TLDR
If you are learning how to goldfish Commander deck lists, play 10 uninterrupted solo games under consistent conditions. Mulligan normally, sequence honestly, and record land drops, color access, ramp, draw, commander timing, interaction readiness, win setup, and dead cards. Look for repeated failures rather than reacting to one bad hand. Goldfishing can expose structural problems, but it cannot measure resilience, politics, threat assessment, or performance at a real multiplayer table.
Goldfishing is a diagnostic rehearsal, not evidence that your deck is powerful. Its job is to answer a narrower question: when nobody interferes, does the deck reliably produce the mana, resources, and sequencing its plan requires? MTGEDH’s playtesting guidance likewise recommends trying multiple hands, recording repeated failures, and validating revisions through real games rather than treating solitaire results as final proof.
Commander is fundamentally multiplayer, and Wizards’ official Commander format overview emphasizes matching expectations before play. That context matters because a clean solo sequence says nothing about whether three opponents will consider your engine threatening, remove your commander, race your setup, or force you to spend mana defensively.
Set consistent test conditions first
Use the actual 100-card deck you intend to play, including its commander. Shuffle thoroughly before every game and avoid arranging a convenient opening sequence. If you know the next card after ending a test, reshuffle before starting another one.
Define the deck’s intended environment before recording results. A battlecruiser deck, upgraded precon, high-power combo deck, and cEDH deck should not be judged against the same clock. Your chosen endpoint should give the deck enough time to demonstrate its normal plan without producing pages of irrelevant late-game solitaire.
Turn eight is a reasonable starting endpoint for many broad casual tests, but it is not a universal benchmark. A faster deck may need a shorter window focused on early setup and interaction. A slower deck may need additional turns to show whether it can transition from development into a credible finish. Keep the endpoint unchanged throughout each 10-game batch so the results remain comparable.
Alternate between treating yourself as the starting player and as a later-position player. This is not an official testing standard; it is simply a way to avoid building every observation around one play position. Follow the current applicable Commander and mulligan rules rather than relying on remembered wording. Wizards notes that the Comprehensive Rules are updated with set releases, making current official material the appropriate authority for detailed procedures.
How to goldfish Commander deck lists in 10 games
Run each game as though your decisions matter. The point is not to demonstrate the strongest possible sequence after seeing future draws. It is to make reasonable choices with the information available on that turn.
- Shuffle and draw an opening hand.
- Evaluate the hand against the deck’s real plan, then mulligan as you would in an actual game under the current applicable rules.
- Record the kept hand’s lands, usable colors, early plays, ramp, card selection or draw, interaction, and expensive payoffs.
- Play one turn at a time without looking ahead. Use normal land drops and pay all costs.
- Sequence lands, ramp, draw, and setup cards according to what you know at that moment.
- Record milestones when they happen, including missed land drops, commander deployment, interaction readiness, and engine setup.
- Stop at the predefined turn or when the deck demonstrates its uninterrupted win setup.
- Write one diagnosis for the game before shuffling again.
- Repeat until you have 10 completed games.
- Review patterns only after the batch is complete.
Evaluate the opening hand instead of counting lands alone
A hand with three lands is not automatically functional. Ask whether those lands cast the spells that matter. A three-color deck with three lands but only one usable color may be worse than a two-land hand containing fixing, cheap selection, and an appropriate ramp spell.
Record both total lands and functional mana. Note when each required color becomes available, whether tapped lands delay your sequence, and whether a ramp spell can actually be cast on schedule. If color access repeatedly fails, inspect the deck’s land counts, color sources, and fixing rather than automatically adding another generic land.
Judge mulligans by plan access, not by whether every card is individually good. A keepable hand usually needs a plausible first few turns: mana, something relevant to do, and a route toward the deck’s engine or commander. An opening hand filled with five- and six-mana payoffs may contain powerful cards while still being a poor keep.
Log how many mulligans you take and which kinds of cards are returned or placed on the bottom. If seven of 10 opening hands contain too many expensive payoffs, that pattern says more than one spectacular game in which you curve out perfectly.
Sequence mana, draw, and the commander honestly
Choose the line that most reasonably enables the next relevant turn. Sometimes that means playing a tapped fixing land before a basic. Sometimes it means casting selection before ramp because you are missing the next land. In another hand, ramping first may unlock the commander and leave the draw spell for later.
Record the sequence, not merely the result. “Turn-two ramp, turn-three draw spell, turn-four commander” is more useful than “commander on turn four” because it shows what resources made the cast possible. If early acceleration consistently determines whether the deck functions, compare the observed hands with an appropriate Commander ramp package rather than assuming one extra mana rock solves everything.
Track the commander’s first actual cast turn. Do not write down when it could have been cast under a different sequence or with a favorable topdeck. If the commander is central to the deck, also note whether you could afford a recast after a hypothetical removal spell. This does not simulate disruption, but it reveals whether the mana base has any recovery margin.
Card advantage deserves similar scrutiny. Note the first turn you cast a draw or selection spell, how many cards it accesses, and whether you had to choose between refilling and advancing the board. A deck that empties its hand quickly but cannot replenish it may need a different mix of repeatable engines and burst draw. The appropriate starting package depends on curve and commander access, as discussed in this guide to card draw counts in Commander.
Track meaningful interaction even without a target
Goldfishing cannot tell you whether a removal spell will hit the correct permanent, but it can show when you could realistically hold it open. Record the first turn on which interaction was available without abandoning your entire development plan.
“Meaningful interaction ready” should reflect actual casting requirements. A counterspell is not ready if you lack the required color. A four-mana removal spell is technically castable on turn four, but holding it may be unrealistic if the deck must deploy its commander that turn to function. Free interaction, flexible removal, and modal cards should be recorded according to their real conditions rather than treated as universally available.
Do not invent an opponent’s board and declare that every answer was perfectly timed. Instead, write observations such as “single-target removal available on turn three” or “board wipe castable on turn five.” Target selection and timing belong in multiplayer testing, where Commander threat assessment involves incomplete information and competing loss conditions.
Record win setup without calling it a real win
Identify the point at which the deck assembles a credible engine, combat setup, lock, or combo line. Record the required pieces, mana, and cards spent. Use “uninterrupted win setup on turn seven,” not “this deck wins on turn seven.” The first statement describes the test; the second implies resistance and table conditions the test never included.
For combat decks, producing a large board is not the same as defeating three opponents. For combo decks, assembling a line is not the same as protecting it through interaction. For value decks, drawing many cards is not itself a win plan. Goldfishing should help you determine whether the deck reaches its intended payoff and whether that payoff can plausibly close a game, but opponents must test the rest.
Use this 10-game notes table
Keep each entry short. The goal is a comparable dataset, not a turn-by-turn match report. Combine related milestones in one cell if necessary.
| Game | Mulligans and keep | Mana and land drops | Commander and sequence | Interaction and win setup | Dead cards and diagnosis |
|---|---|---|---|---|---|
| 1 | |||||
| 2 | |||||
| 3 | |||||
| 4 | |||||
| 5 | |||||
| 6 | |||||
| 7 | |||||
| 8 | |||||
| 9 | |||||
| 10 |
In the mana column, record the first missed land drop, when all necessary colors came online, and the first ramp spell. In the commander column, include its first cast turn and the important ramp or draw sequence that enabled it. In the interaction column, record when a relevant answer could be held open and when the engine or win setup appeared.
A dead card is not simply a card you did not cast. Mark why it was stranded: wrong colors, too expensive, missing a prerequisite, redundant with another effect, irrelevant without an opponent, or weaker than the available line. A board wipe sitting in hand during an empty solitaire board is not automatically dead; a seven-mana payoff repeatedly stranded while the deck misses its fifth land may indicate a genuine structural mismatch.
Turn repeated observations into targeted revisions
Commander deck construction is easier to diagnose through functional roles such as mana, card advantage, interaction, enablers, payoffs, protection, and recovery. After 10 games, group failures by those roles instead of cutting whichever card annoyed you most recently.
| Repeated observation | Question to investigate | Possible revision direction |
|---|---|---|
| Early land drops repeatedly missed | Is the mana count too low, or are expensive hands being kept? | Add mana sources, lower the curve, or improve mulligan decisions. |
| Colors arrive too late | Does the deck have enough untapped sources and fixing for early costs? | Adjust color sources or replace slow lands. |
| Ramp appears but does not help | Is its cost or color requirement mismatched with the curve? | Use cheaper or better-aligned acceleration. |
| Commander consistently arrives late | Is the commander too expensive for the current mana package? | Add suitable acceleration or reduce dependence on an early commander. |
| Hand empties without recovery | Are draw effects too expensive, conditional, or sparse? | Add efficient draw or repeatable card advantage. |
| Interaction cannot be held open | Are answers too expensive or color-intensive? | Choose cheaper, broader, or modal interaction. |
| Payoffs are repeatedly stranded | Are there too many top-end cards or too few enablers? | Trim redundant finishers or strengthen setup. |
| Deck functions only with its commander | Can the core plan operate after removal? | Add redundancy, protection, or independent engines. |
Change one coherent package at a time. For example, replace two slow lands and one colorless utility land with better color access, then repeat the same 10-game protocol. If you simultaneously change the mana base, curve, draw suite, interaction, and win condition, you will not know which revision caused the difference.
When expensive cards or unowned revisions are involved, test the proposed swap slots before buying them. If the next step is producing physical test cards, printed playtest copies from ProxyFoundry can support that stage. Keep the strategic question first: you are testing whether the revisions improve repeatable deck function, not whether one favorable draw feels exciting.
What goldfishing cannot tell you
A deck can look excellent in solitaire and struggle badly in a pod. Goldfishing removes the forces that make Commander a multiplayer game.
- Disruption: You do not learn whether the deck rebuilds after removal, counters, graveyard hate, or a board wipe.
- Politics: You cannot test deals, temporary alliances, attacks based on table perception, or the consequences of becoming the archenemy.
- Threat assessment: There are no competing engines or loss conditions demanding that you identify the most urgent target.
- Combat incentives: Solo testing cannot reproduce blockers, crack-back risk, life-total pressure, or opponents choosing to attack elsewhere.
- Interaction timing: Holding a removal spell is different from choosing whether to spend it now or risk losing later.
- Real pod pace: An uninterrupted setup turn does not establish that the deck can keep up with a particular group.
- Power and expectation fit: Goldfish speed does not determine whether a strategy, combo, tutor package, or lock is welcome at the table.
Do not use a clean goldfish record to claim a power level, bracket fit, expected win turn, or win rate. Wizards presents Commander as a multiplayer format in which pregame alignment helps match expectations. Use a direct Commander pregame conversation to establish table fit, then use actual games to evaluate performance.
The practical stopping rule
Stop goldfishing when you can describe the deck’s recurring uninterrupted pattern and have a specific revision to test. Ten games are enough for an initial diagnostic batch, not enough to prove a universal conclusion. If the results are inconsistent, run another unchanged batch before editing the list.
Your final notes should sound restrained: “In 10 uninterrupted tests, the deck missed its fourth land drop four times and held castable interaction by turn three twice.” That observation supports a focused mana or curve review. It does not prove how the deck will perform against opponents.
Make one revision package, rerun the same protocol, and compare the new observations with the old ones. Once the deck can execute its basic plan with reasonable consistency, move to real games. That is where you test resilience, target selection, politics, pacing, and whether the deck belongs at the table for which it was built.