Why PINtPOINT picks a better pint.
PINtPOINT is built to help you choose what to drink next, from the beers actually on tap in front of you. These four essays explain that approach from four angles: the problem with most beer recommendation systems, the engine behind PINtDEXTER, the flavour vocabulary that makes recommendations easier to read, and the drinking experience that turns a tap list into a better decision.
The four subjects
Where to start
The four essays
Why collaborative-filtering recommenders — the ones built on top of Untappd ratings or RateBeer reviews — keep producing the same disappointing pours. Cold start, popularity bias, geography-blindness. A response to two public academic attempts, and the design constraint PINtPOINT optimises against instead.
The engineering essay. How PINtDEXTER learns taste quickly, builds a profile for each beer, and matches both against what is actually on tap nearby. The mechanics, the architecture, and the bug that exposed the need for a better fermentation signal.
A practical taxonomy for the ingredient that does most of the personality work in modern beer. Six teaching categories (citrus, tropical, pine & resin, stone fruit & berry, floral & herbal, earthy & spicy), the chemistry behind them, and per-variety cards for the hops you'll actually meet on a tap list — the vocabulary you'll need to read a recommendation or steer one.
A reading method for the pint in front of you. Eight axes — Body, Bitterness, Sweetness, Roast, Fruit, Finish, Softness, Complexity — plus drinker cues, a drinkability bucket, and one plain-English "what it feels like" sentence per beer. Tuned for the beers you'll actually find on most tap lists; honest about the sour and Belgian-yeast styles it doesn't cover.
Source material
- Ethan Haley — Untappd as a recommender (RPubs, 2021). The argument that ratings are noise-dominated and location is the honest signal.
- Stan Hieronymus — Another way to think about hop presence (Appellation Beer). The "tyranny of the tasting note" framing.
- NYC Data Science — NINKASI (2020). Collaborative-filtering attempt that the Problem essay responds to.
- Beer-writing pillars referenced across the arc: Martyn Cornell, Roger Protz, Randy Mosher, Garrett Oliver's Oxford Companion to Beer, the BJCP style guidelines.