Is Cache CT-sided or T-sided in CS2? 827 pro maps

Bottom line: Cache is mildly CT-sided. Across 827 professional CS2 maps and 17,979 rounds since 25 June 2026, the Counter-Terrorists won 52.7% of rounds (95% margin ±1.0). That is fourth of the seven active-duty maps by the point estimates, though the margins allow third: well below Nuke (56.0%), level with Ancient (53.0%) and above Dust2 (50.6%) and Anubis (49.4%). Teams ban it about two times in three, a share that has eased from 69% in July to 63% in October; only Anubis is banned more. The round data behind every figure is in the CS2 data API.

What counts as CT-sided

A map is CT-sided when the Counter-Terrorist side wins more than half of its rounds. Every round has exactly one winner, so the rate is CT rounds won divided by rounds played, and 50% is an even map.

Cache is the newest map in the pool. Valve announced on 22 June 2026 that it would replace Overpass in the Active Duty pool for Premier Season 5 (report). The Valve update that made the swap shipped on 8 July, and tournament organisers moved to the new pool on different dates between late June and mid-July, so there is no single start date for pro play. It is the first time Cache has been in the pool since 2019, and the map is a remade version released in April, so figures for the old Cache say little about this one. The first Cache map in our data is from 25 June, ahead of Valve's update, so the window here runs from 25 June to 17 October 2026.

How the figures were built:

• Completed matches, played maps only (walkovers and technical results are not maps played), with round-by-round data.
• Round data exists for 827 of the 1,031 played Cache maps (80%). Tiers are the event tiers in our data, S for the top events down to C: all 53 S-tier maps are covered, 69 of 70 in A, 363 of 414 in B, 342 of 466 in C, and none of 28 from unranked events. 826 of the 827 have every round on record; one is short.
• Rounds inside one map are not independent, since a team on a roll wins several in a row, so every margin below treats a whole map as one observation.

The answer: slightly CT-sided

CTs won 9,479 of 17,979 rounds, 52.7%, with a 95% range of 51.7% to 53.7%. An even Cache is well outside that range. On a typical 22-round map the lean is worth about 0.6 of a round to the CT side.

The pistol rounds, the two per map where both teams have the same minimal buy, are close to level: CTs won 857 of 1,650 (51.9%, margin ±2.4). That fits an even split as well as it fits the map's overall lean.

How it compares with the other six maps

Cache sits in the middle. It cannot be told apart from Ancient (0.3 points apart), it is clearly less CT-sided than Nuke (3.3 points) and clearly more than Dust2 (2.2) and Anubis (3.4). Mirage (1.4 points above) and Inferno (1.6 below) are close calls, just outside the margin of error on each gap (about 1.3 points), so read the middle of the table as a band rather than a strict order.

No map in the pool is clearly T-sided. Anubis is lowest at 49.4%, and its range of 48.3% to 50.4% includes an even split. Taken together the seven maps are 52.5% CT (155,767 rounds on 7,247 maps).

That is a change for Anubis, with two things behind it. Across the whole CS2 era in our data it was T-sided, at 45.7% CT over 4,016 maps with round data, but it was out of the active-duty pool from July 2025 to January 2026 (we have no Anubis round data for August to December 2025), and when Valve brought it back in a January 2026 update it changed the map (it moved the bridge drop and reversed the Mid doors, among other edits), which trade-press coverage tied to its T-side lean. Since the return it has run at 47.3% (January to June 2026) and 49.4% (July on, 685 maps). The rise is partly pool-wide: the five long-standing maps (Ancient, Dust2, Inferno, Mirage, Nuke) went from 49.9% CT in the first half of 2025 to 52.8% in the first half of 2026 and 52.9% since, so Anubis's gap to them narrowed from 5.4 points (January to June 2026) to 3.5 (July on) rather than closing. It is still the lowest map. Older all-time tables, including our map win rates page, show it as the T-sided outlier, which was true of the old map and is no longer clearly true.

Is Cache's lean fading?

Before 1 September Cache ran at 53.7% CT (379 maps, margin ±1.5). Since then it has run at 51.9% (448 maps, ±1.4). The drop is 1.8 points with a margin of about ±2.0 on the difference, so it falls just short of the usual bar (p = 0.07). The other six maps together did not move: 52.5%, then 52.6%. The result also depends on where the line is drawn: splitting at 1 August, 15 August, 15 September or 1 October gives p of 0.17, 0.03, 0.22 and 0.19.

A hint, not a finding. The table leaves out the five maps from 25 to 30 June. Event tier does not explain the drop (holding the tier mix fixed leaves it at 1.8 points), so if it is real it comes from something else, such as teams learning how to play the T side of a remade map or a map update; we have tested neither. Single weeks swing by two to three points on about a thousand rounds, so do not read any one of them. A few more weeks of maps will tell us more.

Does the tier matter?

Not visibly. On Cache, S and A tier maps run at 51.4% CT (122 maps, ±2.5) and B and C tier at 52.9% (705 maps, ±1.1). The 1.5-point gap sits inside the margin of the smaller group, and across all seven maps the two groups are level at 52.6% and 52.5%. With 53 S-tier Cache maps, the sample cannot be cut finer than that.

How often teams ban Cache

From 7 July, the day after Premier Season 4 ended, 2,554 completed matches have a veto that lists Cache. It was banned 65.9% of the time (1,682), picked 24.3% (621) and left as the decider 9.8% (251). Among the seven maps only Anubis is banned more (71.6%), and Dust2 is banned least (46.6%). Each share has a margin of about 1.8 points. The Cache share has eased month by month, from 69% in July to 67% in August, 64% in September and 63% in October, which puts it level with Inferno (63% to 65% throughout).

In all but two of the series here the veto lists all seven maps, so these shares are comparable across maps, which raw totals spanning years are not (the veto study explains why those mislead). They say nothing about why teams choose as they do. Cache's side balance is mid-table, so side balance alone does not explain the bans. For how a veto works, see the map veto entry.

Compute it yourself

Round-by-round team stats are in every match record. With the Python client (pip install esportsodds), the bulk route returns 25 matches a page with their maps and rounds:

from esportsodds import Client

client = Client()  # reads ESPORTSODDS_API_KEY from the environment

ct_rounds = ct_won = 0
for match in client.matches.iter_bulk(
    include="maps,rounds", has="rounds", status="completed", date_from="2026-06-25", limit=25
):
    cache_maps = {
        m["map_number"] for m in match["maps"] if m["map_name"] == "de_cache" and m["status"] == "played"
    }
    for r in match["rounds"]:
        if r["map_number"] in cache_maps and r["team_side"] == "ct":  # one ct row per round
            ct_rounds += 1
            ct_won += r["won"]

print(f"CT won {ct_won} of {ct_rounds} Cache rounds ({ct_won / ct_rounds:.1%})")

Every round has one row per team, so counting only the ct rows counts each round once. About 3,200 matches since 25 June have round data, so the loop makes roughly 130 requests of about 1.5 MB each; new matches keep arriving, so a rerun will land a little past the figures here. Each page counts as one bulk-export request, and the 14-day money-back guarantee on a first subscription holds only while an account has made fewer than 1,000 requests and no more than 20 bulk-export requests since its first payment, so the full window ends it. In those first two weeks, move date_from closer to today so the run stays under 500 matches (20 pages), and swap de_cache for any other map name. For the margins above, keep one pair of numbers per map (CT rounds won, rounds played) and use a standard error that treats each map as one unit.

Questions

Is Cache CT-sided or T-sided in CS2?

Mildly CT-sided. Across 827 professional maps since 25 June 2026, the Counter-Terrorists won 52.7% of 17,979 rounds (95% margin ±1.0), fourth of the seven active-duty maps by the point estimates.

Which CS2 map is the most CT-sided?

Nuke, at 56.0% CT over 958 maps, followed by Mirage at 54.1%.

Which CS2 map is the most T-sided?

None clearly is. Anubis is lowest at 49.4% CT, and its range of 48.3% to 50.4% includes an even split. Across the whole CS2 era in our data it was T-sided, at 45.7% CT, but Valve changed the map when it returned in January 2026 and it now runs close to even.

Where do these numbers come from?

From the round-by-round and veto data in the EsportsOdds CS2 data API, as of 17 October 2026. They describe past maps and say nothing about the next one.