A gap analysis for NorCal garage teams ready to compete at the championship level
| Pillar | Team 25667 | Gap to Worlds | Priority |
| Fabrication | 5 / 10 — Good plastics, some metal, laser-cut frames. Needs more aluminum plate work. | 4 points — Significant but achievable with makerspace access | HIGH |
| CAD & Design | 6 / 10 — ~50% Onshape coverage. 3 iterations shows strong design cadence. | 3 points — Move to 100% assembly-first CAD before fabrication | MEDIUM |
| Auto & Software | 6 / 10 — Has Limelight 3A hardware; autonomous software framework unknown. | 3 points — Highest ROI: odometry pods + path following framework | HIGH |
| Sensors & Telemetry | 7 / 10 — Limelight 3A + AprilTag detection is world-class hardware. FTC Dashboard TBD. | 2 points — Close the loop: use AprilTags for relocalization in auto | MEDIUM |
| Strategy & Scouting | 5 / 10 — Alliance Captain at Regionals shows alliance selection intuition. No formal scouting system visible. | 4 points — Build a structured scouting system before next season | HIGH |
| Team & Process | 7 / 10 — Inspire Award winner at qualifier. Strong portfolio and outreach. Drive practice hours unknown. | 2 points — Quantitative testing data; 25+ drive practice hours | LOW |
Each layer depends on the one below it. You cannot effectively use AprilTag relocalization if you don't have a good position estimate to correct. You cannot run Road Runner or Pedro Pathing without odometry pods. Start from the bottom.
| Feature | Regional Level | Worlds Level | Team 25667 Status |
| Auto framework | Dead reckoning / timer-based sequences | Road Runner / Pedro Pathing | Unknown — key gap to resolve |
| Localization | Motor encoders only (drift accumulates) | Three-wheel odometry pods + kalman filter | Unknown — key gap to resolve |
| Vision / camera | Basic detection or none | Limelight 3A + AprilTag pose relocalization | ✓ Hardware already running |
| Autonomous structure | Sequential blocking commands | Event-driven state machine with fallbacks | Unknown — key gap to resolve |
| Teleop assists | Fully manual — driver controls everything | One-button macros for multi-joint sequences | Unknown — likely gap |
| Telemetry & tuning | Basic string logs to Driver Station | FTC Dashboard — browser-based, real-time graphs | Unknown — likely not set up |
| Driver practice | Under 5 hours on actual robot | 20+ hours of structured practice cycles | Unknown — needs measurement |
| AprilTag relocalization in auto | AprilTags not used, or used for alignment only | Continuous pose correction during auto run | Hardware present — software use TBD |
Worlds-level portfolio work is maintained throughout the season — not assembled in October before competition. Each iteration should document: the failure mode that motivated it, the design alternatives considered, the selection criteria used, and the measured improvement after implementation. The Inspire judges at championship have reviewed hundreds of engineering notebooks; quantitative data is the differentiator.
Drive practice is the biggest free upgrade available to any FTC team. Worlds-level drive teams log 20–30 hours of structured practice on the actual competition robot before their first qualifier. This is not free driving — it is deliberate practice with metrics tracked.
| Practice Metric | What to Track | Worlds Target |
| Cycle time | Time from scoring one game element to scoring the next — average and standard deviation | < 5s variance, consistent rhythm |
| Auto consistency | % of auto runs that hit target score (run 20+ times per configuration) | > 85% hit rate on target auto |
| Teleop score/match | Track your raw teleop contribution in every practice match | Top-quartile for your game position |
| Defense recovery | Practice with a defender actively blocking. Track score delta under defense. | < 15% score reduction under defense |
| Failure recovery | Practice specific failure scenarios: auto miss, element drop, mechanism jam | Driver has a rehearsed response to each |
Three robot iterations in one season already matches worlds-level teams. The critical discipline: each iteration must have a specific measured failure mode from the previous version that motivated it, not a vague "make it better." Document V1's failure modes → set V2's goals → measure V2 against V1 → document. Repeat. This creates the quantitative narrative that wins Inspire awards at championship level.
Alliance Captain at NorCal Regionals shows the team already has competitive instincts for alliance selection. The upgrade is a structured scouting system that makes those decisions data-driven. Before the next event, build a scouting app — Google Sheets works fine — that tracks per-match for every robot observed:
| Scouting Field | What to Record |
| Auto performance | Score achieved, consistency (did it work?), which route |
| Teleop scoring | Estimated elements scored, scoring position, cycle speed |
| Endgame | Climb level achieved, reliability |
| Robot breakdown | Any visible mechanical failures, recovery time |
| Defensive capability | Can they play defense? Did defense stop them? |
| Alliance fit | Would they complement Team 25667's strengths? |
FRC teams commonly build two full robots per season: a competition bot and a practice bot. FTC teams rarely do this. The reasons to consider it:
Game piece flow failures — wherever your robot collects, transfers, and delivers game elements — are the most common class of mechanism problem at every competitive level, and one of the most preventable. Better CAD process catches these before the robot is built. Here's how.
| Tool | What It Simulates | FTC-Relevant Use | Cost |
| Onshape mate animation | Range of motion, interference | Full robot assembly, verify arm reach | Free |
| Onshape with game piece model | Ball/sample travel path | Game piece clearances through collection and transfer path | Free |
| Blender rigid body | Physics-based game piece rolling/falling | Game piece retention in delivery mechanisms, collection funneling | Free |
| Fusion 360 motion simulation | Forces under load, motor torque required | Transfer motor sizing, arm counterbalance | Free (edu) |
| Fusion 360 FEA | Stress in parts under load | Aluminum bracket failure point | Free (edu) |
Currently the team's design direction is heavily influenced by one or two members — whoever speaks first or with the most confidence drives the concept. This works until it doesn't: good ideas from quieter members get filtered out, and the team converges on the first plausible idea rather than the best one. Worlds-level teams use explicit processes so that design decisions are made on merit.
Use a Pugh matrix to compare mechanism concepts objectively. Rate each concept against criteria: +1 = better than baseline, 0 = same, -1 = worse. Multiply by weight and sum. Example below for a scoring mechanism decision:
| Criteria | Weight | Concept A (Flywheel) | Concept B (Catapult) | Concept C (Direct Push) |
| Scoring consistency | 3× | +1 | 0 | −1 |
| Cycle time | 3× | +1 | −1 | +1 |
| Mechanical complexity | 2× | −1 | +1 | +1 |
| Recovery from jam | 2× | 0 | −1 | +1 |
| Build feasibility | 1× | 0 | −1 | +1 |
| Weighted total | +4 | −4 | +5 |
+1 = better than baseline, 0 = same, -1 = worse. Weight is the multiplier applied to each score before summing.
Teams consistently use what they already know. This is not laziness — it is a natural consequence of having no structured process for discovering what else exists. The gap between a regional team and a worlds team is often not the ideas they had but the options they considered. A team that knows goBILDA makes a 5-turn dual-mode servo with 1800° of rotation will design fundamentally differently than one that thinks 300° is the limit — and both teams are reading the same catalog.
Describe your mechanical challenge to an AI (Claude, ChatGPT, Gemini) and ask what components, mechanisms, or approaches might work. AI consistently surfaces options the team has never considered — from vendors, mechanism types, and software patterns.
Example prompts that work well:
Caution: AI does not know current pricing or availability. Treat suggestions as a starting point for vendor research, not a final parts list.
Try: claude.ai · chatgpt.com
Once per off-season, spend one session per major vendor reading their full product catalog — not just the parts you usually order. This is where you discover that goBILDA makes 5-turn servos, that REV has a laser distance sensor with 4m range, that Axon makes a 360° continuous servo with feedback.
Vendors to review annually:
Each review: assign one vendor to one team member. Present a one-page “what’s new / what we didn’t know existed” to the full team.
Many worlds-level FTC teams share their full Onshape CAD publicly. Opening their assemblies shows you exactly which components they used, how they mounted things, and what vendor parts they chose. This is the most direct way to discover “I didn’t know that part existed.”
| Use Case | Example Prompt / Application |
| Equipment discovery | “What FTC-legal sensors can detect a game piece in an intake without stopping robot motion?” |
| Mechanism selection | “I need to score at 36 inches with one motor — what mechanism approaches have teams used?” |
| Code review | “Here’s my autonomous OpMode — what edge cases am I not handling?” |
| Game analysis | “Here are the point values for each game task — rank them by expected points per 30 seconds of cycle time” |
| Portfolio writing | “Help me structure the design alternatives section showing why we chose mechanism A over B and C” |
| Troubleshooting | “My Road Runner auto overshoots by 2 inches to the left after this specific action — what would cause this?” |
| Driver practice planning | “Design a 20-minute driver practice session structure focused on improving cycle time consistency” |
| Pre-season scouting | “What mechanism approaches are likely to score most efficiently given these game tasks and point values?” |
Having good discovery methods means nothing without assigning someone to do them. At any team session, some members are actively building or coding while others are waiting — for a print to finish, for a mechanism to be ready, for code to deploy. That waiting time is currently lost. It should be structured.
Every team member has a defined task at every session — either on the robot, or on the knowledge board. Rotate assignments each session so everyone covers different methods over time.
| Assignment | What to do during the session |
| Catalog reviewer | Read one assigned vendor catalog fully. Document 3–5 components the team has never used. Add to the shared discovery doc. |
| CAD studier | Open a top team's public Onshape assembly. Identify components you don't recognize. Look each one up and document it. |
| AI consultant | Run a structured AI session on a current design challenge. Document suggestions and flag which ones are worth investigating further. |
| Community monitor | Check FTC Discord #mechanism-design and #hardware. Post open questions. Summarize anything relevant to current design challenges. |
| Scrimmage coordinator | Research nearby teams on ftc-events.firstinspires.org. Draft and send outreach. Track responses in the shared doc. |
Session structure: 5-min kickoff — research assignees state what they're working on → build and research run in parallel → 5-min close where discoveries are shared. A shared Google Doc captures findings across sessions. Discoveries feed the next design review.