One week of Ranked: 140,000 matches and activity up 40%
Ranked is one week old. We measured what it actually changed: match volume, how the game is played, and the heroes whose value collapses as soon as points are on the line.
Ranked launched late in the day on July 30. A week later, there are enough matches on record to move past impressions and look at what actually changed. We queried the public data from deadlock-api.com across the whole period, separating ranked matches from the rest of the game every time.
Four matches in ten are now ranked
In seven days, 140,721 ranked matches were played. The mode took hold immediately: on the very first Friday it accounted for 46% of all activity, before settling around 41%, a figure that has not moved since. For a mode Valve gates behind proving yourself elsewhere first, that is massive adoption.
The game runs 40% hotter, but the effect is eroding
Comparing each day against the same weekday a week earlier, across all modes, the curve tells a clear story — in two acts.
On Saturday August 1, the game logged 57,052 matches against 40,566 the previous Saturday, or +40.6%. Sunday, +38.2%. Then the slope turns: Monday +29.3%, Tuesday +19.0%, Wednesday +12.0%.
The novelty effect is real and large — Ranked lifted activity by nearly half — but it deflates steadily, by roughly seven points a day. At the rate seen over this first week, the game would return to its previous level within ten days or so. Where the curve actually settles is the real question, and it needs a few more weeks of hindsight.
Ranked is not played like the rest of the game
The contrast is sharp, and it shows up on almost every metric.
The abandon rate drops to 0.33%, against 0.74% in unranked: players quit a ranked match half as often. Matches are also shorter — 37 minutes 24 against 38 minutes 52 — and noticeably less violent, with 6.4 average kills per player against 7.0.
Fewer fights does not mean less care, though. Weapon accuracy rises to 52.4% against 49.9%, denies go up 5%, and the mid boss ends up dying in 99% of ranked matches, against 94% elsewhere. Ranked produces a more careful, cleaner, more objective-driven game.
The ranked leaderboard is not the normal one
A hero who is strong in unranked is not automatically strong once points are on the line. The top three hold up — Victor at 55.3%, Seven at 55.2%, McGinnis at 54.6% — but the gaps between the two modes are sometimes wide.
Vyper is the big winner of Ranked: 51.2% win rate against 47.9% in unranked, 3.3 points gained purely because players are trying. Sinclair follows at +2.4 points.
The other way round, Haze loses 2.8 points (48.6% in Ranked against 51.4% elsewhere) and Infernus 2.5. Graves, the best hero in the game in unranked at 56.2%, drops to 54.3% as soon as points are involved — and gets picked noticeably less.
At the bottom, Mirage closes the table at 44.0%, ahead of Venator, Silver, Mina and Bebop, all under 46%.
The most banned heroes are not the strongest
This is the most counter-intuitive result of the week. Across all recorded bans, Haze alone accounts for 9.4% — the most banned hero in the mode. The Doorman follows closely at 9.1%.
Yet both lose more ranked matches than they win: 48.6% for Haze, 46.5% for The Doorman. Players are spending their bans on a reputation, not on results. Haze is even picked more often in Ranked than elsewhere, despite performing worse there.
Only McGinnis, the third most banned hero, fully earns the treatment: 54.6% win rate, 1.4 points better than in unranked. Meanwhile Victor, who has the best win rate in the mode, is banned half as often as Haze.
Who is actually playing Ranked
The rank breakdown holds a surprise. From Initiate to Emissary, the brackets are remarkably even, each accounting for 8 to 13% of matches. Then comes a huge block at Oracle, which alone makes up 24% of ranked matches for only 17% of the players.
In other words, the top of the ladder plays far more than everyone else. Ranked has already become their default mode, while it remains one option among several for the middle of the table.
What to take away
Ranked is a usage success — 41% of matches, a stable share — and it lifted overall activity by 40% at its peak. The open question is durability: the effect erodes by around seven points a day, and it will take a few weeks to see where it lands. What it does deliver is exactly the behaviour you want from a competitive mode: fewer abandons, better accuracy, objectives taken every time.
One problem is left on Valve's desk. When the two most banned heroes in a mode win less than half their games, the issue is not balance — it is how players read the game. A week is still short for that reading to correct itself.
Methodology. Figures collected through the public deadlock-api.com API, covering the period from the mode's release on July 30 at 19:14 UTC to August 5. Volume comparisons use complete days only. Ban statistics come from replay analysis and cover a sample of matches, not all of them.
Correction, August 9, 2026. The first version of this article, published on August 5, concluded that the activity spike had fallen back below its pre-Ranked level. That was wrong. The API's data backfills over several days: measured on August 5, Tuesday August 4 showed 36,478 matches; measured on August 9, that same Tuesday shows 47,548, or 30% more. The most recent days were therefore systematically undercounted. The volume figures above have been recomputed on consolidated data, and the conclusion reversed: activity did rise, it had not fallen back. The article's other measurements — abandon rate, accuracy, duration, win rates, bans, rank distribution — are averages, unaffected by this backfill: they were rechecked and have not moved.