How it works

Deck in, draw odds out

3,720:1 answers one question: how often does a deck deliver the cards you want, by the turn you want them? It shuffles your deck thousands of times, deals out the draws, and counts how many of those games met your target. Load a deck, describe the target, run.

1 · Load a deck

The header holds everything about the deck: name, leader and base art, aspect pips, total and unique card counts. A bundled sample deck loads on first visit so you can experiment straight away.

swudb deck — paste a share URL into the box at the top left and press Load (or hit Enter). Either form works: https://www.swudb.com/deck/bnuCUeHzSnGgw or just the bare id bnuCUeHzSnGgw.
melee.gg decklist — the same box takes a tournament decklist URL, e.g. https://melee.gg/Decklist/View/2570eda7-3a7b-480e-8602-b4980186ac7a (or just the id). melee lists cards by printed name rather than by set number, so each one is matched against the swu-db card database on import; reprints resolve to their earliest printing, which can change the art but never the cost or the odds. The list has to be publicly viewable.
swumetastats decklist — the same box takes a tournament decklist URL, e.g. https://www.swumetastats.com/decklists/702ffccc-96db-4920-ba07-b492001d8d4c. Paste the whole URL: swumetastats and melee both identify decks by UUID, so a bare id is read as a melee one. (In practice swumetastats reuses melee's ids, so a bare id usually finds the same deck anyway — but the URL is the reliable form.) Cards are matched by printed name, exactly as for melee.
swubase deck — the same box takes a deck URL, e.g. https://swubase.com/decks/92e4a4df-da90-4638-a7bc-52ae584242f4. Paste the whole URL, for the same reason as swumetastats. swubase identifies cards by its own slug rather than by name, so each is matched against the swu-db database in slug space.
Sample deck — Use sample deck restores the bundled deck at any time.
Recent — the last six imported decks you loaded appear as chips under the box; click one to load it again, or its × to forget it. The chip for the deck currently on screen is highlighted.
Leader and base are metadata, not shuffled into the library. Click either one to inspect the art full size; a double-sided leader gets a Flip button.
Deck sheet / Text + Images switches the deck panel between an art grid and a compact list with types, traits, costs and copy counts. Both are clickable in the same way.

Loading a deck clears the current target and result — a target built from one deck's cards wouldn't mean anything against another. If a URL can't be fetched or parsed, the error appears in red next to the Load button and the previous deck stays put.

2 · Main deck and sideboard

Under the deck sits a Sideboard bar showing its size; click it to expand. Everything that follows works in either view mode, and in both directions.

Move a copy — hover a card and press SB ▼ to send one copy to the sideboard, or MD ▲ on a sideboard card to bring one back to the main deck. One copy per click, so a 3-of takes three clicks to move entirely. Clicking a main-deck card still targets it, exactly as before.
Sideboard cards can't be targeted — they're not in the deck being shuffled, so clicking one does nothing. Move a copy to the main deck first, then target it there.
Counts follow along — the per-card ×N badge, the deck total and the unique count in the header, and the sideboard's own totals all update immediately.
Reset deck appears in the Deck panel header as soon as anything moves, and puts every card back where it was imported.
No legality checks — deck size, copy limits and aspect rules are yours to manage. The only thing enforced is conservation: you can rearrange the copies the deck came with, not conjure new ones.

Runs are scored against whatever is in the main deck at the moment you press Run simulation — that's the point of moving cards around. The readout names the deck size it used. The CLI line beneath it always describes the deck as imported, since a swapped main deck has no command-line equivalent; it says so when the two differ.

3 · Build a target

A target is the condition each simulated game is scored against, checked against every card drawn by your chosen target turn. On the main page you assemble it from two kinds of term:

Card terms

Click any card in the deck panel to target it. It appears as a chip under Target with a − 1 + stepper — raise it to require multiple copies. The stepper stops at the number of copies in the deck, since more than that can never be drawn. Click the card again, or the chip's ×, to drop the term.

Resource cost terms

Set Resource cost and Count, then press Add — that's “at least N cards that cost this much.” Good for curve questions (“will I have two cheap plays?”) where the exact card doesn't matter. Cost terms show up as chips like ≥2× cost 2.

Combining terms

With two or more terms, the Combine multiple terms switch decides how they're scored:

Any (OR) default. The game counts as a hit if at least one term is met. This is the “do I draw one of my openers?” question.
All (AND) The game counts only if every term is met — combo pieces, or a card plus a supporting cost requirement.

With a single term the switch has no effect. If you need mixed logic — “this card, plus either of those two” — open Advanced: nested AND / OR, explained in Building nested targets.

4 · Set up the simulation

Setting	What it does
Target turn	How far into the game to draw before scoring. `2R` is the opening hand; each later round adds 2 cards (`3R`, `4R`…). A plain number is a draw count instead — `1` means the opening draw, so `1` and `2R` are the same thing. The arrows step it and keep whichever format you're using.
Initial hand size	Cards in the opening hand, default 6. Total cards seen by round N is `hand + 2 × (N − 2)`, capped at the deck size.
Mulligan	When on, a game whose opening hand already meets the target keeps it; otherwise the hand is shuffled back and redrawn once at the same size. Play then continues from the kept hand. It's a free re-roll of the opening hand — no card penalty is modelled.
Iterations	How many games to simulate, default 10,000, capped at 100,000 here. More games means a tighter estimate: the ± figure shrinks with the square root of the count, so 4× the games halves the error.

Press Run simulation. The readout gives the probability with its standard error, then the hit count, the target it scored, the horizon round, how many cards that drew, and whether the mulligan was on. The last line is the equivalent command for the CLI version of this tool — copy it if you want to rerun the same question locally, at any iteration count.

Worked examples

These use the bundled sample deck (a 50-card Admiral Piett list). The steps are the same for your own.

01 “How often do I open with my key card?”

Click Long Live the Empire in the deck panel — one chip, count 1.
Leave Target turn at 2R and Initial hand size at 6.
Run.

In plain EnglishOut of every 100 games, how many opening hands of 6 contain at least one copy of that card? With 3 copies in a 50-card deck, expect about a third of them — roughly 32%.

02 “Either of my two-drops, by turn 3” or

Click Imperial Dark Trooper and GNK Power Droid — two chips.
Leave the combine switch on Any (OR).
Set Target turn to 3R (6 + 2 = 8 cards seen).

In plain EnglishA game is a hit if either card shows up in the first 8 cards. Adding more “any of these” options is the cheapest way to raise a probability — it's why redundancy beats a single silver bullet.

03 “Both halves of the plan” and

Click the two cards you need together.
Switch Combine multiple terms to All (AND).
Set the turn you'd need them by, e.g. 4R (10 cards).

In plain EnglishOnly games holding both pieces count. Two-card requirements are far rarer than they feel — run it before building a deck around one.

04 “Do I have enough cheap plays?”

Set Resource cost to 2, Count to 2, press Add.
Target turn 2R, hand size 6.

In plain EnglishHow often does the opening hand hold at least two 2-cost cards? No specific card involved — this is a question about the curve, so it answers “can I actually spend turn two?”

05 “Does cutting this card hurt my opener?”

Build a target and run it — note the probability.
Hover the card you're thinking of cutting and press SB ▼ once per copy to move it to the sideboard.
Run again. Same target, one card lighter.
Reset deck puts it back.

In plain EnglishThinning the deck raises the odds of drawing anything else in it, so a cut helps every other target a little. This tells you how much.

06 “Is a specific card worth mulliganing for?”

Build the target and run once with Mulligan off — note the probability.
Flip Mulligan on and run again with everything else unchanged.
The gap is what the free re-roll buys you.

In plain EnglishTwo runs, one variable changed. The same trick works for any setting: compare hand sizes, compare turn 2 vs turn 3, or compare two decks against the same target.

Tip — read the ± figure. 32.61% ± 0.47% means the true odds are near 32.6%, and the simulation's own noise is about half a percent. If two setups differ by less than that margin, raise the iterations before calling one better.

What the browser remembers. Your light/dark choice and that recent-deck list are stored locally in this browser, under a single key — nothing is sent to the server and no account is involved. Clearing site data forgets both.

Tip — the odds are about the deck, not the draw order. Everything here is scored against the set of cards drawn by the target turn. It models what you'll have seen, not the sequence you saw it in, and it doesn't model searching, filtering, or discarding.

Deck data and card art from swu-db.com · Advanced: nested AND / OR targets →