Magic Estimation for Agile Teams
The card-stacking variant of planning poker that lets your team size 50, 100, even 200 user stories in a single session — almost entirely without discussion.
What Is Magic Estimation?
Magic estimation — sometimes called the magic estimation game, speed estimation, or silent estimation — is an agile estimation technique that uses the same Fibonacci scale as planning poker but trades per-story discussion for raw throughput. It is designed for the moment when your backlog has grown faster than your refinement capacity and you need to size 100+ stories before next quarter's PI planning.
The mechanic is brutally simple. Lay out the Fibonacci sequence (1, 2, 3, 5, 8, 13, 20, 40, 100) on a wall. Hand every team member a pile of stories. They silently place each story under the Fibonacci value that feels right. Anyone who disagrees with a placement silently moves the story. Two or three silent rounds, then a short discussion round on the contested stories. That is the entire process.
Magic estimation works because the silent placement removes anchoring bias even more aggressively than planning poker does. Nobody hears anyone's opinion before forming their own. And because the team only discusses stories where there is genuine disagreement, you skip the 80% of stories everyone already agrees on — which is exactly where planning poker spends most of its time.
When to Use Magic Estimation
Magic estimation is the right tool when speed matters more than precision. Specifically:
- Quarterly or PI planning where you need to size every story before kickoff
- A new backlog dump from product that nobody has estimated yet
- After a major refactor that invalidates older estimates
- Onboarding a new team to an existing product with hundreds of refined stories
- Any time planning poker would take 6+ hours and the team would lose energy
It is the wrong tool when the team needs the discussion that planning poker generates. If your stories are about to be sprint-committed, run them through planning poker — the per-story conversation surfaces hidden complexity that magic estimation deliberately skips.
How to Run a Magic Estimation Session
1The silent placement round
Lay out the Fibonacci scale on a long table or wall: 1, 2, 3, 5, 8, 13, 20, 40, 100. Hand every team member a stack of story cards. They silently place each story under the Fibonacci value they think fits. No talking, no debate. Timebox: 15 minutes for ~50 stories.
2The silent re-shuffle round
Now the team walks the wall and silently moves any story they disagree with. A story might move from 5 to 8 because someone with backend context thinks the database work is bigger than the original placer realized. Or a story slides down to 2 because someone has built it before. Still no talking. Another 10 minutes.
3The discussion round (only outliers)
After two silent rounds, identify stories that have moved more than once or that anyone wants to flag. Discuss only those — five minutes per story max. The rest of the placements stand. The team trusts the silent rounds to handle the easy cases.
4The product owner pull-aside
The product owner pulls anything that looks suspiciously sized — a "100" that should be split, a "1" that hides scope. Stories that need clarification go back to refinement, not the magic estimation board.
5Lock the estimates
Whatever Fibonacci column a story ends up under is its estimate. Capture them in your backlog tool. The team has just sized 50–200 stories in under an hour.
Magic Estimation vs Other Techniques
Magic estimation sits in the middle of the agile estimation toolbox. It is faster than planning poker, more precise than affinity estimation, and works at scales planning poker simply cannot reach in a normal-length session.
| Technique | Stories / hour | Precision | Discussion | Best for |
|---|---|---|---|---|
| Planning Poker | 5–15 | High | Per-story | Sprint-level commitment, refined stories |
| Magic Estimation | 50–200 | Medium | Outliers only | Quarterly planning, large backlog dump |
| Affinity Estimation | 20–100 | Low | After silent grouping | Brand-new backlog, T-shirt buckets |
| Bucket System | 40–150 | Medium | During placement | Distributed teams, async sessions |
Tips for a Successful Session
Set up reference stories first. Before placing the new stories, the team should agree on one or two well-understood reference stories per Fibonacci value. These anchors prevent the wall from drifting. A reference 5 keeps everyone's mental model of “5” calibrated.
Keep stories printable. One title per card. The team should be able to read a story in two seconds and place it. If a story needs three paragraphs to understand, it is not ready for magic estimation — send it back to refinement.
Limit the silent rounds to 2–3. Diminishing returns kick in fast. After three silent rounds, the placements stabilize and additional rounds just create ping-pong without new information.
Use a stopwatch. Timebox the discussion round to 5 minutes per story. Without a timer, outlier discussions sprawl and the speed advantage evaporates.
Capture the result, then run planning poker before sprint commitment. Magic estimation gives you a roadmap-quality estimate. Planning poker gives you a sprint-quality commitment. Use ScrumChamps planning poker on the next-up stories before you commit them into a sprint.
Right tool, right backlog stage.
Magic estimation for the long view. Planning poker for the next sprint. Pair them and you have an agile estimation toolkit that scales from solo backlog to PI-wide planning.