ML Breakout Probability vs Breakout Score in Rank 51-200
Breakout score ranks today's band candidates by momentum and rank climb. ML success probability adds historical lift when enough labeled history exists.
Updated
Dashboard users sometimes conflate breakout score (a daily composite for the rank band) with ML breakout probability (a model output trained on labeled historical examples). They answer related but distinct questions. Kvantrank shows both when data allows, and blends ML into confidence rather than replacing transparent rank-band logic.
Key Takeaways
- Breakout score = interpretable composite of 7d hype momentum and rank velocity in the band.
- ML success probability = historical pattern match when labeled history exists; may be empty early on.
- Prediction confidence merges coverage, ML confidence (if present), and historical depth when available.
- Neither output is a trade signal; both support research watchlists only.
Breakout score (canonical definition elsewhere)
Breakout score ranks band coins where attention accelerates and market-cap rank improves toward the top 50. Full definition, tables, and workflow context live in the mid-cap breakout candidates pillar and daily watchlist workflow. This post covers only how ML layers on top.
What is ML breakout probability?
When enough labeled history exists, the pipeline may attach success probability estimates per coin. The model compares today’s snapshot against historically labeled patterns and outputs a probability that a coin’s setup resembles past positives.
Early installs or thin history:
- ML columns may be empty or NaN.
- Confidence falls back to signal coverage and historical rules.
- Breakout score still runs on Tier 1 feeds.
How ML interacts with confidence
Unified prediction confidence blends:
- Signal coverage (which feeds fired today)
- Price vs hype conflict penalties
- Historical support when labeled history exists
- ML model confidence when predictions are present (weighted blend with coverage-based confidence)
When ML confidence is available, Kvantrank blends it with coverage-based and historical confidence before displaying the final percentage. See the confidence spoke for the research framing.
Success probability on the dashboard may display ML output when present; otherwise a non-ML heuristic may appear until the model is ready.
When to trust which metric
| Situation | Lean on | Why |
|---|---|---|
| First 1-2 weeks of data | Breakout score + hype | ML labels immature |
| Mature history, high ML confidence | Breakout score + ML prob | Agreement reduces false positives |
| High breakout, low confidence | Confidence diagnostics | Missing Tier 2 or conflict penalty |
| Stablecoin row | Neither | Confidence forced to zero |
Always exclude stablecoins from breakout comparisons.
ML vs vendor black boxes
LunarCrush Galaxy Score and similar metrics are vendor-defined composites. Kvantrank keeps breakout score transparent and treats ML as an optional lift layer when historical data allows. Compare external tools in the Galaxy vs AltRank post but run daily sorts on one stack.
Practical workflow
- Sort by breakout score.
- Note success probability when populated; treat absence as “model not ready.”
- Require minimum confidence for automated alerts.
- Cross-check divergence on top rows.
- Review the full scoring stack weekly.
Kvantrank treats ML as historical assist, not autopilot. Breakout score stays explainable when the model disagrees.
Frequently asked questions
Does high ML probability guarantee top-50 entry?
No. Labels describe historical patterns in a specific band and regime; markets shift.
Why is ML probability blank for my coin?
Insufficient history, missing data, or model not ready for that snapshot day.
Is ML the same as breakout score?
No. Breakout score is always computed from momentum and rank rules; ML is additive context.
How do I improve ML quality?
Accumulate daily runs so historical labels and model inputs can mature over time.
Related reading
Not financial advice. For informational purposes only.