Winrate of CS2's strange Competitive ranks

Winrate of CS2's strange Competitive ranks

I primarily play the Competitive gamemode in Counter-Strike 2. The game won't even let me play Premier with my friends due to some not having a rank. Despite using the same rank icons and names as CS:GO, it's widely known that there's something very different (or just plain wrong) about Competitive's ranking system.

Recently, CS Stats released a Competitive leaderboard, which shows the highest ranked players on each map. When they Tweeted out that the first Legendary Eagle player was spotted on Mirage. I was curious what it would take to get a rank that high, as my rank barely changes after placement. In the case of the first LE, he was just boosted by cheaters, but he also had a very high winrate.

Winrate vs rank

I decided to look into the correlation between winrate and CS2 Competitive ranks. CS Stats notoriously works to prevent data scraping, but it was the only way to get my answer. So I spun up a script to look through all the players on both of a map's leaderboard, and then scrape the players winrate filtered on Competitive games on that map.

Scraping this data takes a while, and your IP address eventually gets temporarily banned, so I scraped only Anubis, Nuke, and Mirage, which range from the least to most popular map.

A candlestick chart showing a positive correlation between rank and winrate on Anubis
Data as of 2024-02-13

Anubis shows a smooth curve, but only tops out at Master Guardian 2, as the lesser played maps don't see the highest ranks show up. The MG2 players have an average 87% winrate.

A candlestick chart showing a positive correlation between rank and winrate on Nuke
Data as of 2024-02-13

Nuke is middle of the pack in popularity. It gets players up to Master Guardian Elite, but only a handful. You can see the ridiculous 90%+ winrate MGE players need to have to get there.

A candlestick chart showing a positive correlation between rank and winrate on Mirage
Data as of 2024-02-13

Being the most popular map, Mirage sees a steeper curve, but there's also a lack of data in the Gold Nova Master rank. Extrapolating the data, it seems like to get to LEM you may need a near 100% winrate (AKA cheat).

Latest developments

Believe it or not, we recently saw the first LEM player on Dust 2, sporting a 99% winrate with 68% of matches tracked on CS Stats. I haven't looked myself, but word is (in the moderator deleted Reddit comments) that they are cheating.

Maybe some day we'll see the first Global Elite in Competitive when cheaters figure out how to get a 110% winrate.

What about CS:GO

Maybe this winrate phenomenon is normal, and CS:GO only had Global Elites due to the larger playerbase. Well, I scraped the CS:GO winrate and final rank for the final year of CS:GO matchmaking, looking at the same players from the CS2 datasets.

A candlestick chart showing a positive correlation between rank and winrate in CS:GO
Data from the final 365 days of CS:GO, before the release of CS2

While similar looking, it's clear that you did not need a 99% winrate to get LEM or higher in CS:GO. It's about what you'd expect, where the average rank wins about 50% of their games.

Something is definitely wrong with CS2's Competitive ranks.