What these reports contain
A metagame report is a summary of what players are using and how it performs. Most contain two basic columns for each deck, character or strategy: how often it was played (play rate, sometimes called presence) and how often it won (win rate). Some add matchup grids, which show how each option performs against each other.
The two columns answer different questions. Play rate tells you what you are likely to meet. Win rate hints at how strong something is. Mixing them up is the first common error.
Play rate is about opponents, win rate is about strength
A popular deck is not necessarily a strong one; it may be easy to learn, heavily promoted, or simply fashionable. A rare deck is not necessarily weak. If you are choosing what to prepare against, play rate matters most, since it tells you what to expect. If you are choosing what to play, win rate matters, but only when interpreted with care.
The two interact. A strong deck draws players, which raises its play rate, which in turn leads other players to adapt against it, which can lower its win rate. So high play rate and falling win rate often occur together, and this can be the sign of a healthy cycle, not of a problem with the deck. The shape of that cycle is the subject of the archetype triangle.
The sample size problem
The most common source of misreading is a small number of games. Win rate is a proportion, and proportions are noisy when the denominator is small.
A rough illustration: if a deck’s true win rate were exactly 50 percent, then over 20 games you would not be surprised to see 12 wins (60 percent), or 8 wins (40 percent). Over 2,000 games, a result of 60 percent would be extremely unusual. The standard rule of thumb is that the margin of uncertainty shrinks with the square root of the sample size, so quadrupling the games roughly halves the uncertainty. That means a table row showing 58 percent from 40 games says very little, while 52 percent from 4,000 games is more informative, though it may not be a large effect.
A good report states how many games underlie each row. If it does not, treat the numbers cautiously. The general principle of sample size when judging a change applies to every row of a table.
Other ways the table can mislead
Who is playing. Data from top-level events reflects strong players; data from a ladder reflects everyone. The two can differ, because a deck that is hard to play well might look weak on average but strong in expert hands. A report should say which population it describes.
Selection effects. Players who choose a deck are not a random sample. If a deck is mostly picked by experienced players, its win rate partly reflects them. Comparing decks played by different groups is not a fair test.
Mirror matches. When a deck plays itself, one side wins and one loses, so mirrors pull a deck’s win rate toward 50 percent. Whether or not the report excludes them changes how to interpret a figure.
Survivorship. In multi-round events, players who lose early stop playing. Late-round data therefore contains mostly winners, and decks that did well early appear stronger in later rounds. Be sceptical of any table that is restricted to a stage that only some players reach.
Time. Data from before a patch can be stale. Changes made in balance patches reset many assumptions, and a report that straddles one is harder to read.
Reading matchup grids
A matchup grid shows each deck’s win rate against each other deck. These are more informative than overall win rates, because they show why a deck wins. A deck with a mediocre overall rate might be excellent against one popular deck and weak against everything else.
But grids multiply the sample problem: splitting 4,000 games across many pairings leaves each cell small. Look for cells with large counts, and be wary of extremes in cells with few games. Direct comparisons between particular opponents are a version of head-to-head records and share their pitfalls.
Estimating your expected result
If you know the play rates of the decks in the field and your deck’s win rate against each, you can compute a weighted average: multiply each matchup win rate by how often you expect to face that deck, and add them up. A simple, made-up example: if you face deck A 60 percent of the time and win 55 percent of those games, and face deck B 40 percent of the time and win 45 percent, the result is 0.6 × 0.55 + 0.4 × 0.45 = 0.33 + 0.18 = 0.51, or 51 percent. This is an expected value calculation, and it is only as good as the inputs. The figures in the example are illustrative, not drawn from any real report.
Not confusing balance with fun
A flat, roughly even win rate across options is often cited as evidence of good balance. It is a reasonable starting point for win rate balance, but not a complete one. Options may have equal win rates and still feel very different to play against, and some matchups may be unpleasant even when the numbers are balanced.
Practical habits
- Check the sample size behind every figure you rely on.
- Check who the data describes and over what period.
- Prefer trends across several reports to a single snapshot.
- Treat small gaps, such as 1 or 2 percentage points, as likely noise unless the sample is very large.
- Remember that your own skill and familiarity with a deck can matter more than a few percentage points of average win rate.
A metagame report is a summary, and a summary hides detail. It is a good prompt for questions, but a poor substitute for playing and reviewing your own games.