Parasocial Charts Methodology
The Parasocial charts rank the live streaming ecosystem every week. This page explains what each chart measures, how the data is collected, and the rules that govern who appears. The charts are powered by Borealis, Parasocial’s creator intelligence platform.
Chart methodologies evolve over time to reflect shifting industry economics and consumer behavior. When the inputs or rules change, the change is announced with the edition it takes effect.
How creators are tracked
No registration is required to appear on any Parasocial chart. Borealis automatically tracks the streaming ecosystem above its detection thresholds, and any creator streaming at meaningful scale is measured whether or not they have ever heard of us.
Verified profiles are coming. Creators will soon be able to register with Avarithim to confirm their identity, claim their place on the charts, and ensure their accounts are tracked. Until then, every ranking on this page is generated the same way, from public measurement, for verified and unverified creators alike.
Newly discovered creators are held out of the charts until a human reviews them. A creator found automatically is observed immediately but does not rank until someone confirms they belong. This keeps discovery aggressive without letting the charts inherit whatever an automated sweep found overnight. A creator in this state is not being judged — they are in a queue.
The tracking week
The Streaming 100 and the Streaming Surge 50 cover a fixed weekly window: Monday 00:00:00 ET through Sunday 23:59:59 ET. The edition is computed early the following Monday and published that morning, Eastern time.
The boundary is Eastern time, not UTC, and it follows US daylight saving — so twice a year a tracking week runs 23 or 25 hours. We keep the local Monday boundary rather than a fixed offset because a creator’s week is a local week; a stream that starts Sunday evening belongs to the week a viewer would say it belongs to.
The Snapchat charts run on a different window. The Snap 100, Snap Global 100, Spotlight 100 and Spotlight Global 100 cover Saturday 00:00 UTC through Friday 24:00 UTC — a fixed UTC boundary that does not move with daylight saving. They publish Mondays at 9:00 AM ET. Their measurement is set out in full on the Snap methodology page and the Spotlight methodology page.
If a published edition is corrected, the correction is appended as a visible note rather than silently edited. How corrections work is set out in full below.
How watch time is measured
Estimated hours watched is the core measurement behind the charts. It is derived from concurrent-viewer sampling taken at a fixed, uniform interval applied identically across every platform we track. Holding the method and the interval constant across all platforms is what makes the numbers comparable from one creator to the next, and it is the foundation the charts are built on.
In plain terms: the number of viewers on a stream multiplied by how long they were there, summed across everything a creator broadcast that week. Ten thousand people watching for three hours is thirty thousand hours watched.
Platforms tracked at launch: Twitch, Kick, and YouTube Live. A creator’s numbers are aggregated across all of their linked channels on these platforms, and simulcast audiences are summed — if someone streams the same session to Twitch and YouTube at once, both audiences count toward their total. They are real people watching, and the chart does not decide which platform they should have chosen.
Accounts on X/Twitter, Instagram and TikTok are recorded where we know them because they help confirm identity, but nothing on those platforms contributes to a ranking. They have no live-viewership signal comparable to the other three.
Too little data to rank
A creator we barely observed is unknown, not zero. A creator needs at least five live observations in the week to be ranked. Below that they are marked insufficient data and left out of both the ranking and the comparison set. We would rather omit a creator than publish a number we do not stand behind.
There is nothing to game here: the remedy is to stream. We publish the figure so that a creator who does not appear can tell whether the reason was their week or our measurement of it.
The Parasocial Streaming 100
The Parasocial Streaming 100 ranks absolute weekly volume: the hundred streamers with the biggest week. Size is never penalized, and there is no momentum or recency weighting anywhere in this chart. A creator who streams heavily and commands watch time holds the top of the chart for as many consecutive weeks as they earn it. The only thing that lowers a ranking is a quieter week, which produces smaller numbers on its own. Pure velocity belongs to the Surge chart, not here.
What the ranking weighs, by category: estimated hours watched (the dominant input); average concurrent viewers, which captures audience size independently of how long someone streamed; engagement, measured through share-and-clip activity on the platforms that expose it; and follower growth across linked platforms.
Also collected, and carried at zero weight today: views on VOD and stream content. It is measured and stored so that calibration can decide whether and when it should count; it does not currently affect any position.
Used to break ties, not to rank: where two creators finish level, the order is settled by total hours watched, then peak concurrent viewers, then alphabetically.
The exact weighting of these inputs is held in versioned configuration and is not published, in the same way industry music and box-office charts disclose their inputs but not their formula. When weightings are reviewed and adjusted, the adjustment is announced with the edition it takes effect.
Comparing unlike things, in plain terms
The inputs are measured in different units — hours, viewers, clips, followers — so they cannot simply be added together. Two different methods put them on a common scale.
Ranking against the field. For most signals, what matters is where a creator sits relative to everyone else that week. A creator in the top 10% of clip activity scores high on clips whether the top of the field was busy or quiet. This is deliberate: these signals are prone to extreme outliers, and a ranking is a more honest reading of them than a raw count.
Measuring against the leader. Hours watched is treated differently, because size is never penalized. If the week’s biggest creator drew four times the hours of the second-biggest, that gap is real and the chart shows it rather than flattening the two into adjacent ranks. Hours watched is scored as a proportion of the week’s leader, on a curve that keeps large gaps visible without letting a single enormous week erase everyone below.
This distinction was not theoretical. During pre-launch calibration, ranking hours watched against the field compressed a 3.4x difference in actual viewership into eleven points of score and placed the week’s second-largest audience around 41st. That result was wrong, and the method changed before launch.
The comparison set
A creator’s score depends on who else was measured that week — that is what “relative to the field” means. The field is every eligible creator we observed that week, not only the hundred who make the chart.
The Parasocial Streaming Surge 50
The Parasocial Streaming Surge 50 ranks velocity rather than volume: week-over-week growth in watch time and engagement. It exists to surface the next wave, the streamers climbing fastest, before they reach the flagship chart.
A creator is eligible for the Surge 50 when all of the following hold:
- They are not on this week’s Parasocial Streaming 100.
- They have never peaked inside the top 25 of the Parasocial Streaming 100. A single past appearance in the lower half does not disqualify a creator, because a comeback is itself a surge story — but reaching the top 25 graduates a creator permanently. Reaching the top 10 of the Surge 50 in two separate editions graduates a creator the same way. As platform-specific charts launch in the future, a second graduation rule activates automatically: repeated top-tier finishes on a platform chart also graduate a creator out of the Surge.
- They meet a minimum scale floor, so a very small channel posting a large percentage jump does not chart. A creator going from 10 viewers to 200 has grown 1,900% without having an audience the chart is about. The floor is measured against a creator’s typical live audience: the median number of concurrent viewers across their live observations over a trailing window — seven days where we have enough of them, thirty days otherwise, requiring at least thirty live observations either way. The median, not the average, so that one unusually large stream does not stand in for a creator’s normal level. From the edition of 10 August 2026 the thresholds are 3,000 concurrent viewers on Twitch, 3,000 on Kick, and 1,500 on YouTube Live. Clearing it on any one platform is enough. Where we cannot measure a typical audience at all — too few live observations in either window — the creator is not eligible, because the threshold is unmet rather than met. We publish these figures. A creator cannot fabricate a sustained typical audience without actually having one, so the numbers are not an evasion route, and withholding them would make a real eligibility outcome unexplainable to the creator it affected.
- They have a measurable prior week. Growth needs two weeks of data.
The Surge surfaces genuine signal, not noise. Rules 1 and 2 mean it is not a junior version of the Streaming 100: it is a chart of creators on the way up, and it lets them go.
What the charts do not measure
The integrity of the charts depends on a hard separation between measurement and everything else.
Parasocial’s editorial coverage of a creator has no effect on that creator’s ranking; the newsroom and the charts share no scoring pathway.
Avarithim verification status does not affect ranking; a verified badge marks data quality and identity, and never adds or subtracts a single position.
Sponsors and advertisers have no access to chart data and no pathway into scoring of any kind.
Who is excluded, and why
Channels that are not individual creators. Relays, rebroadcast loops, network and publisher feeds. These are distribution, not creators streaming.
Event and tournament channels. From the edition of 3 August 2026, tournament broadcasts and event feeds are not eligible for either chart. An event channel is a broadcast slot rather than a creator whose audience follows them — a tournament feed’s week reflects the tournament schedule, not a person’s draw. They continue to be observed and recorded; they are filtered when the chart is built, and the remaining creators are re-scored without them rather than simply having a row removed.
Channels whose audience signals fail eligibility review under the chart’s anti-manipulation policy. These are excluded at the data layer and the affected edition is regenerated so all positions recompute. This is the one exclusion applied case by case rather than by category.
Each exclusion is recorded with the policy category it falls under, the evidence behind the decision, who made it, and when. That record is what makes a disputed exclusion answerable: we can say what was found, on what date, and under which rule. It is held internally rather than published — a creator subject to an exclusion is given the decision and its basis; the detection signature is not made public.
Corrections
A published chart is never edited in place. Every edition is written to a permanent record. When an edition is corrected, a new version is written alongside the old one, the chart begins serving the new version, and the superseded version is retained in full. Corrections are appended as a visible note rather than silently edited.
This means the question “what did the chart say on the day, and what does it say now” always has an answer, and both answers are on the record. All four editions published so far have been corrected, three of them more than once.
Corrections carry a stated reason, and where a correction changes a published ranking it is accompanied by a public note explaining what went wrong and what changed — including when the fault was ours rather than a creator’s.
Five such notes have been published to date:
- Inaugural edition — a channel excluded under the anti-manipulation policy · 6 July 2026
- Edition 4 — a measurement fault that scored an incomplete field · 28 July 2026
- Edition 2 — two creators wrongly ruled ineligible for the Surge 50 · 30 July 2026
- Edition 3 — one creator wrongly ruled ineligible for the Surge 50 · 30 July 2026
- Edition 4, second correction — one creator seated at two positions · 30 July 2026
All integrity notes are indexed at /charts/integrity/.
Reconstructibility. From the edition of 3 August 2026 onward, each published edition also freezes the full set of creators it was scored against, so the ranking can be reconstructed exactly as published even after the roster has moved on. Earlier editions retain their positions, scores and correction history, but the comparison set behind them was not captured at the time and cannot be reproduced in full. We would rather state that limit than imply an audit trail we do not have.
Versioning
When the rules change, the change is recorded as a new methodology version with a date it takes effect, and every published edition carries the version it was produced under.
Versions are never applied backwards. An edition published under version 1 keeps version 1 for good, even after a later version supersedes it. This is why an old edition is still readable as what it was, rather than being silently rewritten by a rule that did not exist yet.
| Version | In effect from | What changed |
|---|---|---|
| 1 | 13 June 2026 | The launch methodology. |
| 2 | 3 August 2026 | Event and tournament channels are no longer eligible for either chart. |
| 3 | 10 August 2026 | The Surge 50’s minimum audience threshold is now derived from observed live viewership rather than from a signal that had drifted from it, and the threshold was recalibrated to match. |
When a methodology change makes an edition incomparable with the one before it, the chart says so. Week-over-week movement — climbed, fell, weeks on chart — is withheld for that edition rather than published as if it were performance. A creator who moves because the measurement changed has not moved. The first use of this is the Surge 50 of 10 August 2026; the Streaming 100 of that same edition is unaffected and reports movement normally.
Designations
Charts carry standard designations that mark notable weeks: the highest new entry, the largest climb, a creator’s return after time off the chart, and milestones for weeks spent at number one. The inaugural edition of each chart is composed entirely of debuts, with movement beginning the following week.
What we publish, and what we do not
Published:
- Every input that contributes to a ranking, and which one dominates.
- How unlike signals are made comparable, in the terms above.
- Every eligibility rule, including the graduation rules for the Surge 50, the minimum-audience thresholds and how that audience is measured, and the minimum number of observations required to be ranked at all.
- Which platforms count, which do not, and that simulcast audiences are summed.
- The tracking week and publication schedule.
- Every methodology version, what changed, and when it took effect.
- The version each published edition was produced under.
- Every correction to a published edition, with its reason.
- That a creator was excluded under eligibility review, and under which category.
Withheld by design:
- The exact weight on each input. Published weights are a specification for gaming the chart: they convert “be a bigger streamer” into “buy this specific quantity of that specific signal.” The relative importance is stated; the arithmetic is not.
- Anomaly and manipulation detection thresholds. The precise point at which viewership data is treated as inconsistent with an organic audience. Publishing it publishes the evasion boundary.
- The evidence in an individual eligibility case, beyond the category. The decision and its basis are disclosed to the affected creator and retained so a dispute can be answered; the detection signature is not made public.
The line: a rule a creator must SATISFY is published. A detector a bad actor would EVADE is not.
An eligibility threshold tells a creator what is required of them, and knowing it does not help anyone meet it dishonestly — you cannot counterfeit a sustained live audience. A detection threshold tells a manipulator exactly how far to go without tripping the alarm. The first belongs to the people being measured; the second does not.
Disputes and corrections desk
If you believe a position, an exclusion, or an absence is wrong, the Parasocial charts desk will answer it.
Where to send it: [email protected]
What to send. The chart and edition date, the creator or channel, and what you believe is wrong. If you have figures of your own — a third-party analytics export, your own platform dashboard — send them; they are treated as evidence worth reconciling against ours, not as a competing claim to be dismissed.
If you believe the issue will affect a forthcoming edition, say so. We will look at it before the next Monday even where the full answer takes longer. Stopping an error recurring and explaining it are different jobs, and the first one should not wait for the second.
What happens to it.
- Acknowledged within two business days, with a reference.
- Reviewed against the retained record for that edition — the positions and scores as published, the correction history, and, for editions from 3 August 2026 onward, the full set of creators the edition was scored against. Review is against what the chart actually did, not a re-run against today’s data.
- Answered within ten business days with one of three outcomes: note, on the same terms as any other correction. application — those go to a methodology decision and are answered on this page rather than case by case.
- No change, with the reasoning and the figures the position rests on.
- A correction, published as a new version of the edition with a public
- Referred, where the question is about a rule rather than its
What we will not do. We will not remove a creator from a chart on request, adjust a position by negotiation, or disclose the detection signature behind an eligibility decision. The decision and its basis go to the affected creator; the detector does not go to anyone.
A dispute never disappears. If it results in no change, the reasoning is on the record and can be re-opened with new evidence. If it results in a correction, the correction is public.
