Methodology
About the Math
How the catalog earns its scores, how the calculator models a drink, and where the math deliberately stops.
How we score the catalog
Every scored spec in the catalog can carry two independent computed views. The circular hero is Computed balance: its percentile among cocktails with a producer score, colored by a Gold, Silver, or Bronze tier. It comes from a computed perceptual-balance score — a modeled blend of acid, sugar, ethanol, and dilution chemistry. No tasting panel touched these numbers. The compressed raw score is not presented as a grade out of 100.
Service resilience is a separate operational axis: how well the engine expects the build to hold up under dilution, ice, and technique variation, ranked within its own technique cohort — stirred against stirred, shaken against shaken. It is not a taste grade, so a Gold balance tier can coherently sit beside low service resilience. The service percentile maps onto five bands:
- Bulletproof90th percentile and upSurvives almost any service reality.
- Bar-Ready75–89Holds up under normal volume and variance.
- Solid55–74Reliable with reasonable care.
- Technique-Sensitive35–54Execution quality moves the result.
- Fussybelow 35Small drifts show up in the glass.
(The catalog filter groups the middle bands as Service-Aware and the lowest as Fragile for browsing.) The displayed percentile caps at 99 — no spec out-ranks 100% of its own cohort — and percentiles computed over tiny cohorts are suppressed rather than over-claimed. Specs without a percentile read "Resilience pending" instead of wearing a band.
Each recipe page also breaks out a quality profile — Forgiveness, Acid-Sugar Balance, Ice stability, and Technique stability, each 0–100 — and shows engine provenance (qualification version and the date the spec was last scored). When the underlying data can't support a subscore, it reads "Pending" with the reason, never a silent guess.
All scores are computed estimates, not preference predictions. They are still being calibrated and refined; the public labels describe the axis and the stored evidence without claiming that a model knows whether you'll like the drink.
Why this exists
Most drinks are built on intuition. This tool exists for when intuition is not enough.
Daily Libations is designed for batch consistency. It turns dilution, freeze point, sugar, and acidity into visible constraints so operators can make repeatable decisions across shifts, teams, and service volume.
This page explains framework-level behavior. Proprietary coefficients and calibration constants are intentionally not published.
Core calculator models
Freeze Point (Frozen)
Freeze behavior is modeled from alcohol strength, sugar load, and acidity using the same manifest-backed freeze-floor model used by the internal Daily Libations engine. Fructose-heavy sweeteners are handled differently from sucrose-style sweeteners to better match observed machine behavior.
The public frozen gauge is computed by the Daily Libations engine as a service estimate for a batch recipe or finished cocktail profile.
Dilution (Stirred / Shaken / Built)
Non-frozen techniques estimate in-glass dilution from style and starting strength. Stirred and shaken use dynamic service estimates; built uses low, stable dilution assumptions.
Operators can optionally layer service calibration multipliers (ice profile, vessel profile, ice-state, agitation, and custom venue calibration) on top of the baseline model to better match in-house outcomes.
Serving counts are based on expected in-glass volume (pre-service volume plus modeled dilution), not only raw pre-ice volume. Default serve sizes are technique-aware: Frozen 10 oz, Stirred 3 oz, Shaken 4 oz, Built 6 oz, Draft 8 oz.
Balance Score (calculator)
The calculator's Balance Score is a live 0–100 composite across ABV, Brix, acidity, and the technique-specific Gauge 3 signal. Weighting and thresholds are technique-calibrated so the score reflects service outcomes, not just textbook ranges.
This is the workbench's own gauge for the spec you are building right now — separate from the catalog's Service-Resilience grading and from the Acid-Sugar Balance subscore on recipe pages.
Technique context
Frozen mode evaluates the mix exactly as it goes into the machine. Stirred, shaken, and built modes evaluate pre-ice inputs and then model dilution separately. If a warning appears, verify which technique you are in first.
Solver targets follow the same rule: frozen targets are in-batch; non-frozen targets are pre-ice targets that assume dilution will happen during service.
Sources
Model assumptions are aligned with established cocktail and food-science references, including peer-reviewed cold-behavior data, sugar and freezing references, and field validation from service workflows.
This is an operator tool, not an academic simulator. The model prioritizes service realism and repeatability.