Casino Streamer Data Methodology: Collection, Filtering & Heuristic Models

1. Scope & Dataset Architecture

StreamCasAnalyst indexes 4,871 Kick channels and 3,799 Twitch channels broadcasting in casino-related categories (Slots & Casino, Slots, Poker, Virtual Casino). Across cross-platform identity resolution, these form 8,559 verified creator profiles as of snapshot 5 Nov 2026.

2. Primary Metric Definitions & Data Provenance

• Registered Followers: Total follower count reported on the broadcaster's public channel profile. On Twitch, daily deltas are archived via official Helix API snapshots.
• Highest Observed Peak Viewers: The largest concurrent viewer count recorded by our live scanner during monitored broadcasts. Because polling occurs at discrete ~2.5-minute intervals rather than every second, observed peak is a conservative lower bound of the true all-time peak.
• Peak ÷ Followers Ratio: Calculated as Observed Peak ÷ Followers. Ratios exceeding 100% on smaller channels indicate that peak concurrency exceeded registered followers (which can result from platform front-page placements, raids, external embeds, or artificial inflation). It is presented strictly as a descriptive ratio, never a standalone verdict.
• Average Stream Duration & VOD Count: Mean broadcast length and archive count computed from platform video-on-demand (VOD) listings and monitored live sessions.
• Disclosed Sponsor & Referral Links: External domains and affiliate parameters (ref, code, btag, promo) extracted verbatim from public channel bios and panels.

3. Real-Time Telemetry & Chat Ingestion

Our background telemetry collector polls Kick and Twitch casino categories every ~2.5 minutes to log concurrent viewership and active session boundaries, while connecting to public chat feeds (Twitch IRC and Kick realtime chat) for top live broadcasts. From public chat events, the scanner aggregates:

• Messages per minute and unique active chatters per 10-minute sliding window.
• Lexical uniqueness (ratio of distinct messages to total messages) and top-5 chatter concentration.
• Account creation age distribution (where platform user IDs permit public estimation).
• Moderation actions (timeouts/bans), gifted subscription events, and chat burst spikes (msgs/10s).

Hidden Viewer Counters: When a Kick channel disables the visual viewer counter on its consumer web page, the underlying concurrency figure remains present in the platform's public category listing API; StreamCasAnalyst displays this public API figure and marks the channel as “counter hidden”.

Visual Milestones: We do not record or redistribute video streams. Stream snapshots are standard platform preview thumbnails captured at 10-minute intervals (or during chat bursts) for broadcasts above 300 CCV, watermarked with UTC timestamp and viewer count for archival verification.

4. Audience Diagnostic Signals (6-Vector Heuristic Formula)

For monitored broadcasts, StreamCasAnalyst computes a 0–100 Audience Diagnostic Score as a weighted mean of up to six normalized public signals $s_i in [0, 1]$:

• Vector 1 — Chat Participation (Weight: 30%): Unique chatters over 10 minutes divided by concurrent viewers. Full score ($1.0$) at $ge 3.0%$ participation.
• Vector 2 — Viewers vs. Followers Ratio (Weight: 20%): Evaluates whether live concurrency is proportionate to the channel's follower base. Full score ($1.0$) at $le 10%$ ratio; scales down toward $0$ at $ge 100%$.
• Vector 3 — Message Uniqueness (Weight: 15%): Share of non-duplicate chat lines inside the sampling window. Full score at $ge 80%$ unique text; $0$ at $le 25%$ (repetitive macro spam).
• Vector 4 — Account Freshness (Weight: 15%): Share of chatting accounts estimated to be newly created ($< 30$ days old). Full score at $0%$ throwaway accounts; $0$ at $ge 25%$.
• Vector 5 — Chatter Concentration / Top-5 Dominance (Weight: 10%): Share of total chat volume generated by the 5 loudest accounts. Full score when top 5 accounts represent $le 15%$ of total messages.
• Vector 6 — Viewer Curve Variation / CV% (Weight: 10%): Coefficient of variation ($sigma / mu$) of viewer samples across a 40-minute window. Natural human audiences fluctuate organically; flat step-function plateaus score lower.

Missing Vector Handling & Interval Integration: When live chat or curve vectors are unavailable for an offline or unmonitored channel, those vectors are explicitly marked Unavailable and excluded from the weighted denominator Σ(w_i) — never replaced with synthetic favorable constants. Hours Watched and Average Viewers are integrated across clamped polling intervals (NOMINAL_INTERVAL_SEC = 150s, MAX_GAP_SEC = 375s), strictly separating Measured Only scanner sessions from Archive Estimates (~est.).

5. Legal & Statistical Disclaimer on Heuristic Scores

Important Legal Notice: The Audience Authenticity Score and Estimated Organic Viewers are automated statistical heuristics and analytical opinions derived exclusively from public telemetry. They are not calibrated against private platform server logs (IP addresses, device fingerprints, or authenticated session tokens) to which only Kick and Twitch have access. Accordingly, StreamCasAnalyst does not assert as a statement of fact that any specific viewer is a bot, or that any broadcaster purchases, operates, or authorizes artificial viewership. Low scores may reflect legitimate phenomena such as embedded players, quiet “lurker” audiences, raids, or giveaway rules. Always inspect the underlying 6-vector breakdown and raw session curves directly.

6. Automated Throw-Away Channel Filtering

To prevent disposable spam accounts from polluting directory rankings, our ingestion pipeline automatically excludes channels matching any of four objective thresholds:

• Observed peak of 300+ viewers with fewer than 50 followers, or 1,000+ viewers with fewer than 200 followers.
• Observed on a single day only, with zero social links, a bio under 10 characters, and fewer than 100 followers.
• No avatar, no bio, no social links, and zero recorded VODs.
• Fewer than 10 followers, zero VODs, and zero social links.

7. Cross-Platform Identity Resolution & Corrections

A Kick channel and a Twitch channel are merged into a single dual-platform profile (107 confirmed profiles) only when one channel explicitly links to the other or both publish the exact same verified social handle (X/Twitter, Instagram, YouTube, TikTok, Telegram, or Discord). Channels that merely share a display name are never merged without cross-link verification. To request a manual audit, profile merge/split, or data correction, contact [email protected].