ReportUpdated

Dust2 B is the hardest bombsite to retake in CS2.

Dust2 B is the hardest of all 14.

1 in 5Dust2 B retakes won116 of 580 in gun rounds
1 in 4expected with the same players aliveabout 145 of 580
1sthardest of 14 bombsitesranked by the gap to expected
Retake routes: 1 B Doors and 2 B Window (together 88% of 231 entries), 3 Upper Tunnels (12%). Red: the default plant.

Dust2 B is the only one of 14 bombsites where pro CTs win fewer retakes than the players alive at the plant predict, by more than luck.

Its CTs start further from the bomb than at any other site. Below are the ranking and the distances, then retake or save, kits and teams.

Get 7 days of Pro free and first access to x-ray.gg when it launches.

01The ranking
i
Expected
What CTs won across all 14 sites with the same players alive
Gap
Share won minus expected, in percentage points
Range
95% bootstrap over demos (2,000 draws)
Clear
Short of expected after testing all 14 sites (Holm p under 0.05)
Dust2 B
Ranked 1st. Could be 1st to 3rd. Hardest in 63% of draws.
Filters
Players alive is summed in your browser. Gaps only, without ranges or Clear.

Only Dust2 B falls short of expected by more than luck

CTs don't arrive at every site with the same players alive, so raw win rates can't be compared site to site. Each site is measured against expected: what retakes across all 14 sites win with the same players alive at the plant.

On raw win rate, Dust2 B (20%) and Mirage B (20.1%) are level. Mirage B's CTs are a player down a little more often, so a little less is expected of them. Dust2 B's CTs are a player down at 58% of plants, against 56% across all 14 sites. They win about 29 retakes fewer than expected.

By the numbers
Exactly

Gun rounds, all pro demos, May 2025 to Oct 2026, 6,980 retakes on 14 sites

The 14 bombsites: retakes won against expected for the same players alive, biggest shortfall first
#SiteWonRateExpectedGap to expected, pointsClear of luck
1Dust2 B116 of 58020%25%−5.0 points (95% range −7.3 to −2.5)Clear
2Inferno B134 of 53625%29%−3.6 points (95% range −6.4 to −0.5)
3Mirage B87 of 43320%24%−3.5 points (95% range −6.5 to −0.7)
4Dust2 A160 of 62126%27%−0.9 points (95% range −3.6 to +1.9)
5Inferno A121 of 49325%25%−0.4 points (95% range −2.9 to +2.1)
6Nuke B146 of 44433%33%−0.2 points (95% range −3.1 to +2.8)
7Anubis B101 of 45222%23%−0.2 points (95% range −3.3 to +2.8)
8Cache A84 of 32426%25%+0.7 points (95% range −2.7 to +4.1)
9Cache B65 of 26525%23%+1.1 points (95% range −2.4 to +4.6)
10Ancient B187 of 65329%27%+1.5 points (95% range −0.9 to +4.0)
11Anubis A98 of 36827%25%+1.6 points (95% range −1.7 to +4.8)
12Nuke A112 of 44025%24%+1.9 points (95% range −1.0 to +4.7)
13Mirage A186 of 71826%23%+2.4 points (95% range +0.2 to +4.8)
14Ancient A209 of 65332%28%+4.0 points (95% range +1.5 to +6.7)

Dust2 B is 1st and clear in all 4 other views

  • Top 30 v top 30815 demos77 of 414won (19%)

    Expected 25%1st, clear

  • Last 180 dayssince 14 April 2026105 of 515won (20%)

    Expected 25%1st, clear

  • Last 90 dayssince 13 July 202659 of 296won (20%)

    Expected 25%1st, clear

  • All rounds, pistols and ecos too13,445 retakes200 of 1,010won (20%)

    Expected 25%1st, clear; Inferno B also falls short (−4.1)

02Distance

Dust2 B's CTs start further from the bomb than at any other site

At the plant, Dust2 B's living CTs are on average 1,783 units from the bomb. That's about CT Spawn to Long Doors, or 8 seconds running. Across all 14 sites the average is 1,460, roughly CT Spawn to B Doors. The difference is about 1.5 seconds of running. Won retakes get their first kill about 3 seconds sooner after the plant than lost ones.

The three sites whose CTs start furthest from the bomb (Dust2 B, Inferno B and Mirage B) are the three that fall furthest short of expected. Nuke's two sites are the closest, at 929 and 955 units, and win about what's expected.

Dust2 B's CTs attempt the fewest retakes of any site: 65%, against 76% expected. They also win less than expected of the ones they attempt (31% against 35%).

379 of 580attempted (65%)

Expected 76%Lowest of 14

116 of 379won when attempted (31%)

Expected 35%

Average distance from the living CTs to the bomb at the plant, all 14 sites, game units: Dust2 B furthest at 1,783, the average 1,460
SiteUnits
Dust2 B1,783
Inferno B1,752
Mirage B1,707
Cache A1,674
Cache B1,601
Dust2 A1,558
Anubis A1,558
Mirage A1,486
Inferno A1,473
Anubis B1,356
Ancient B1,340
Ancient A1,306
Nuke B955
Nuke A929
Average distance from the living CTs to the bomb at the plant, in game units.
03Players alive

CTs win 37% of attempted retakes at even numbers, 13% one down and 3% two down

Scores are CTs first: 2v3 is two CTs against three Ts alive at the plant. At even numbers CTs attempt nearly every retake. They win 41% at 2v2 (239 of 589) and 31% at 3v3.

One down, CTs win 14% at 2v3 (67 of 485) and still attempt 71% of them. Two down, they attempt 54% of 2v4 retakes and win 9 of 324.

CTs attempted 355 retakes three or more players down and won none of them. A full 5v5 retake is won 21 of 61 times (34%).

Retake or save: CTs' win chance when they attempt the retake, by players alive at the plant. Rows: CTs alive. Columns: Ts alive. Each cell: won when attempted, how often attempted, won of attempts.
CTs v Ts1 T2 Ts3 Ts4 Ts5 Ts
5 CTs5v1, Fewer than 20 attempts: few6 of 6, attempted 100%5v2, Retake: 96%23 of 24, attempted 96%5v3, Retake: 76%44 of 58, attempted 100%5v4, Retake: 49%34 of 69, attempted 99%5v5, Retake: 34%21 of 61, attempted 95%
4 CTs4v1, Retake: 100%30 of 30, attempted 100%4v2, Retake: 83%106 of 128, attempted 100%4v3, Retake: 58%145 of 251, attempted 100%4v4, Retake: 33%70 of 210, attempted 91%4v5, Long shot: 10%13 of 129, attempted 66%
3 CTs3v1, Retake: 94%125 of 133, attempted 100%3v2, Retake: 70%286 of 406, attempted 100%3v3, Retake: 31%162 of 530, attempted 92%3v4, Long shot: 11%38 of 342, attempted 62%3v5, Save: 3%5 of 183, attempted 54%
2 CTs2v1, Retake: 80%225 of 282, attempted 99%2v2, Retake: 41%239 of 589, attempted 97%2v3, Long shot: 14%67 of 485, attempted 71%2v4, Save: 3%9 of 324, attempted 54%2v5, Save: 0%0 of 153, attempted 63%
1 CT1v1, Retake: 52%92 of 177, attempted 97%1v2, Long shot: 16%53 of 331, attempted 81%1v3, Save: 3%6 of 226, attempted 48%1v4, Save: 0%0 of 154, attempted 46%1v5, Save: 0%0 of 48, attempted 47%

Reading a square: 2v2Pick any square to read it here.

41%239 of 589, attempted 97%
  1. 41%Won when they attempt it
  2. 239 of 589Retakes won, of those attempted
  3. 97%How often they attempt it (the bar)
Retake: 25% or more wonLong shot: 10 to 25%Save: under 10%Even numbers

Gun rounds, all 14 sites, 6,980 retakes

One down, attempted retakes win 16% within 1,000 units and 6% from 1,500

The distance is the closest CT's at the plant. Within 1,000 units of the bomb, CTs one down attempt 792 retakes and win 130 (16%). From 1,500 units out they attempt 163 and win 10 (6%).

Closest CT to the bomb, in unitsHow often they attempt itWon when they attempt it
Under 1,000up to CT Spawn to A Site, 4.5 s
86%
16% 130 of 792
1,000 to 1,500up to CT Spawn to B Doors, 7 s
65%
9% 31 of 332
1,500 to 2,000up to Long Doors to the A default plant, 9 s
47%
9% 9 of 99
2,000 to 2,500up to the B plant to A Site, 11.5 s
43%
2% 1 of 48
2,500+over 11.5 s running
19%
0% 0 of 16
04When to save

At least one CT is alive at the end of 98% of saves

In 1,601 of 1,637 saves (98%), at least one CT was still alive when the round ended.

Of the four things below, players alive split retakes the most. One down, the next is how far the closest CT is from the bomb. Maps differ in how often pros save, and attempted retakes win about as often on every map.

  1. Even numbers: retake.

    37%Won when attempted584 of 1,567
    95%Attempted1,567 of 1,657
  2. One down: a long shot up close, a save from further out.

    13%Won when attempted171 of 1,287
    70%Attempted1,287 of 1,838

    Won when attempted, by how far the closest CT has to run

    • Closest CT under 4.5 s away16%130 of 792
    • 4.5 to 7 s away9%31 of 332
    • Over 7 s away6%10 of 163
  3. Two down: save.

    3%Won when attempted20 of 733
    52%Attempted733 of 1,401

Map by map

No map's attempted retakes win more or less often than the rest by more than luck. Ancient one down comes closest: 18% won against 13% expected. Maps differ in how often pros save. One down, Dust2 CTs attempt 59% of retakes against 70% expected for the same players alive. Nuke CTs attempt 81%.

Retakes won when attempted, by map, at even numbers, one player down and two down, and how often CTs attempt one down against what the same players alive would expect
MapEvenOne downTwo downAttempted one down
Ancient39%123 of 31518%44 of 2483%4 of 148
78%Attempts more
expected 70%
Anubis39%71 of 18312%22 of 1842%2 of 94
73%
expected 70%
Cache41%50 of 12311%11 of 1013%2 of 61
64%
expected 71%
Dust234%84 of 25013%24 of 1831%1 of 111
59%Saves more
expected 70%
Inferno37%94 of 2559%16 of 1722%2 of 88
63%Saves more
expected 70%
Mirage37%95 of 25813%30 of 2333%4 of 125
72%
expected 70%
Nuke37%67 of 18314%24 of 1665%5 of 106
81%Attempts more
expected 70%

What moves the line

  • Players alive

    Won when attempted: 37% at even numbers, 13% one down and 3% two down.

  • Distance

    One down: 16% within 1,000 units of the bomb (130 of 792) and 6% from 1,500 units out (10 of 163).

  • Kit

    One down: 14% with a kit (133 of 969) and 12% without (38 of 318).

  • The map

    Dust2 CTs save the most one down and Nuke CTs the least. Won when attempted is within luck on every map.

05Context

Players alive explain Mirage B, Nuke B and Ancient A

Across all 14 sites, CTs are a player down at 56% of plants and a player up at 20%. Where a site strays from that, its expected rate moves with it.

  • Mirage B

    87 of 433won (20%)

    Expected 24%: 3.5 points short, within luck

    This siteCTs a player down or more 60%, even 24%, CTs a player up or more 17%.
    All 14 sitesCTs a player down or more 56%, even 24%, CTs a player up or more 20%.

    CTs are a player down at 60% of plants. So less is expected of it than of Dust2 B, whose raw win rate it matches.

  • Nuke B

    146 of 444won (33%)

    Expected 33%: in line with expected

    This siteCTs a player down or more 46%, even 23%, CTs a player up or more 31%.
    All 14 sitesCTs a player down or more 56%, even 24%, CTs a player up or more 20%.

    CTs are a player up at 31% of plants, the most of any site. So more is expected of it than of any other site.

  • Ancient A

    209 of 653won (32%)

    Expected 28%: 4 points over, within luck

    This siteCTs a player down or more 50%, even 28%, CTs a player up or more 22%.
    All 14 sitesCTs a player down or more 56%, even 24%, CTs a player up or more 20%.

    CTs are a player down at only 50% of plants. They still win 4.0 points more than expected, the most of any site. That is within luck.

06Won v lost

CTs get the first kill in 81% of won retakes and 23% of lost ones

This compares won and lost retakes with the same players alive and the same buys. Won retakes start closer to the bomb and carry more kits. Their CTs get on site together more often. CTs get the first kill after the plant in 81% of won retakes and 23% of lost ones, and part of that is the win itself. Two things don't separate them: the grenades CTs still held at the plant, and the flashes thrown before entering the site.

Won against lost retakes, 1,746 won and 4,148 lost, with the same players alive and buys
MeasureWon 1,746Lost 4,148Difference
What separates them
Kitliving CTs with one58.5%52.5%+6.0 points
Closest CT to the bombabout CT Spawn to Mid Doors v B Doors to Mid Doors824 units973 units0.7 s sooner
TogetherCTs on the site within 2 s of the first1.391.18+0.21
First killCTs get the first kill after the plant81%23%partly the result
TradesCT deaths traded58%11%partly the result
What doesn't
Grenades heldper CT at the plant1.021.06no difference
Flashes before enteringper retake0.440.43no difference
Grenades before enteringper retake1.151.11no difference
07Kits and grenades

Retakes with no kit win about 4 points less often

With the same players alive, retakes where none of the living CTs had a kit win about 4 points less often than retakes with one. With no grenades left the gap is 0.5 points, within luck. In 62 retakes a picked-up kit was the only reason the defuse was in time.

  • −4 points

    Without a kit

    20% won v 24% with one · 1,865 retakes

  • ±0

    Without grenades

    22% won v 23% with them · 1,044 retakes

  • 62

    Defused in time by a picked-up kit

    740 retakes had a kit picked up after the plant

08Teams

A team's retake record in one half of the library doesn't carry into the next

Rank the teams on the first half of the library and their order in the second half is unrelated: a correlation of 0.04 across 25 teams. Beyond luck, teams' retake rates differ by about 0.7 points.

Teams that attempt more retakes also win more of the ones they attempt. Across the 25 teams with 100 or more gun-round retakes on all seven maps, the rank correlation is 0.50 after allowing for the players alive. Luck alone would give that less than 1 time in 100. The data can't say which way it runs: teams that trust their retakes may simply be better at them. Team Vitality are the exception. They attempt 73% of their retakes and win 43% of those, the highest rate of the 25.

Go or save: how often each team attempts the retake, and how many of those it wins25 teams, all maps
Go or save: how often each team attempts the retake, and how many of those it wins, 25 teams, all maps
TeamAttemptedWon when attempted
B8 Esports353 of 436135 of 353
FURIA243 of 31090 of 243
Team Falcons238 of 30381 of 238
FaZe Clan245 of 30093 of 245
Aurora Gaming229 of 29692 of 229
Astralis214 of 27873 of 214
Team Spirit212 of 27276 of 212
MOUZ206 of 27079 of 206
Natus Vincere203 of 26571 of 203
G2 Esports191 of 26259 of 191
GamerLegion200 of 25857 of 200
magic195 of 25655 of 195
PARIVISION175 of 24954 of 175
MIBR197 of 23973 of 197
Team Vitality171 of 23373 of 171
9z168 of 23251 of 168
FUT Esports170 of 21159 of 170
Legacy154 of 20754 of 154
The MongolZ140 of 19337 of 140
BetBoom Team139 of 17544 of 139
3DMAX86 of 12531 of 86
M8078 of 11225 of 78
BIG81 of 11128 of 81
Ninjas in Pyjamas82 of 10330 of 82
Monte73 of 10323 of 73

Every site 60+ retakes, 29 teams

Every site: retakes won against expected, the top and bottom three teams
TeamWonExpectedGap
Top of the table
Aurora Gaming93 of 29679+14
B8 Esports135 of 436119+16
Team Vitality73 of 23362+11
Bottom of the table
Liquid13 of 7119−6
3DMAX31 of 12537−6
magic56 of 25763−7

Dust2 B 12+ retakes, 20 teams

Dust2 B: retakes won against expected, the top and bottom three teams
TeamWonExpectedGap
Top of the table
FUT Esports10 of 216+4
Legacy6 of 164+2
FURIA7 of 306+1
Bottom of the table
magic5 of 318−3
Astralis4 of 317−3
G2 Esports2 of 304−2
Example: x-ray.gg Pro Library, Team Falcons' plays on Dust2 site by site: A site won 77% (74 of 96 hits), B site 67% (58 of 86), Mid 64%
09One to watch

9z retake B from Team Spirit at 3v3 in overtime

Overtime at the IEM Cologne Major 2026. donk opens B with two kills (max and luchov) before dgt catches him just ahead of the plant. The bomb goes down at 3v3 with a kit on every 9z player. dgt picks off sh1ro in CT Mid 5 seconds after the plant. 9z wait until HUASOPEEK's molotov is on Big Box and meyern's flash pops over B Car.

They step on 1 second after the flash, two of them together. zont1x kills dgt, HUASOPEEK trades him 1.6 seconds later and meyern finishes tN1R with the AWP. HUASOPEEK defuses with 3.3 seconds left.

9z v Team SpiritRound 26, IEM Cologne Major 2026

IEM Cologne Major 2026, 13 June 2026Open the full replay in x-ray.gg

10Takeaways

What to take into your next game.

  1. 01From the data

    Carry a kit, and look for dropped ones.

    Won retakes had a kit on 58.5% of living CTs, lost ones 52.5%. With the same players alive, retakes with no kit win about 4 points less often. In 62 retakes a picked-up kit was the only reason the defuse was in time.

  2. 02From the data

    On Dust2, play closer to B as a CT.

    In won Dust2 B retakes the closest CT was 845 units from the bomb at the plant, about CT Mid to the B plant (4 seconds running). In lost ones it was 1,043, about CT Spawn to A Site (4.5 seconds running).

  3. 03From the data

    Retake together.

    Won retakes put 1.39 CTs on site within 2 seconds of the first, lost ones 1.18 (Dust2 B: 1.56 v 1.27). 58% of CT deaths were traded in won retakes and 11% in lost ones. Part of that is the win itself.

  4. 04From the data

    Retake at even numbers, treat one down as a long shot, save two down.

    CTs win 41% of the retakes they attempt at 2v2 and 31% at 3v3. One down, they win 14% at 2v3. Two down, 3% at 2v4.

  5. 05Advice

    Keep at least a flash for the retake.

    Won and lost retakes hold about the same grenades at the plant (1.02 v 1.06 a CT), so the data can't show this wins retakes.

  6. 06Advice

    Practise retakes on a retakes server.

    Clear every angle a T can hold, then do it again with one fewer teammate.

Where these numbers come from

x-ray.gg reads CS2 demos round by round: where all ten players stood, every grenade, every duel and every retake. Every number links to the rounds behind it.

The Pro Library keeps the 15 newest demos of every top-30 team on each map they play. Your own demos and your opponents' are read the same way.

Founders get in first when x-ray.gg launches. Pro is free for 7 days, then your founding price stays while you're subscribed.

Get early access
  • ScoutThe same read on any opponent, site by site, with the rounds behind it.
  • LearnAny role on any map as the pros play it: where they stand and what they do in each situation.
  • Nade sheetsEvery lineup the pros throw, with where they stand and the setpos to copy it.
  • ImproveYour own demos next to the pros', so you can see where your rounds go differently.

Questions

What is the hardest bombsite to retake in CS2?

In pro CS2, Dust2 B. In gun rounds CTs won 116 of 580 retakes there (20%), 29 fewer than the 145 expected for the same players alive. It is the only one of 14 bombsites to fall short by more than chance (x-ray.gg, 1,218 pro demos, 21 May 2025 to 11 October 2026).

Why is Dust2 B so hard to retake?

Its CTs are furthest from the bomb at the plant: 1,783 units on average (about CT Spawn to Long Doors, 8 seconds running) against 1,460 across all 14 sites. They attempt 65% of retakes (76% expected) and win 31% of those (35% expected).

Is Mirage B as hard to retake as Dust2 B?

On raw win rate they are even: Mirage B 87 of 433 (20.1%), Dust2 B 20.0%. But CTs arrive a player down at 60% of Mirage B plants, so 23.6% is expected there. It is 3.5 points short, within chance.

How often do CTs win a 2v3 retake?

In pro gun rounds, CTs attempt 71% of 2v3 retakes and win 14% of those (67 of 485). At even numbers they win 41% at 2v2 (239 of 589) and 31% at 3v3 (162 of 530).

When should CTs save instead of retaking in CS2?

In pro gun rounds CTs win 37% of the retakes they attempt at even numbers and 13% one player down. Two down they win 3% (20 of 733), so two down is a save. One down, the closest CT's distance splits it: 16% within 1,000 units of the bomb and 6% from 1,500 units out. In 1,601 of 1,637 saves at least one CT survived the round.

Which matches is this based on?

1,218 public pro demos from HLTV, 21 May 2025 to 11 October 2026: up to the 15 newest per top-30 team on each map. The biggest events in it are ESL Pro League Season 24 (96 demos), Esports World Cup 2026 (92), IEM Cologne Major 2026 (89) and BLAST Open Porto 2026 (66).

Does saving more lead to better retakes in CS2?

In pro CS2 the teams that attempt more retakes also win more of the ones they attempt. Across the 25 teams with 100 or more gun-round retakes, the rank correlation is 0.50 after allowing for the players alive. The data can't say which way it runs: teams that trust their retakes may simply be better at them. Team Vitality are the exception. They attempt 171 of 233 retakes and win 73 of them (43%), the highest rate of the 25.

Which CS2 team is best at retakes?

No team stands out by more than chance. Teams' retake rates vary by about 0.7 points beyond chance. A team's order in one half of the library says nothing about the next (a correlation of 0.04 across 25 teams). Aurora Gaming (93 of 296), B8 Esports (135 of 436) and Team Vitality (73 of 233) top the table, all too close to call.

How it's measured

Retake
The bomb is planted with at least one CT alive. Gun rounds only.
Expected
What CTs win across all 14 sites with the same players alive.
Sample
1,218 pro demos, May 2025 to October 2026.
Full method

Alive is five minus the side's players killed at or before the plant. The score is CTs first: 2v3 is two CTs against three Ts.

Expected is the combined win rate for the same CTs and Ts alive, over all 14 sites. The gap is won minus expected, in percentage points. Ranges are 95% bootstrap intervals over demos within each map, 2,000 draws. A site is Clear when its Holm-corrected p is under 0.05 across the 14 sites; p < 0.01 is the floor at 2,000 draws. Dust2 B ranks 1st; it could be 1st to 3rd, and is hardest in 63% of draws.

Attempted means a CT was on site after the plant, a T was killed after it, or a CT died within 1,200 units of the bomb. A retake is 25% or more won when attempted, a long shot 10 to 25%, a save under 10% (20 or more attempts).

Distance is the living CTs' average distance to the bomb at the plant; the bands use the closest CT. The Dust2 examples are straight lines between callout centres. Running is about 215 to 225 units a second with a rifle out, so 300 units is about 1.4 seconds.

Time, rounds, top 30 and attempted are the analysis' own views, with ranges and Clear. Players alive is summed in your browser against the all-time expected rates: gaps only, and no gap for a site with fewer than 10 retakes.

Won v lost uses the same players alive and buys, a 95% bootstrap and a false discovery rate over 24 measures. No kit means none of the living CTs had a defuse kit at the plant. 740 retakes had a kit picked up after the plant and CTs won 394; most of that gap is that you must be alive to pick one up.

Teams need 12+ retakes on a site and 60+ over every site, and small samples are pulled toward the site's rate. For carry-over, the library is split at 11 July 2026, with 25 teams having 30+ retakes in each half.

Demos are public pro demos from HLTV only: no pugs, nothing private. The library keeps up to the 15 newest demos per top-30 team on each map, so a team's perma-ban has none. Each map's start differs; Dust2 begins 22 January 2026. 6,980 gun-round retakes come from 1,202 of the demos.