Series: FRC Technical Foundations Unit: 5 — Making It Win Session length: ~3 hours, plus a full-day mock event if you can afford it Prerequisites: Lessons 1–9
There is a specific moment in most teams’ seasons where the robot first completes a full scoring cycle. It is genuinely exciting, and it is also the moment the real work starts, because a mechanism that works once is perhaps a third of the way to a mechanism that works in a match.
The arithmetic is unforgiving. A robot that completes a cycle successfully 90% of the time, attempting eight cycles per match across ten qualification matches, will fail roughly eight times. Each failure costs you the points from that cycle plus the time to recover. Two of those failures will happen in matches decided by less than that margin.
Ninety percent sounds good. It is not good. Competitive reliability is 98% and above, and getting there is entirely a matter of deliberate testing rather than talent.
Rule 10.1 — Reliability is measured, not felt. If you do not have a number, you do not have reliability. You have optimism.
Failure Mode and Effects Analysis sounds corporate. It is actually just a structured version of asking “what could go wrong and what should we do about it first?”
For each subsystem, list every way it could fail. Then score each one from 1 to 10 on three axes:
Multiply the three to get a Risk Priority Number. Sort descending. Work top-down.
The value is not in the arithmetic. It is in two things the process forces out:
Detectability is where the insight lives. A moderately likely, moderately severe failure that you would never see coming is far more dangerous than a severe one you can spot in a pre-match check. Teams that do this exercise properly usually end up adding sensors or checks rather than redesigning parts.
It surfaces disagreement. When the mechanical lead scores something 3 for occurrence and the drive team scores it 8, you have found something important: the drive team is seeing failures the builders are not hearing about.
Failure mode | Sev | Occ | Det | RPN | Action |
|---|---|---|---|---|---|
Intake chain derails | 7 | 6 | 8 | 336 | Add chain guard, check tension pre-match |
Battery not fully charged | 8 | 4 | 3 | 96 | Voltage check on pit checklist |
CAN wire loosens | 9 | 3 | 9 | 243 | Strain relief, add CAN status to dashboard |
Elevator belt snaps | 9 | 2 | 6 | 108 | Inspect every 50 cycles, carry spare |
Note the ranking. The intake chain and the CAN wire outrank the snapped belt, not because they are more severe but because you would not see them coming.
Adapted directly from the source curriculum, and the most valuable single exercise in this entire series.
Run the complete scoring cycle fifty times, consecutively, without uncontrolled maintenance, and log every single interruption.
Rules that make it work:
What teams find, essentially universally:
Every one of those findings is a match you would otherwise have lost.
Rule 10.2 — If you have not run 50 consecutive cycles, you do not know whether your robot works.
[ANECDOTE SLOT] — Something a long endurance test revealed that no amount of short testing would have. Thermal problems and cumulative loosening make the best examples because they are genuinely invisible over ten cycles.
Ask a team how long their scoring cycle takes and most will guess low by 30–40%. Ask where the time goes and most will guess wrong entirely.
Record a match — or a practice session — on video, and break one full cycle into components:
Component | What it is |
|---|---|
Travel out | Driving from scoring position to the game piece |
Acquisition | Intaking, including any missed attempts |
Travel back | Driving to the scoring position |
Alignment | Positioning accurately enough to score |
Mechanism | The scoring action itself |
Driver delay | Hesitation, decision-making, comms |
Do this for ten cycles and compute the mean of each component.
The result is almost always surprising. Teams expect the mechanism to dominate and it rarely does. Alignment and driver delay are usually the largest components, which is why swerve’s alignment advantage matters so much and why driver practice yields more improvement than most mechanical changes.
This should change what you work on. If alignment is 4 seconds of an 11-second cycle, an automated alignment routine or twenty hours of driver practice will do more for your ranking than making the elevator 20% faster.
Rule 10.3 — Improve the largest component, not the most interesting one.
Once the robot reaches basic functionality, every change competes with practice time. The source curriculum puts this well: after the robot works, every change must justify the testing and practice time it consumes.
A rule that works: from a fixed date — typically two weeks before your first event — no new features. Only reliability fixes, and every fix must be re-validated with cycles.
Drivers need hundreds of cycles to become good, and there is no substitute. A mediocre robot with a great driver beats a great robot with an untrained one, consistently and visibly, at every event.
Your pit is a repair system with a five-minute design requirement.
Layout. Tools in the same place every time, ideally shadowed or labelled. Batteries on a charging station with a clear charged/uncharged separation. A clear working area that two people can use simultaneously.
Spares. Driven by your FMEA. Anything with an occurrence score above about 4 should have a spare on the shelf, pre-built where possible. Pre-built subassemblies beat loose parts — swapping a complete intake assembly takes three minutes; rebuilding one takes thirty.
The pre-match checklist. Written, physical, and signed off by a named person every single match. Minimum:
Assigned roles. Under pressure, “someone should check the battery” means nobody does. Name the person.
Inspection is not adversarial and it is not hard, but failing it costs you practice matches.
Read the current inspection checklist in week one, not week six. The recurring items:
Run a mock inspection against the real checklist two weeks out. Every item that fails is a thing you would otherwise be fixing at 8 a.m. on competition day.
Time: 75 minutes. Equipment: Whiteboard or shared spreadsheet, every subteam present including drive team.
Build a full FMEA for your robot. Every subteam contributes failure modes for their own subsystem and scores someone else’s. Compute RPNs, sort, and identify the top five.
For each of the top five, decide: redesign it, add a check that detects it, carry a spare, or accept it. All four are legitimate answers; “accept it” is legitimate as long as it is a decision rather than an oversight.
Evidence of learning: A completed FMEA sheet, a spares list derived from it, and a maintenance interval schedule.
Time: Half a day. Equipment: Complete robot, field elements, multiple charged batteries, a logging sheet, a stopwatch.
Run it as described above. One student’s entire job is logging — nothing else. Log for each cycle: number, time, outcome, any anomaly, battery voltage at start.
Afterwards, produce:
Evidence of learning: The full log, the cycle-time plot, and a before-and-after comparison for the implemented fix.
Time: 60 minutes. Equipment: Video recording of driving practice, a stopwatch or video analysis software.
Break ten cycles into the six components in the table above. Compute means. Produce a stacked bar chart.
Then propose three improvements, each targeting a specific component, each with an estimated time saving and an estimated cost in build hours. Rank by saving per hour spent.
Evidence of learning: The component breakdown, the chart, and a ranked improvement list. This document should drive what your team works on next.
Time: A full day. Equipment: Everything. Field elements, multiple teams if you can arrange it, volunteer inspectors and judges.
Run a complete event simulation: inspection, judging interviews, practice matches, qualification matches with real match timing, alliance selection, and eliminations. Include a deliberate pit emergency — a mentor breaks something between matches.
Nothing else in this series prepares students for competition pressure the way this does. The technical problems are the smaller half; the logistics, communication, and time pressure are what actually catch teams out.
Evidence of learning: A completed inspection sheet, match logs, a pit repair log with times, and a written event action plan listing everything that needs to be fixed before the real event.
The end state is not a winning design. Designs are season-specific and they get copied, and a copied design teaches nobody anything.
The end state is a team that can build, wire, program, tune, test, repair, and explain a competition robot — and that can improve its own design using its own evidence. That capability transfers to next season, and to the season after, and long past robotics entirely.
The most successful students I have seen are not the ones who build the most complicated robots. They are the ones who test frequently, measure honestly, and refine small details. That is as true at 125 lb as it was at 2 kg.
If you are teaching this sequence, the single highest-value thing you can do is run Exercise 10.2 earlier than feels comfortable — even on a half-finished robot. Nothing else reveals the gap between a robot that works and a robot that competes.
