mLkoach / Guides
How to analyse your CS2 demo: what the scoreboard doesn't tell you
A scoreboard gives kills, deaths and an ADR. It doesn't say when those kills happened, or what they changed. The demo contains every shot, every buy and every movement — and answers questions the score can't ask.
Where you really stand
Here are three measures everyone knows, taken from 1,015 players whose FACEIT level is known, using the demos of their matches. These are medians: half the players at that level do better, the other half do worse.
| FACEIT level | ADR | K/D | KAST | Players |
|---|---|---|---|---|
| 1 – 2 | 62 | 0.67 | 70 % | 121 |
| 3 | 72 | 0.93 | 74 % | 216 |
| 4 | 80 | 1.00 | 75 % | 303 |
| 5 | 78 | 1.05 | 75 % | 204 |
| 6 | 88 | 1.11 | 75 % | 103 |
| 7 and up | 94 | 1.32 | 75 % | 68 |
Median per player, over their measured matches. Levels 7 to 10 are grouped: beyond that, the sample becomes too thin to honestly tell an 8 from a 10.
Three things stand out in this table, and none of them is what you'd expect.
ADR climbs from 62 to 94 between the bottom and the top of the ladder, +51 %. It's the measure that separates levels best — ahead of K/D, and far ahead of everything else.
K/D only reaches 1.00 at level 4, not before. Below that, killing more than you die isn't the norm: it already means playing above your level.
KAST barely moves — five points across six levels, against thirty-two for ADR. A level 2 player is useful to their team almost as often as a level 7; what changes is what they do when they are. KAST measures presence, not weight.
The level 5 trap. Median ADR drops slightly below level 4. It isn't a measurement error: at level 5, opponents get noticeably better while the player has only just arrived. Every level reached comes with a dip — and that's exactly when you think you're getting worse.
The four numbers that explain a loss
The score says who won. These four say why, and none of them is on a scoreboard.
1. The opening balance
The opening is the first death of the round, whoever it is. The balance counts the times you get the kill, minus the times you're the one who dies. Positive, you give your team the 5 against 4; negative, you give it to the enemy. An entry player at −3 over a match has cost three rounds before they even started.
2. Utility taken to the grave
The shop price of the grenades still in your pocket when you die. It's money spent that produced nothing — and sometimes picked up by the enemy. Our base's median is around 200 $ per round; above that, it's no longer caution, it's forgetting.
3. Avoidable deaths
Dying with a grenade or knife in hand, or right after switching weapons: duels lost in advance, where you couldn't shoot back. They account for a few percent of deaths — and, in our base, they don't clearly drop as the level rises. It's one of the few things that doesn't fix itself as you improve: you have to look at it to see it.
4. Traded deaths
The share of your deaths a teammate trades within five seconds. High, you die in playable spots and your team wins the exchange back. Low, you die alone, and every death costs double. In our base, that share barely moves from one level to the next — around 16 % at the bottom against 15 % at the top. The gap is too small to draw a rule from.
How to read your own demo
FACEIT demos can be downloaded from the match room for a few days after the match. After that they disappear: that's the main constraint on everything below.
- In the game. Open the demo in CS2 and watch yourself in third person. Great for understanding one specific round, useless for measuring anything — nobody reviews twenty-four rounds in detail after every match.
- With an analyser. A program reads the demo tick by tick and extracts measures: that's what mLkoach does. The advantage isn't precision — the game gives that too — it's repetition: forty matches compared with each other tell you what a single match never will.
That's exactly what mLkoach does. You sign in with Steam, hand over a FACEIT demo, and it's read round by round: openings, utility, clutches, crosshair placement, and the real impact of every action on the chance of winning the round. Your numbers are compared with those of 1,483 players measured the same way — not with made-up targets.
Read next: what a good ADR is in CS2 and what really separates FACEIT levels.