Case Study: Team 1678 Citrus Circuits' Scouting Pipeline

Why study 1678#

Team 1678, Citrus Circuits (Davis, CA), publishes a detailed scouting whitepaper most years and open-sources much of its tooling. It is one of the most-cited references for a fully electronic, high-volume scouting operation, and its architecture is worth reverse-engineering even if you build something smaller.

The pipeline shape#

The system follows a clean collection -> consolidation -> server -> viewer flow, with distinct apps for distinct roles rather than one monolithic app. The 1678 GitHub release reflects this with separate repositories such as Match Collection, Pit Collection, Server, Viewer, and Stand Strategist:

• Stand / match collection app: a fleet of devices, one scout per robot, tapping actions live. Output is encoded as QR codes, the same offline-first idea as QRScout, so no venue wifi is needed.
• Super-scouts: dedicated scouts who capture subjective, hard-to-quantify information (driver skill, defense quality, robot reliability, strategy) that objective taps cannot, kept separate from the objective counts.
• Pit collection: robot specs, dimensions, mechanism details, and reliability notes gathered in the pits.
• Server / consolidation: scanned QR data is aggregated, scout records are reconciled, and derived metrics are computed automatically because matches happen faster than humans can compile by hand.
• Viewer / Stand Strategist: strategists browse computed team profiles, rankings, and predictions to drive picklists and match plans.

The key design principles to steal#

1. Separate objective and subjective data. Stand scouts record raw counts; super-scouts record judgments. Mixing them lets bias leak into your hard numbers, so keep the streams distinct and label the subjective ones.
2. Multiple roles, not one super-app. Splitting collection, super-scouting, pit, server, and viewer keeps each tool simple and each scout focused on one job, which is the single biggest accuracy lever.
3. Automate aggregation. The whole point of the server is that no human can keep up with the match pace; computed metrics and predictions are generated continuously.
4. Measure your own data accuracy. Their whitepapers explicitly analyze data accuracy and document lessons learned, treating the scouting system itself as something to debug and improve season over season.
5. Plan for second-pick ability. Beyond raw scoring, they derive how useful a robot is as a second pick, recognizing that the best alliance is built from complementary roles, not three copies of the highest-rated robot.

What to copy at your scale#

You do not need 1678's headcount or custom server to use the architecture:

• A rookie/mid team can replicate the shape with QRScout for stand collection, a Google Form for super-scout subjective notes, and a single aggregation spreadsheet acting as server + viewer.
• The principles, separate objective/subjective, one job per scout, automate aggregation, measure accuracy, evaluate second-pick fit, scale down perfectly to a paper-plus-spreadsheet operation.

Read their published whitepapers and open-source repositories at citruscircuits.org/scouting; they are written to be reused by other teams, and each season's release reflects the current game. The lesson is not "build exactly this"; it is "adopt the architecture and discipline, sized to your team."