Planning poker cards and which deck to use
The deck you choose shapes the conversation. Numbers invite precision, letters invite ranges, and the gaps between values decide how much argument a single story can generate.
Why the Fibonacci sequence
The classic deck runs 0, 1, 2, 3, 5, 8, 13, 21. The gaps grow as the numbers grow, and that is the entire point. Nobody can reliably tell the difference between a 20-point story and a 21-point story, so a scale that offers both invites a pointless argument. By the time you are at 13, the honest options are 13 or 21, and picking between them is a real decision about scope.
The practical effect is that large numbers become uncomfortable, which is healthy. A story that lands on 21 is usually telling you it should have been three stories.
Comparing the decks
Fibonacci
Sprint-level story points
- 0
- 1
- 2
- 3
- 5
- 8
- 13
- 21
- ?
- ☕
Gaps widen as numbers grow, which matches how estimation accuracy falls off with size.
Modified Fibonacci
Teams estimating a wide range
- 0
- ½
- 1
- 2
- 3
- 5
- 8
- 13
- 20
- 40
- 100
- ?
- ☕
Adds a half point at the bottom and 40 and 100 at the top for epics that need a rough figure.
T-shirt sizes
Early or rough sizing
- XS
- S
- M
- L
- XL
- XXL
- ?
- ☕
Removes the false precision of numbers, which keeps a first pass fast.
Sequential
Small, similar tasks
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- ?
- ☕
Even steps suit work where items genuinely are comparable in size.
Powers of two
Infrastructure and spikes
- 1
- 2
- 4
- 8
- 16
- 32
- 64
- ?
- ☕
Doubling forces a real decision at each step instead of splitting hairs.
When T-shirt sizing beats numbers
T-shirt sizes work when the team is estimating something it does not understand well enough to put a number on. Early roadmap work, a backlog nobody has groomed, or a discovery phase all fit. Sizes carry an honest amount of information: bigger, smaller, about the same.
The trade-off is that sizes do not add up. You cannot sum a sprint of M and L stories into a velocity figure. Most teams start on T-shirts during discovery and switch to Fibonacci once the work is understood well enough to plan a sprint around it, and the full trade on precision, speed and forecasting is set out in planning poker vs T-shirt sizing.
The two cards people forget
Every deck here includes a question mark and a coffee cup. Both matter more than they look.
The question mark means “I do not have enough information to estimate this”. Without it, an uncertain voter picks a middle number to avoid looking lost, and that guess gets averaged in as though it were a real opinion. SprintsPoker counts question marks towards agreement but leaves them out of the average, because averaging an admission of uncertainty produces a fake number.
The coffee cup asks for a break. Estimation quality falls off sharply after about forty minutes, and giving people a socially easy way to say so is cheaper than the bad estimates you get from a tired room.
Choosing one and sticking to it
The worst outcome is switching scales mid-backlog. Points from a Fibonacci sprint and points from a sequential sprint are not comparable, so your velocity history stops meaning anything. Pick a deck, use it for a quarter, and only change it deliberately. Keep it consistent across formats too: the printable planning poker cards use these same five decks, so a paper deck on the table and a room in the browser never end up on different scales.
For what the numbers themselves should represent, read the story point estimation guide.