SWU.Report FAQ & Methodology
The short version
SWU.Report turns Star Wars: Unlimited tournament results into a handful of core numbers:
- APR (Adjusted Performance Rating): Is a deck overperforming or underperforming based on its turnout?
- The Bounty Rating: A 0-100 score showing “How close is this deck to being the best in the current format?”
- Tiers (S/A/B/C/D): A quick categorization built off Combined APR.
- HHI (meta concentration): How concentrated or healthy the field currently is.
Where the data comes from, and where it still has gaps, is below. There are gaps. Anyone telling you their meta model has none is selling something.
APR: Adjusted Performance Rating
What problem is this solving?
Raw meta share tells you how popular a deck is. Raw win rate tells you how it performs in isolation. Neither tells you whether a deck’s results are earned. A deck can be 55% of the field solely because it’s easy to build and forgiving to pilot, but still underperform half the field below it. A deck can also sit at 4% of the field and be the best performer in the format but folks might not notice if it’s not played enough.
APR tries to relate those two concepts to each other. So, is deck performing (making cuts/winning) at the rate you’d expect for how many people are playing it?
How it’s calculated
APR is broken into two stages, matching the two cuts that happen at a tournament:
- Stage 1 (S1): Field → Top 8. Does field presence translate into cut appearances at a higher or lower rate than its size predicts?
- Stage 2 (S2): Top 8 → Win. Once in the Top 8, does it close out events at a higher or lower rate than that Top 8 presence predicts?
Each stage runs the same formula:
PR = (Next_stage% − Current_stage%) / Current_stage%
APR = PR × (log₁₀(count + 1) / log₁₀(total + 1))
The first line illustrates PR (Performance Rate), which is just an over/underperformance ratio. The second line created APR by introducing logarithmic dampening. The log term is doing real work as a volume dampener, so that a deck with 3 Top 8s running hot to 100% doesn’t move the needle nearly as hard as a deck doing the same across 30. It keeps a small, noisy sample from muscling its way to the top of a leaderboard it hasn’t earned.
Combined APR is just S1 + S2. The number you’ll see most, and what the S/A/B/C/D tiers are built from.
(A small floor of 0.5 gets applied to Top 8 and win counts so a deck with zero results doesn’t produce a divide-by-zero. Gotta have some failsafes.)
What does a positive or negative number mean?
- Positive: Wins more than its field share predicts. Earns its spot.
- Negative: Results lag behind popularity. More common than its record justifies.
- Near zero: Performs almost exactly as its size predicts.
Why not just use win rate?
It flattens too much. A 2-1 record and a 40-20 record read identically under straight win rate, and it says nothing about whether popularity is inflating or deflating how good a deck looks. “Wins a lot” and “wins more than its size predicts” are different questions.
Why not use Conversion Rate?
Conversion rate (Top 8s ÷ field entries, or wins ÷ Top 8s) has the same core problem as win rate: no sense of scale.
- No baseline. A 20% conversion rate means nothing alone. Good, bad, or average depends on what the rest of the field is converting at.
- Small samples look identical to real trends. 1-for-2 and 40-for-80 both read as 50%. Conversion rate can’t tell you which one means something.
- Isolated from the rest of the metagame. No reference point for whether the number is earned or lucky.
- It cannibalizes. Popular decks are more likely to face each other over the course of a tournament and if they do so at the end of that stage, only one of the two can convert.
In some wasy, it’s the same complaint as win rate, just one level more granular. Neither accounts for volume, and APR exists to bridge that gap.
Tiers (S / A / B / C / D)
| Tier | Combined APR |
|---|---|
| S | ≥ 80% |
| A | ≥ 40% |
| B | ≥ 10% |
| C | ≥ -10% |
| D | below -10% |
They exist for scanning speed, quick refernce, and discussion. Individual stats always the more precise read, and two decks in the same tier can still be meaningfully different. Don’t lean on the letter grade to settle an argument the underlying number could settle better.
The Bounty Rating
What is it?
A single 0-100 score that answers a different question than APR: not “Is this deck earning its results,” but rather “How close is this deck to being the single best deck in the format, right now.” This is adapted from a method Vicious Syndicate runs for Hearthstone, with two additions of our own. Also, it’s worth really noting that “right now” clause. Bounty Rating is built so that the best deck in the current meta is the anchor point for rating everything else. You can’t really compare two different meta periods this way.
Where this comes from, and where it doesn’t
The core mechanic is Vicious Syndicate’s, straight out of their Data Reaper Report. Check out their FAQ for more information (coincidentally it inspired this FAQ too). For Power Score, whichever deck holds the highest win rate gets set to a fixed 100, the floor is 100% minus that same number, a mirror around 50%. For Frequency Score, field share is used on a plain 0-to-highest-popularity scale. Combine the two with a simple average and that’s their vS Meta Score. We use both of those exactly how they define them, but we also did a bit more as well.
What we add are two more lenses built the same way, mirror-and-scale against a ceiling, applied to something Vicious Syndicate has no equivalent of. Hearthstone’s ladder has no bracket, so there is not a Top 8 for them to build a stage-conversion measure against. But SWU tournaments do and we’ve already talked about APR as a useful stat, so S1-APR (the Field→Top 8 stage) and S2-APR (the Top 8→Win stage) get the same treatment as Power and Frequency, each scored on its own mirrored 0-100 scale.
These four lenses are then averaged into the final number:
| Lens | What it measures | Source |
|---|---|---|
| Power Score | Game win rate, mirrored around 50% | Vicious Syndicate’s own, unchanged |
| Frequency Score | Field share, 0-to-highest scale | Vicious Syndicate’s own, unchanged |
| S1-APR | Field→Top 8 stage APR, standalone | Ours |
| S2-APR | Top 8→Win stage APR, standalone | Ours |
The sample size floor
Only archetypes with at least 50 field entries and 3 Top 8 appearances count. Below that, one lucky run can dress itself up as a 100% win rate and drag the whole scale sideways, since every other score gets measured against whoever’s on top.
A word on the name
Ok, I couldn’t help being a little bit cutesy when trying to think of a fun and thematic name for this. Most of the top (and problematic) decks in the history of the game have been Bounty Hunter related (Boba Fett, Jango Fett, Cad Bane). It only seemed right to lean into that.
HHI (meta concentration)
Borrowed from antitrust economics, repurposed to describe how top-heavy or spread out the metagame currently is. Tracked two ways: Field HHI and T8 HHI (which runs hotter, since strong decks cluster once you reach the cut).
| Effective # of archetypes | Label |
|---|---|
| HHI < 5 | Open |
| HHI 5-10 | Healthy |
| HHI 10-20 | Moderate |
| HHI ≥ 20 | Concentrated |
These bands are calibrated to an “effective number of archetypes” framing, not the raw antitrust thresholds they come from. We tried the textbook version first. It basically never left “wide open,” even during stretches of this game’s history that were anything but. We recently tweaked and lowered the thresholds once again and that’s an option that I want to keep open for the future as well, since we’re feeling out how to best apply this to a card game meta.
Where the data comes from
- SWUMetaStats.net — aggregates paper tournament results from Melee.gg. Primary source for field counts, Top 8s, and event wins.
How often does it update?
Weekly during an active PQ season. Matchup data (Avg Matchup Win Rate, Field-Weighted Matchup Expectation) follows a point-in-time snapshot rule: each week’s numbers come only from that week’s own pull and are not retroactively recalculated.
Known limitations
- Small samples are volatile, even with the log dampener. A handful of recorded games can still swing hard week to week. Treat a low-count APR figure like any small sample: interesting, not decisive.
- Generic bases get bucketed by color, not tracked card by card. (“Vigilance 30 HP” bases become “Blue 30.”) Rare and unique bases with their own text stay tracked individually. This is something we’ll have to change come Homeworlds. (Not looking forward to that, honestly.)
Does something look off, or do you have a question this didn’t answer? Reach out on Discord.