weight in future aggregations.
What Voting Cannot Do
This is the important part. Voting is more constrained than it might appear.
Votes affect weight — not the score itself. When you vote on a review, you are influencing how much that reviewer's future scores are trusted. You are not changing the numerical score they gave. If a critic gave 3/5, that 3/5 is fixed. What changes is how much that 3/5 pulls on the weighted average.
You cannot target a single movie through voting. Votes are cast on reviews, not movies. A reviewer's accuracy multiplier is the aggregate of their votes across all movies they have ever reviewed. Downvoting every one of a critic's reviews for a single film you disliked does not isolate the effect to that film — it moves their weight globally, across the entire catalogue. This makes surgical manipulation of a single movie's score through voting extremely difficult.
One vote per review, per IP address. The voting system enforces one vote per (review, IP address) pair. You can change your vote from agree to disagree (or vice versa), but you cannot stack multiple votes from the same connection. Your IP is stored as a one-way SHA-256 hash — we can enforce the deduplication constraint without ever logging your actual IP.
Bayesian smoothing resists early manipulation. Because the formula adds 5 pseudo-agrees to 10 pseudo-total-votes before any real votes are counted, the first vote a reviewer receives does not dramatically move their multiplier. Influence is earned through sustained community signal across many reviews — not one viral incident where a fanbase piles on a review they disagreed with.
A downvoted reviewer cannot be silenced. Even a reviewer whose multiplier falls to the floor (0.10) still contributes to the aggregate. Their scores count at one-twentieth of the weight of a trusted reviewer — but they count. No vote total can remove a review from the calculation entirely. This is intentional: it prevents a majority from silencing a dissenting critic whose take may ultimately prove correct.
What This Looks Like in Practice
A critic who calls every film a masterpiece gets disagree-voted consistently. Over time their multiplier falls, and their inflated scores carry less and less weight. The community naturally corrects for chronic over-enthusiasts.
A critic known for nuanced takes that the community agrees with accumulates a high multiplier. When they call a film average, that assessment has more pull on the final score than an equivalent review from an unvalidated source.
A film's fanbase cannot flood one critic's review with downvotes to inflate the aggregate score. They would have to downvote every one of that critic's reviews across all films — which, even if sustained, only reduces the critic's weight rather than removing them from the calculation.
The Score Is a Living Number
The TollyPrema Score updates as new reviews are added and as the community votes on existing ones. It is recomputed dynamically from the underlying data — there is no cached or locked-in number. This means a film's score can move over time, and we believe that is correct behaviour: the critical consensus on a movie is not fixed on release day.
Our goal is a number that becomes more trustworthy the more the community engages with it — not a static star rating that reflects whoever reviewed it first.