How We Score — Beat Pistol Hit Predictor

01

The Methodology

100% audio-based scoring — no metadata, filenames, or artist names influence results. We analyze the actual sound, not the label on the file.

🎧

Client-Side Analysis

Your audio is decoded and analyzed entirely in your browser using the Web Audio API. Nothing is uploaded to our servers — the raw file never leaves your device.

📊

Spectral + Rhythmic

We extract frequency-domain and time-domain features via FFT (Fast Fourier Transform) and onset detection. The same signal processing used in professional audio tools.

🎯

Two-Layer Scoring

A weighted distance model gives a continuous Hit Potential %. A second rule-based layer checks five objective thresholds, earning bonus points (max 65 pts → 0–100 scale).

🏆

Benchmark Calibrated

Feature targets are calibrated against charting tracks spanning 25 years of Billboard Hot 100 data. Optimal values reflect what those tracks average.

02

Scoring Categories & Weights

Seven audio features are extracted per track. Six carry weight in the Hit Potential score. Word Count is an additional lyric-layer analysis reported in the Features section.

Feature	Weight	Optimal Range	What It Measures
Danceability	25%	55–88%	Beat regularity + tempo sweetness + energy combo. Reflects how well the track drives movement on a dancefloor.
Energy	20%	45–85%	RMS-based loudness intensity. Measures perceived power and intensity throughout the track.
Tempo	18%	90–140 BPM	BPM via onset detection and autocorrelation. Optimal target is 120 BPM — the sweet spot for most radio-ready formats.
Loudness	15%	55–92% (dBFS normalized)	Integrated loudness normalized from dBFS (-35 to 0 dBFS range). Reflects mastering quality and streaming readiness.
Valence	12%	25–78%	Spectral brightness + tempo + energy composite. Correlates with perceived emotional positivity — higher-valence tracks trend well on streaming.
Word Count	Context	Varies by genre	Estimated vocal word count derived from speechiness × duration. Used for Narrative Pop classification and lyric density context. Does not affect the Hit Score directly.

03

The Five Rules (Billboard DNA Layer)

On top of the weighted score, five objective rules are checked. Each rule that passes earns bonus points. Max total: 65 points, then normalized to the 0–100 badge scale.

🔥

Dance + Energy Combo

Danceability above 70% and Energy above 60%. The most heavily weighted rule — this combination is the single strongest predictor of chart performance in the dataset.

+20 pts

🎯

Genre Tempo Match

Tempo falls within the optimal window for the detected genre. For pop/hip-hop, that's 110–130 BPM. Genre-aware tempo scoring means a Lo-Fi track at 75 BPM isn't penalized for missing the pop sweet spot.

+15 pts

✨

High Valence

Valence score above 60%. Bright, positive-feeling tracks historically over-index on streaming replays and social sharing — two key signals for algorithmic playlist placement.

+10 pts

🎤

Vocal Presence

Vocal presence detected — confirms human singing or rapping is present, not purely instrumental or spoken-word content.

+10 pts

📻

Radio-Ready Mastering

Integrated loudness above −8 dBFS (normalized). Tracks mastered below this threshold are typically perceived as quiet on streaming platforms and radio.

+10 pts

04

Badge System

Badges are awarded based on the rule-based score (0–100). A track earns a badge when it clears a score threshold — there's no human curation involved. Three tiers, each with a clear meaning.

0–39

40

–59

60–

–89

90+

No Badge Certified Premium Bullseye

🔫

Pistol Beat Certified

40–59

Passed at least 2–3 of the five rules. Solid audio fundamentals — commercially viable with refinement.

Threshold: 40+ pts

⭐

Premium Pick

60–89

Passed 3–4 rules including the high-weight combo. Strong hit indicators — production quality shows.

Threshold: 60+ pts

🎯

Bullseye Hit

90+

Passed 4–5 rules with strong scores across all features. Exceptional alignment with proven hit characteristics.

Threshold: 90+ pts

05

Trend Score

Two signals, one number. The Trend Score you see is the average of Billboard DNA (Layer 1) and Chart Trend Intelligence (Layer 2). It's the final, blended read on your track's commercial potential.

Most scoring tools run a single model. Beat Pistol runs two independent layers and blends them. Billboard DNA looks backward — 25 years of proven hits. Chart Trend Intelligence looks forward — where the sound is right now. The Trend Score is their average.

📀

Layer 1: Billboard DNA

Weighted distance from the centroid of 25 years of Hot 100 chart data. Five rules add bonus points up to 65 pts. Normalized to 0–100. This is the historical hit benchmark.

📈

Layer 2: Chart Trend Intelligence

Compares your features against a live-weighted fingerprint of current charting trends — which genres are rising, which are fading, which tempo/energy combinations are getting traction right now.

Why blend two layers? A track that nails Billboard DNA but misses the current moment might have been a hit in 2015. A track scoring high on Trend Intelligence but skipping fundamentals might ride a micro-trend and fade. The Trend Score rewards tracks that do both — sound current and hit the proven structural markers.

06

AI Music Benchmark

Most Beat Pistol users are AI and indie producers. Billboard calibration alone misses what's actually performing in the AI music ecosystem. The AI Music Benchmark is a separate comparison against seven trending AI-music genre profiles — showing how your track stacks up against what's getting plays on Suno, Udio, and Spotify's AI-curated playlists.

Seven AI Genre Profiles

AI Genre Profile	Energy	Dance.	Valence	Notes
AI Pop	0.65	0.70	0.73	Upbeat, accessible, high valence
AI Electronic	0.83	0.86	0.62	Club-ready, high energy, driving
AI Lo-Fi / Chill	0.32	0.48	0.66	Quiet, textural, lo-fi aesthetic
AI Cinematic	0.52	0.22	0.55	Orchestral, low danceability, atmospheric
AI Hip-Hop	0.72	0.80	0.40	Heavy beat, trap tempo, rhythmic
AI Folk / Acoustic	0.38	0.44	0.72	Organic, gentle, warm
AI Indie / Alt	0.66	0.54	0.58	Guitar-forward, moderate everything

AI music listeners over-index on valence and danceability — they skip darker, low-energy tracks faster. The AI score weights valence and danceability slightly higher to reflect these platform-specific patterns.

07

Genre Classifier v5 — 31 Genres

Genre prediction uses a KNN (K-Nearest Neighbors) classifier with K=9, trained on 165+ reference tracks spanning 31 genres. Your track's six audio features are compared against every reference track using weighted Euclidean distance — closer tracks vote for their genre, and the top 9 voters determine the result.

How KNN classification works: We compute a distance score between your track and every reference track in the library. The 9 closest matches vote for their genre, weighted by how close they are. The top 3 vote-winners become your genre predictions, with confidence percentages showing the vote share.

AI Cross-Validation

When the AI Music Benchmark identifies a strong genre match, that signal cross-validates the KNN result. If both agree, confidence is boosted. If they disagree, the result is flagged as a hybrid or "Best Guess." This prevents misclassification when a track sits between genres.

All 31 Genres

Top 40 Pop Hip-Hop / Rap Trap / Modern Hip-Hop R&B / Soul EDM / Dance Rock Alternative Rock Indie / Alternative Indie Pop Country Latin Metal Classical Jazz Funk / Disco Reggaeton K-Pop Lo-Fi / Chill Synthwave / Retrowave Ambient / Atmospheric Bedroom Pop Downtempo / Trip-Hop Drill / UK Drill Afrobeats Phonk Hyperpop Punk Singer-Songwriter Hard Rock New Wave / Synth-Pop Soft Rock

Note on Narrative Pop: Narrative Pop is a special 32nd classification that is not available through KNN alone. It requires the Word Count gate: 300+ estimated words with sufficient vocal density. This is because narrative density is a lyric-layer signal that audio features alone can't reliably detect. When both conditions are met, Narrative Pop overrides the KNN result.

08

Word Count Analysis

Word Count estimates how many words are in a track's vocals — derived from vocal density × duration. It's displayed in the Audio Features section as a lyric density signal, with context relative to the detected genre's typical word count.

🔢

How It's Calculated

words ≈ vocal_density × duration × 150
Where 150 is the average words-per-minute rate for sung/rapped vocals. Result is clamped and smoothed per genre.

🗣️

Why It Matters

Lyric density tells a different story than energy or tempo. A 400-word storytelling track and a 60-word hook-driven pop song can have identical audio features but very different creative profiles.

Narrative Pop Gate: When estimated word count exceeds 300 words and vocal density is high enough for song classification, the genre classifier overrides its KNN result and labels the track Narrative Pop. This catches story-driven tracks — like folk confessionals or spoken-word-adjacent hip-hop — that audio features alone would misclassify as ambient, indie, or country.

09

Copycat Check v2

Copycat Check v2 detects when your track sounds like an existing hit — not by comparing raw features, but by comparing genre-filtered residuals. It's the difference between "this track sounds like pop" and "this track sounds like this specific pop hit."

How Genre Fingerprint Filtering Works

1

Detect genre centroid. The detected genre's average feature profile (the "centroid") is established. For example, a Rock centroid has high energy, low danceability, moderate tempo.

2

Compute residuals. Your track's deviation from the genre centroid is calculated for each feature. This filters out common genre characteristics — leaving only what makes your track distinctive within the genre.

3

Compare residuals to reference tracks. The residual vector is compared against reference tracks' residuals (each also deviation-corrected from their own genre centroid). Similarity is measured on the filtered signal, not raw features.

4

Apply the 47% threshold. Any match below 47% similarity is filtered out — it's genre overlap, not copying. Only matches above the threshold are meaningful enough to surface.

The 47% Threshold

Genre Overlap — Hidden

Flag Above 47%

47%

Below 47%: the similarity is explained by shared genre characteristics — not meaningful. Above 47%: the track occupies a similar distinctive sonic space to a specific reference hit.

Single-Match vs. Multi-Match Logic

1–2 Matches Above 47%

Flagged as a genuine copycat signal. The track shares specific sonic characteristics with 1–2 reference hits that go beyond genre. Results are shown with match percentages. Use it as positioning intel.

3+ Matches Above 47%

Suppressed entirely — this is genre overlap, not copying. If your track simultaneously sounds like 3+ hits above the threshold, you're just firmly in a genre. The Copycat section is hidden in this case.

What "75%+ match" means: Copycat Check only appears in results when the top match scores 75% or higher. Below that threshold, the section is hidden even if 1–2 matches are technically above 47%. When it does appear, a match is not plagiarism — it's a signal that your track occupies a similar sonic space to an established hit.

10

Unique Style Label

When the Genre Classifier can't confidently assign a primary genre, most systems either force a label or report an error. Beat Pistol instead surfaces a 🎯 Unique Style label — signaling that your track is deliberately genre-defying, not unclassifiable.

Trigger Conditions (both must be true)

• Top genre confidence < 45% — no single genre dominates

• Spread between #1 and #3 genre confidence < 15% — the scores are evenly distributed

Why this is positive, not a warning: Genre-defying tracks are a creative signature — they're harder to place, which can be an asset in algorithm-driven playlisting and sync licensing. Unique Style replaces the old "Low Confidence — Genre Blend Detected" message that falsely implied a problem. It's not a problem. It's a data point.

11

Benchmark Reference Values

Target ranges are calibrated against feature profiles of charting tracks spanning 25 years of Billboard Hot 100 data. The optimal value reflects the mean of those tracks; the range covers the 25th–75th percentile.

Feature	Optimal (Target)	Range (Lo–Hi)	Unit
Danceability	0.72	0.55 – 0.88	0–1 normalized
Energy	0.68	0.45 – 0.85	0–1 normalized
Tempo	120 BPM	90 – 140 BPM	Beats per minute
Loudness	0.74	0.55 – 0.92	0–1 (dBFS normalized)
Valence	0.52	0.25 – 0.78	0–1 normalized

✓ Backed by real Billboard data. Benchmarks are calibrated against 25+ years of Billboard Hot 100 chart data (2000–2023). Genre profiles are computed from thousands of charted tracks and updated as new data is imported. Billboard Validation: 89 Hot 100 tracks tested — Tier 1 (#1–10) avg 85.5%, Tier 2 (#11–40) avg 87.0%, Tier 3 (#41–100) avg 88.3%.

12

Privacy

🔒

Your Audio Never Leaves Your Device

We don't store your uploads. Audio is decoded and analyzed in real-time inside your browser. No audio data is transmitted to our servers at any point during analysis.

No third-party data sharing. Analysis results and feature data are not shared with any third parties. If you choose to share your results or submit your email for a breakdown, only that information is transmitted — never the audio itself.

Scoring is filename-agnostic. Track names, artist names, and file metadata are never factored into any score. Two identical audio files with different names produce identical scores.