Which CS2 stats matter, and which are decoration

A FACEIT profile shows about forty numbers. Perhaps six of them tell you anything about whether you will climb.

The ones that matter

Win rate. The only outcome measure. Everything else is a theory about why this number is what it is. At 50% you are not climbing at all, regardless of how good the rest looks.

Opening duel balance. First kills minus first deaths. Whoever wins the opening plays the rest of the round a man up, so this is one of the tightest correlations with round wins there is. Negative is a problem; how much of a problem depends on how often you take those duels.

Entry rate. What share of rounds you are in the opening duel for. Read this one carefully, because it is the number most often quoted wrongly — including in an earlier version of this guide, which claimed 10% was normal. Measured across 435 profiles, the median is 18.8% to 20.4% at every level from 2 to 10. It barely moves, because one player per side is in the opening duel and each of the five is in roughly a fifth of them. A rate near 20% tells you nothing; a rate at 26% tells you something.

1v1 clutch rate. Decision-making with no team to carry you. Measured medians: 38.4% at Level 7, 39.1% at Level 8, 37.9% at Level 9, 38.7% at Level 10. So around 38% is typical above Level 6, and 41%+ is good. It is also the flattest metric in the whole table — 2.5 points of spread across Levels 3 to 10 — so its value is in the contrast with your opening duel record, not as a rating.

Utility damage per round. Whether you contribute in rounds where you never win a duel. The measured median runs 4.0–4.7 across every level, peaking at 4.66 around Level 6–7 and falling to 4.06 at Level 9. Under 3.5 is low; over 5.5 is above the median for any level. Numbers like “6.0+ is good” predate the measurement and are wrong.

All of these come from the published baselines table, which is worth reading before you compare yourself against anything you saw in a video.

Per-map win rate, on maps with 15+ matches. Below 15 you are reading a good week.

The ones that mislead

ADR. Measures damage, not whether damage mattered. Detail in ADR vs win rate.

K/D. Better than ADR but still blind to timing. A player who dies first every round and a player who dies last can share a K/D and be completely different problems.

Headshot percentage. Says something about how you aim, nothing about whether you win. High HS% with a 50% win rate is extremely common.

MVPs. Rewards the last kill in a won round. Noise.

Aces and multi-kills. Fun, uncorrelated with rank at this level.

Longest win streak. A statement about variance.

Reading them together

Individual numbers rarely mean anything. Combinations do.

Pattern	What it says
Good ADR + 50% win rate	Mechanics fine, decisions are the problem
Negative openings + high 1v1	Wrong role — you need information, and keep going in blind
Positive openings + low 1v1	Real entry with an endgame problem
Low utility damage + low flash count	Not present for your team’s fights
Good stats overall + big ELO swings	Session discipline, not skill

That last row is worth dwelling on, because no single statistic reveals it. You only see it by grouping matches into sessions and looking at net ELO per sitting — which is where most drawdowns turn out to live.

Why “compare against your level” is the whole trick

A number on its own is not information. 74 ADR is comfortably above the median at Level 3 and seven points below it at Level 8 — the same measurement, opposite verdicts, and you cannot tell which applies without knowing the distribution.

This is the thing almost no stats site does, and it is not because it is hard to display. It is because nobody has the distribution. FACEIT does not publish per-level medians, so a site that wants to say “this is low for you” has to go and measure it: sample profiles at each level, filter out the accounts too small to mean anything, take medians rather than means so one smurf does not drag the band.

We did that — 435 profiles, roughly 40 per level, players with 100+ matches and 2,000+ rounds — and published the result, including the parts that were inconvenient. Two of the findings overturned assumptions this very guide used to state as fact, which is the argument for measuring rather than estimating in the first place.

Two things the public data cannot tell you

Economy. FACEIT does not publish per-round buy data. Anyone claiming to measure your force-buy discipline is inferring it. Use the rules instead of waiting for a number.

Grenade breakdown. Non-flash utility is reported as a single count, so HE, molotovs and smokes are combined. Damage per grenade is exact; the split is not available.

Being clear about limits is not a caveat for its own sake — a tool that pretends to know your HE usage will also happily invent the rest.

How much sample you need before a number means anything

The most common mistake with all of these is reading them too early. Rough thresholds:

Lifetime stats: about 100 matches. Below that your ADR and K/D describe a stretch rather than a player. This is why our baseline measurements only sampled accounts with 100+ matches and 2,000+ rounds — including smaller accounts would have measured variance.

Per-map win rate: about 15 matches. Under 15 you are reading a good week. A map at 40% over 9 matches is not a bad map; it is nine matches. Reports on this site show those maps but refuse to judge them, which is the honest thing to do with a number that cannot support a verdict.

Opening duel and clutch rates: hundreds of rounds, not matches. These are per-round events, so they accumulate faster than match-level stats. A few hundred opening duels — 30 or 40 matches — is enough for a direction.

Session patterns: about 30 matches. Enough sittings to see whether the long ones are the negative ones.

The practical consequence: if you have 60 matches and you are trying to work out your role from your 1v1 rate, you do not yet have the data. Play, then look.

Comparing yourself to a streamer is the other trap

Aggregate stats from a Level 10 streamer are not a target, for two reasons that both come from the measured table.

First, above Level 6 the medians barely separate, so the difference between their ADR and yours is mostly not the difference between their rank and yours. Second, their numbers come from a specific role in a specific setup — a designated entry with four people playing around them has an opening duel profile you cannot replicate by trying harder in solo queue.

Compare against the median for your own level, and against your own numbers from three months ago. Those are the two comparisons that carry information.

The one-minute version

Look at four numbers: win rate, opening duel balance, entry rate, 1v1 rate.

If the win rate is around 50% and the ADR is normal for your level, stop looking at mechanics. The answer is in the other three, and it is usually about which fights you take rather than how well you take them.

Have it read for you — the same numbers, ranked by how much each is costing.