Build a deck, run thousands of solitaire games on the play and on the draw, and see how many Rats hit play and how fast Rats, Black Vise and Ankh deal 20. Deal single hands to watch them play out, read the case for the best build, and see exactly how the simulated player makes its choices.
Starting…
Averages over all games, including damage past 20. The dashed line marks 20.
Average damage already dealt by each source when the total reaches 20, grouped by the turn it happens. The bar shows the Rat / Vise / Ankh share.
Each run is added here so you can compare builds side by side. Average Rats in play, on the play / on the draw.
Three Black Vise, three Ankh of Mishra and three Sinkhole alongside Sol Ring, Demonic Tutor and fifteen Plague Rats. It gives up about a third of a turn of goldfish speed against the all-in Rat deck. In exchange about a fifth of its lethal damage on the play (a ninth on the draw) comes from cards that creature removal, blockers and Control Magic can't touch, at least two of its three damage sources (Rats, Black Vise, Ankh) are in play at the end of turn 4 in about two games out of three, and it attacks the opponent's mana on turn 2.
Headline numbers for the recommended build from 1,000,000 games per seat, shown as on the play / on the draw.
The deck has three kinds of damage source: Plague Rats, Black Vise and Ankh of Mishra. At the end of your turn 4 we look at which kinds you have at least one of in play. "2+ kinds" means at least two different kinds, for example a Rat and a Black Vise, so no single answer (creature removal, a blocker, an opponent who empties their hand) shuts off all of your damage. Sol Ring and Sinkhole are not damage sources and don't count. The table shows every combination for the recommended build.
200,000 games per build and seat, goldfish opponent, same play rules and damage model as the Simulator tab. Best value in each row in bold.
With n Rats attacking, each is n/n, so a full swing deals n × n. Damage grows with the square of your Rat count, and so do the losses.
Losing one Rat out of four costs 44% of that turn's attack, not 25%. A Rat stolen by Control Magic is worse still: it keeps counting toward the size of every Rat, so the opponent gets a blocker exactly as big as yours.
| Interaction | Plague Rats | Black Vise | Ankh of Mishra |
|---|---|---|---|
| Spot removal (Lightning Bolt, Swords to Plowshares, Paralyze) | Hit | Unaffected | Unaffected |
| Blockers | Hit | Unaffected | Unaffected |
| Control Magic | Hit twice | Unaffected | Unaffected |
| Creature sweepers | All lost | Unaffected | Unaffected |
| Emptying their hand fast | Unaffected | Shuts it off | Unaffected |
| Stopping at few lands | Unaffected | Unaffected | Less damage |
No single answer stops all three. That is the case for spreading damage across them.
Six slots of Vise and Ankh replace six Rats, yet the deck is no slower on the play because the artifacts do their work on turns 1–4, exactly when Rats can't: a turn-1 Vise and a turn-2 Ankh are worth up to 6 damage each on the play. On the draw both are weaker (the opponent has already played a land and their hand is smaller), so they cost about 0.1 turn there.
Each Sinkhole deals no damage and, under the play rules, takes the whole turn-2 mana ahead of Ankh and Tutor. That is why it is the one card that costs speed in the goldfish. What it buys is time: a turn-2 Sinkhole takes the opponent back a land (to zero on the play, to one on the draw), which delays four-mana answers such as Control Magic and sweepers by a turn and can cut them off a color. The simulator does not model that benefit, so how many Sinkholes to play is a judgment about how interactive your opponents are.
Rerun at 1,000,000 games per build and seat, so average kill turns are accurate to about ±0.003 and percentages to about ±0.1 points. All three share the same 11 non-land, non-Rat cards; only the Swamp / Rat split changes. Bold = best of the three, red shading = worst, green outline = recommended.
Gap = distance from the best of the three on each measure, as a share of the best-to-worst spread (0 = best, 1 = worst), averaged over all measures. The same result holds on the play and on the draw.
If you weight goldfish speed and mana consistency above everything, 15 Swamps is the pick. The case for 14 is that interaction turns fast games into longer ones where a spare Rat matters, and 14 Swamps buys that reserve at the smallest cost in speed and stalls.
Every game is played by the same fixed set of rules, with no look-ahead and no randomness in its choices. The rules are a greedy plan built around one goal: get Plague Rats into play as early and as often as possible, and spend any mana a Rat can't use on the other cards. They are the rules you specified; the page follows them exactly, step by step, in the order below.
A 1-mana turn is almost always turn 1 without Sol Ring: Swamp, Black Vise.
Sinkhole needs both mana to be black, so with Swamp + Sol Ring's colorless it is skipped and Ankh comes down instead.
Then spend what is left, repeating down this list until nothing fits:
"Spend what is left" restarts from the top of the list after every cast, so 4 mana with a Rat and a Black Vise becomes Rat + Vise, and 5 mana with a Rat and a Sinkhole becomes Rat + Sinkhole.
Tutor looks at two candidates, Sol Ring (if it is still in the library) and a Plague Rat, and scores each one in this order:
The library is shuffled after Tutor, as the card says.
| Situation | Fetches |
|---|---|
| Turn 2, two Swamps, Rats in hand, no Sinkhole | Sol Ring (cast next turn) |
| Turn 2, two Swamps, no Rats, no Sinkhole | Plague Rats |
| 5 Swamps, no Rats in hand | Plague Rats, cast it now |
| 3 Swamps, Rat in hand, Sol Ring in library | Rat is cast first; Tutor needs 2 more, so it waits |
| Card | Cost | Notes |
|---|---|---|
| Plague Rats | 2B | needs 1 black |
| Sinkhole | BB | needs 2 black |
| Demonic Tutor | 1B | needs 1 black |
| Sol Ring | 1 | any mana |
| Black Vise | 1 | any mana |
| Ankh / Chaos Orb | 2 | any mana |
Generic costs are paid with Sol Ring's colorless mana first, keeping black mana free for the next Rat or Sinkhole. That is why one Swamp plus Sol Ring (1 black + 2 colorless) is exactly one Rat.
One catch: with a single Swamp, casting Sol Ring that turn uses the only black mana, so no Rat is possible until the next turn.
| Board and hand | Play |
|---|---|
| T1: Swamp, Sol Ring, Ankh | Swamp → Sol Ring → Ankh (2 colorless) |
| T1: Swamp, Black Vise, no Sol | Swamp → Black Vise |
| T2: 2 Swamps, draw Sol Ring, Rat in hand | Sol Ring → Plague Rats |
| T2: 2 Swamps, Sinkhole and Ankh | Sinkhole |
| 2 Swamps + Sol Ring, Rat and Vise | Rat (B+CC) → Vise (B) |
| 3 Swamps + Sol Ring, Rat and Sinkhole | Rat (B+CC) → Sinkhole (BB) |
| 5 Swamps, Tutor, no Rats | Tutor → fetch Rat → cast it |
The opponent starts at 20 life and does nothing but play lands. Damage is counted by round: your turn plus the opponent's turn that follows it. The "turn total damage reaches 20" is the round in which the running total of all three sources first hits 20.
Each turn, the Rats that were already in play at the start of your turn attack. Each one is X/X, where X is the number of Rats in play after this turn's casting, because new Rats pump the attackers even though they can't attack yet.
damage = attackers × Rats in play
Damage carries over from turn to turn. No blockers.
The opponent keeps a 7-card hand at the start and shrinks it by one card per turn, so each Vise deals 3, then 2, then 1, then 0 at their upkeeps.
A Vise cast on turn 1 deals 3 on its first trigger on the play, but only 2 on the draw, because the opponent has already taken their first turn.
Each Vise counts separately.
Each Ankh deals 2 whenever the opponent plays a land, and the opponent stops at 4 lands. Maximum 8 per Ankh.
Cast on turn 2 on the play it catches lands 2, 3 and 4 (6). On the draw it catches lands 3 and 4 (4).
Ankh damage to you from your own Swamps is not tracked.
It is the best play under the rules you set, applied the same way every game, which is what makes builds comparable. It is a fixed, greedy policy, not a search for the best possible line, so a few spots could be played better in a real game:
To test a different policy, change the priority lists or the rule functions in the engine (they are named after each rule), or ask me to add a switch for an alternative rule.