How We Score Games
Most trend tools answer "what's popular?" TrendForge answers a different question: where is the content gap? A game can be soaring on Steam while YouTube coverage is still thin — that gap is where small and mid-size creators get a real chance. This page explains the methodology behind every score on the site.
Three peer signals, each on its own 0–100 scale (higher is always better), blended into one Opportunity score. The blend is run through a fixed calibration curve so the headline number lands on the same 0–100 opportunity scale used by mainstream trend dashboards — a daily top pick reads ~90–94, a saturated evergreen reads ~50 — while the three sub-signals stay on their raw, legible scales. An earlier multiply formula (“Trend × Gap × Freshness × Momentum”) collapsed real-world picks to around 40/100 because any single sub-1 factor halved the score; the blend keeps every signal legible and on the same scale.
Demand — is attention there? (0–100)
Heat is measured by rate of change, not absolute size. A giant like CS2 is always big; that tells a creator nothing. We weight growth highest so "rising now" beats "always large".
Supply — how open is the lane? (0–100, higher = less crowded)
The flip side of competition. We first score how saturated the YouTube shelf is (video density + big-channel dominance, 0–100, higher = more crowded), then invert it so a wide-open game scores high. High dominance means even a good small-channel video gets buried — so it pushes Supply down.
Timing — is now the moment to post? (0–100)
Timing is, above all, a momentum read: how sharply a game is rising right now (Steam player growth trend). A game that's surging scores high; one that's flat or declining scores low. A launch or major-update window adds a bonus on top — the moment a creator wants to be early. Weights are tuned server-side.
Confidence & risk flags — advisory, not a demotion
Some spikes are misleading. Each game is checked for three risk flags, shown alongside the score rather than folded into it — so a flagged pick is never hidden, just labeled, and you decide.
| discount — game is on sale; player spike may be price-driven, not organic | — |
| unsustained — sharp spike, no launch/event, low base: possible one-day meme | — |
| ambiguous — generic/short name; search intent may be split | — |
Confidence is derived from flag count: 0 = high, 1 = medium, ≥2 = low. It is a peer signal to Opportunity, not part of the blend — matching how serious creators actually read a recommendation (a "why" before a "how much").
Fit — realistic for your channel size
A topic realistic for a 100K channel can be a dead end for a new one. The dashboard's channel-size selector re-reads each card's Fit from the same big-channel dominance signal, scaled by your tier: 50K+ (unaffected) → 10K–50K → 1K–10K → Under 1K (most sensitive to dominance). Fit labels: Good / OK / Risky for your size. The feed's default sort stays Opportunity; Fit is a per-card read.
Small-creator radar — a second leaderboard, ranked differently
The main ranking sorts by Opportunity. The radar is a second leaderboard built for small and new channels: it re-reads the same signals but ranks by how reachable the topic is for a small creator right now. Same data, different question — "can I actually surface here?" instead of "how big is the audience?".
Real audience — is anyone actually watching? (0–100)
Whether real viewers are actually watching this game right now — not bots, not hype spikes, not a stale legacy base. A high score means a live, searchable audience is already there; a low score means the interest looks thin or manufactured.
Creator gap — is the lane open? (0–100, higher = more open)
The opposite of one giant sitting on the search results. A high gap means no single channel has locked up the topic — competition is fragmented and a small channel's video can still get surfaced. A low gap means the shelf is already crowded. High = open lane; low = wall.
Big-channel lock — shown only when it bites
How much a handful of big channels dominate the top results. High lock means even a great small-channel video gets buried under their backlog. Only shown when big-channel control is strong enough to matter.
Launch boost — be early to a content vacuum
A fresh launch or major update just opened a content vacuum — early videos get recommended harder because nobody has a backlog yet. Disappears as the shelf fills in.
| Real audience — live verified viewership | — |
| Creator gap — higher = more open lane | — |
| Big-channel lock — higher = harder to surface (advisory) | — |
| Launch boost — present only during a launch window | — |
| Signal | Fragmented niche (small-channel win) | Giant-dominated game |
|---|---|---|
| Real audience | 72 | 80 |
| Creator gap | 88 | 22 |
| Big-channel lock | 12 | 71 |
| Launch boost | now | — |
Illustrative hypothetical, not live games. The giant's audience is bigger, but the niche is where a small channel actually surfaces. That re-rank is the whole point of the radar.
These labels are public so you can read the radar; the underlying per-game scores are part of Pro.
Worked example: rising indie vs a blockbuster
Two games, same shelf, opposite verdicts. This is why raw popularity is the wrong sort.
| Signal (0–100) | Rising indie (launch + surging) | Saturated blockbuster |
|---|---|---|
| Demand | 78 | 55 |
| Supply | 32 | 24 |
| Timing | 100 | 17 |
| Opportunity | 90 | 53 |
The blockbuster has higher raw popularity — a best-seller list would rank it first. TrendForge ranks the indie, because demand is climbing into a wide-open lane at the right moment. That is the judgment a creator actually needs.
Where scoring ends and the LLM begins
Scoring is a deterministic formula — fully explainable and tunable, the same math every day. An LLM is used only on the top-ranked games to turn the score breakdown into concrete title/angle ideas. The model never touches the numbers; that separation keeps the ranking trustworthy and lets us tune weights over time against real creator outcomes.
Data sources & limits
Public signals only: Steam / SteamCharts (players, growth, discounts), the YouTube Data API (recent videos, creator supply), and Google Trends (momentum). Recomputed once a day. Scores are research support, not a guarantee of views or growth — validate the evidence before recording. Not affiliated with Valve, Google, or YouTube.