In short
Fetch's cast rate measures the probability that your Commander deck's spells arrive on the right turn, based on a random draw from your library. Two independent signals summarize this: color (intra-deck median cast probability) and speed (average effective mana produced per turn). The method builds on the hypergeometric probabilities published by Frank Karsten (ChannelFireball), empirically calibrated on real decks. The bands that classify your deck are relative to your bracket, never absolute.
Your deck as a random bag
Every time you cast a spell, you draw at random from your library. The real question: does your deck contain enough lands of the right color so your spells arrive on time, turn after turn?
Frank Karsten, longtime writer on ChannelFireball, answered this question rigorously. His method calculates, for each spell in your deck, the probability of being able to cast it on the right turn. Fetch builds on this foundation and adds empirical calibration using real decks, bracket by bracket.
The result is two signals: color and speed. Each measures something independent.
Two independent signals
Color — for each spell, we calculate the probability that you have both enough lands and enough sources of the right color on the casting turn. The summarized signal is the median: is your typical spell castable? The median is robust — a single exotic WWWW creature doesn't tank the score for your whole deck.
Speed— how much mana your deck produces on average each turn, compared to a deck that just plays "1 land per turn" with no ramp. We use the mean because the median would be identical between a 36-land deck and a 40-land + 8-rocks deck — yet the actual mana speed is very different.
These two signals can diverge. A control deck is often consistent (its spells arrive on time) but slow (little ramp). A green stompy deck may be fast (full of dorks) but fragile on the white of a splash color.
To see both signals on a deck: Browse public decks →
What improves your signals
Four card families improve both your color and speed signals:
- Lands — the foundation. A land that taps for multiple colors (dual, triome) improves your color signal first. A basic Island does nothing for red mana. More lands means safer early turns.
- Rocks (mana artifacts) — Sol Ring, Arcane Signet, Fellwar Stone. They accelerate speed and help color if they tap for the right color. Note: colorless rocks (Sol Ring, Everflowing Chalice) only improve speed, not color.
- Dorks (mana creatures) — Llanowar Elves, Birds of Paradise, Selvala. Same benefit as rocks, with vulnerability to board wipes. Very powerful for accelerating speed as early as turn 2.
- Land tutors— Cultivate, Kodama's Reach, Farseek, Nature's Lore. They search for missing lands in your library. Double effect: more lands in play and fewer in the deck.
The Suggestions tab in the wizard automatically detects mana holes in your deck: Build your first deck →
Reading the bands
Fetch places your deck in one of five bands, calibrated on real decks from your bracket cohort:
- Weak — significantly below your cohort. Your spells will often come late.
- Fragile — below your bracket average. Fine on good draws, rough on bad ones.
- Correct — within the norm. Your spells arrive on time in most games.
- Solid — above average. Rarely delayed casts.
- Strong — top of your cohort. Near-perfect mana for your bracket.
Important: these bands are relative to your bracket. It is completely normal for a Bracket 3 (Upgraded) deck to have a weaker color signal than a Bracket 4 deck — B3 decks typically run fewer dual lands. You are being compared to other B3 decks, not some absolute ideal.
How to read the table: each figure is the upper bound of its band — a value strictly below it belongs to that band, a value at or above it moves to the next one.
Bracket 3
| Turn | Weak | Fragile | Correct | Solid |
|---|---|---|---|---|
| 1 | 64% | 72% | 79% | 88% |
| 2 | 74% | 78% | 82% | 86% |
| 3 | 73% | 80% | 85% | 88% |
| 4 | 74% | 79% | 84% | 90% |
| 5 | 76% | 82% | 86% | 91% |
| 6 | 77% | 85% | 89% | 93% |
| 7 | 80% | 88% | 92% | 94% |
| Reliable playability | 96% | |||
Bracket 1-2
| Turn | Weak | Fragile | Correct | Solid |
|---|---|---|---|---|
| 1 | 61% | 67% | 77% | 88% |
| 2 | 74% | 78% | 82% | 86% |
| 3 | 73% | 79% | 85% | 89% |
| 4 | 75% | 79% | 84% | 90% |
| 5 | 76% | 82% | 87% | 92% |
| 6 | 76% | 85% | 90% | 93% |
| 7 | 80% | 88% | 92% | 95% |
| Reliable playability | 96% | |||
Bracket 4-5
| Turn | Weak | Fragile | Correct | Solid |
|---|---|---|---|---|
| 1 | 64% | 70% | 77% | 85% |
| 2 | 71% | 76% | 81% | 85% |
| 3 | 71% | 77% | 83% | 87% |
| 4 | 72% | 77% | 82% | 88% |
| 5 | 75% | 81% | 85% | 89% |
| 6 | 76% | 84% | 89% | 91% |
| 7 | 78% | 87% | 91% | 94% |
| Reliable playability | 96% | |||
Calibration: where the bands come from
The Commander bracket color bands (1-2, 3, 4-5) are calibrated on the EDHREC 2026-09 measurement: 4 497 decks, drawn randomly and stratified by band across 10 commanders × 3 bands. The Brawl bands (Brawl 60 and Brawl 100) are calibrated separately, one scale per format, on 150 Moxfield decks per format.
The Commander collection is a randomized draw stratified by band over decks published on EDHREC, not a systematic draw of all published decks.
Representativeness caveat: the measurement only covers 10 commanders, whose verdict arbitrates the bands for every other commander — a narrow number, even with a stratified draw and 4 497 decks total.
Quantile convention: weak = first quartile (Q1), fragile = median, correct = third quartile (Q3), solid = 90th percentile (P90), and above = strong — computed on the intra-deck median of each cohort.
Speedbands are not in the table: the 2026-09 calibration finds them quasi-degenerate (values quantized to whole numbers, dominated by "1 land per turn"), with little power to differentiate between brackets. Publishing them would add noise, not verifiability.
The "Reliable playability"line in the table is not a fifth band: it is an absolute playability level (separate calibration, same 2026-09 collection) — the turn where the bracket's average playability curve crosses the "elbow" (T4, about 96% for the three brackets — see the "Reliable playability" row in the table). It does not read like the four columns to its left.
Brawl: one scale per format
Fetch also covers Brawl, in two formats: Brawl 60 (Standard Brawl, 60-card deck + commander) and Brawl 100 (Historic Brawl, 100 cards + commander).
The engine is the same as for Commander — same color and speed signals, same hypergeometric probabilities. What changes is the band scale, calibrated on real Brawl decks, per format.
No bracket in Brawl. The 1-5 system is an official Commander scale (WotC); Brawl has no equivalent. So there is a single scale per format: you are compared to all decks in the format, not to a power-level subgroup.
Acknowledged limit (D5).Direct consequence: a highly optimized deck and a casual deck in the same format are judged by the same scale. The band says "regular vs the format", not "regular for equal power". Keep that in mind when reading your Brawl band.
Inspiration and limits
The mathematical foundation comes from the method published by Frank Karsten on ChannelFireball between 2018 and 2022. It is the community standard for calculating how many mana sources a Commander deck needs.
Our contribution: empirical calibration using real decks collected by bracket (B1-2, B3, B4-5), which adapts Karsten's thresholds to each power level instead of applying a universal cutoff.
Acknowledged limits: board state is not modeled (land destruction, Stax…); variable mana sources (Cabal Coffers, Nykthos…) are flagged but not quantified; multicolor joint probability is approximated by assuming independent sources. These points are under consideration for future improvements.