The 2026 map pool and how to cut it to four

The pool changed twice in a year, which means a lot of players are still making decisions from stats about maps that no longer exist.

Current Active Duty

Ancient · Anubis · Cache · Dust 2 · Inferno · Mirage · Nuke

Recent changes: Train was removed in January 2026, with Anubis returning in its place. Cache replaced Overpass on 6 July 2026. FACEIT mirrors Valve’s pool, so this is what you queue.

The always-current version, with the removal dates, lives on the map pool page.

Your dead data

If you have hundreds of matches on Overpass or Train, those statistics are real and they are also useless. A 60% win rate on a map nobody plays is not an asset, and averaging it into your overall numbers makes you look better than your queueable self.

Exclude retired maps before you make any decision about which maps to play. This sounds obvious and almost nobody does it, because stats pages happily show all of them in one list.

What a real map pool problem looks like

Abstract advice about map pools lands badly, so here is a measured one. This is a Level 8 account from our worked examples, 1,602 matches, 30,560 rounds — a sample large enough that none of it is variance:

Map	Matches	Win rate
Ancient	209	52%
Nuke	65	51%
Dust2	373	50%
Anubis	233	47%
Mirage	507	46%
Inferno	129	45%

Find the largest number in the matches column. Mirage: 507 matches at 46%. Nearly a third of their entire career is played on their second-worst map, and 507 matches is far past the point where that could be a bad run.

Their overall win rate is 48%, which at ±25 ELO per match means they are not climbing at all — a sub-50% win rate has no match count that reaches the next level. Their gap to Level 9 is 90 ELO, which at 48% is infinity and at 55% is 36 matches.

They do not need to improve to reach 55%. They need to stop playing Mirage. Reweighting the queue toward the maps they already win on moves the aggregate without changing anything about how they play.

“Just get better at the map” has already been tested

The instinctive objection is that banning a weak map is avoidance, and the real answer is to learn it.

Over 507 matches, that experiment has run. They have played Mirage more than any other map in their history and it remains their second-worst. Whatever the mismatch is — the pacing, the angles, the utility discipline that map demands — five hundred attempts did not resolve it.

There is a version of “learn the map” that works, and it involves deliberate practice off the back of demo review rather than another hundred ranked games. But it is a project measured in months, and it is a strange project to take on while a free three-to-six points of win rate is sitting in the pick/ban screen.

How to pick the four

The rule is four maps, and the selection method matters more than the number.

Sort by win rate, not by ADR and not by enjoyment. ADR measures how much damage you do; win rate measures whether the damage mattered. A map where you top the scoreboard and lose is a map to ban — that combination usually means you are getting value in rounds that were already decided.

Discard anything under 15 matches. Not “weight it less” — discard it. A map at 57% over 8 matches tells you nothing, and treating it as your best map is how people end up maining something they have never actually tested.

Drop retired maps entirely before you sort. Overpass and Train stats are real history about something you cannot queue.

Take the top four that survive all three filters, then check that you actually have four. Players with a lot of dead-map history often find they have only three qualifying maps, which is useful to know: it means the next map you learn should be chosen deliberately rather than picked up by accident.

Then hold it for thirty matches. A pool you change every week is not a pool, and the win rates never accumulate enough matches to mean anything.

When to ban a map you like

The uncomfortable case is the map you enjoy that sits at 45%.

Ban it anyway, with one exception: if it is below 15 matches, it does not qualify for a verdict in either direction, and playing it more is how you find out. Above 15, enjoyment is not evidence.

What you can do is keep it as a deliberate off-pool map on a night you are not tracking ELO — warmed-up, early, not after two losses. That is a different activity from climbing, and separating the two is more honest than pretending the map is fine.

Cache is the interesting one

A map that has just entered the pool is the one place where preparation still buys an edge. Nobody at your level has the lineups yet. Six smokes and four molotovs learned this week will be worth more than they will ever be worth again, because in three months everyone will have them.

The caveat: a map you have played nine times has no win rate. It has a mood. Fifteen matches is roughly the point where the number starts meaning something — below that you are reading a good week, not a good map.

Build the four-map pool

Step one: throw out the dead maps. Overpass, Train, Vertigo — gone, whatever they say.

Step two: throw out anything under 15 matches. Not enough signal. Play them or ignore them, but do not plan around them.

Step three: rank what is left by win rate. Not ADR. This matters more than it sounds.

Step four: keep four. Permanently ban the worst two.

Win rate, not ADR

The single most common mistake in map selection is picking by ADR, and it is worth being blunt about why it is wrong.

ADR measures how much damage you do. Win rate measures whether the damage mattered. A map where you consistently top the scoreboard and lose is a map where you are farming damage in rounds that are already lost — usually because the map’s structure forces you into a role that does not fit you.

Your best map by ADR and your best map by win rate are frequently different maps. Trust the second one.

Why four and not seven

Below Level 10, depth beats coverage.

Seven maps means seven sets of default setups, timings, common angles and utility lineups. Nobody at 1700 ELO knows all of that on all seven. What actually happens is that four of them are solid and three are coin flips you volunteered for.

Cutting to four does three things. Your reps per map roughly double, so the learning compounds. You stop losing the games you were always going to lose. And your worst-case matchup improves, because a banned map cannot be picked against you.

The cost is real but small: longer queue times, and occasionally a map you enjoy that you do not play. Enjoying a map you lose on is an expensive hobby.

What to do about a map you must play

Sometimes a bad map survives the veto. Two things help more than lineups:

• Pick the simplest possible role. If the map forces you to entry and you are not an entry player, the fix is not to learn the map, it is to take a passive position and let the round come to you.
• Buy for the round after. Bad-map rounds are lost more often, so the economy discipline matters more, not less.

Check your own pool

Dead maps filtered, small samples flagged, ranked by win rate rather than ADR, with a pick/ban recommendation: run your profile.