Scoring explained
Split the G Ranking Methodology
Split the G ranks how closely the foam-to-beer line left after your first sip crosses the center of the G on a Guinness glass. A computer-vision workflow detects the glass, G, beer, and split line, then returns a score from 0 to 5. The score measures this visual pub-game result only, not the untouched pint's pour quality or taste.
What Does Split the G Rank?
Split the G ranks the position of one visible boundary: the line between foam and beer after the player takes a first sip. The target is the vertical center of the G printed on the glass. A line through that center earns the strongest result; distance above or below it lowers the score.
This boundary matters because the game starts only after the pint has been served and the player drinks from it. Split the G does not grade an untouched finished pint. It also does not judge head texture, settling time, color, taste, freshness, tap maintenance, or the pub itself.
If you want the serving method that comes before the game, read the separate evidence-based Guinness pouring guide. For the rules and background of the first-sip challenge, see what Split the G means.
How Is a Split the G Score Calculated?
The app uses a two-stage computer-vision workflow. The first stage detects the Guinness glass, the G, and the beer. A focused second stage examines the G region for the split line. The calculation then follows one of three paths.
| Detected result | Score range | How the score changes |
|---|---|---|
| Split detected inside the G | 2.5 to 5 | Closer to the center of the focused G image scores higher. |
| No split detected inside the G | 0 to 2.5 | The beer level is compared with the detected G center, with distance reducing the score. |
| Required features not detected | 0 | The image cannot receive a positional score without the required detections. |
On the split-detected path, the code measures the detected line's distance from the center and maps that distance to 2.5 through 5. On the fallback path, it compares the top of the detected beer region with the center of the detected G and applies a squared distance decay from 2.5 toward 0. A missing beer or G detection returns 0.
Both paths can return exactly 2.5. It is the shared boundary where a detected split is farthest from the focused image's center or the fallback beer level aligns with the detected G center. The value alone does not identify which path produced it.
Roboflow's technical account of the Split the G workflow explains why the focused second model is used: the curved liquid surface near the glass can make a whole-pint beer boundary differ from the level crossing the G.
What Do Scores from 0 to 5 Mean?
Higher scores mean the detected line is closer to the target center. The scale is continuous, so decimal scores preserve smaller differences between attempts. The ranges below are interpretation aids, not separate judging categories.
- 5: the detected line is centered on the G.
- Above 2.5 and below 5: a split is detected, with the score rising as it approaches the center.
- 2.5: the shared boundary; either scoring path can produce this value.
- Above 0 and below 2.5: the fallback beer-level comparison finds a miss at some distance from the center.
- 0: the attempt is at the edge of the scale or the required image features cannot be scored.
How Should You Compare Scores?
Compare attempts made with the same Split the G system and 0-to-5 scale. Other products can use different targets, model versions, formulas, or percentage scales. Their numbers should not be treated as equivalent.
Image conditions also matter. Keep the glass upright, show the full G, and make the foam-to-beer boundary visible. Reflections, an angled camera, a hidden logo, or a cropped boundary can change the detections. A score is the workflow's measurement of that submitted image, not an independent certification of the drink.
See Split the G leaderboard scoresRanking Methodology FAQ
What does Split the G score?
Split the G scores where the foam-to-beer line rests against the G after your first sip. It does not grade an untouched finished pint, taste, freshness, serving temperature, or overall pub quality.
What is a perfect Split the G score?
A perfect score is 5. The detected foam-to-beer line must pass through the center of the G. Scores decrease as that line moves farther from the center.
Why can two similar photos receive different scores?
The score depends on the detected G, beer level, and split line in each image. Camera angle, reflections, framing, a hidden logo, or an unclear boundary can change what the computer-vision workflow detects.
Does a Split the G score measure Guinness quality?
No. The score measures one visual pub-game result after the first sip. It does not measure flavor, aroma, temperature, freshness, tap condition, or whether the original pour met a professional quality standard.
How should Split the G scores be compared?
Compare scores produced by Split the G on its 0 to 5 scale. Other apps may use different targets, models, or scales, so their numbers are not directly interchangeable.