On the live HLTV Confirmed taping in Cologne today, Striker gave an exhaustive preview of what Rating 3.0 looks like, what it improves on compared to the current Rating 2.1 and who stands to benefit (or suffer) the most from it. Here's the quick rundown in text form. Each hyperlink leads to a screenshot of a slide explaining it visually.
New key term: Round swing/win probability
The biggest innovation of Rating 3.0 is introducing the concept of "round swing" – essentially, how a given action by a player influenced their team's probability of winning a round. This effectively replaces the HLTV's current "Impact rating" and takes into account a lot more data. The most basic way to influence a round is, obviously, to get kills in high-leverage situations. The slide gives an outlook at how valuable kills are depending on the number of players left on each side, assuming full buys for both. Read it like this: at 5v5, the T side's win probability is 48.8% (since CS is a slightly CT-sided game overall); should the Ts get a kill, that increases to 70.2%, meaning the player who got the kill just swung a round by roughly 22% in their team's favor. However, if the player gets the last kill in a 5v1, that barely registers – just a 0.1% swing.
Example: Falcons throw a 3v5 postplant on Inferno B site, kill by kill
Here's a timeline of how round swing works. Falcons kill a B defender, increasing their odds from the starting 49% to 70%, get a bomb plant (89%) and kill another CT (96%), but then throw this advantage by giving away kill after kill. Left in a 1v3, Magisk manages to spray down one player, briefly spiking his team's odds back up, but then dies and with enough time to defuse, the probability plunges to 0. The difference between each point on the graph is the "value" of a kill in the mind of Rating 3.0.
Economy is now a factor: kills with lesser weapons matter more
Rating 2.1 treats every kill in isolation and values them equally no matter the weaponry. Not anymore. The weapons are divided into 6 categories: sniper rifles (AWP, autosniper variations) are the most valuable, followed by, tier 1 rifles (AK/M4), tier 2 rifles (every other assault rifle), SMG (Mac-10, MP9 etc), upgrades (P250, Deagle, Tec-9 etc) and starters (default pistols). This table estimates the probability of winning a given duel based on weaponry from both sides; get a kill with a lesser weapon, get a bigger "round swing" value.
Economy-adjusted damage: real-world examples
Here's an example of how some notable players' ADR stat changes with the switch from 2.1 to 3.0, now taking economy into account.
Winners and losers of Rating 3.0: yes to SMG merchants, no to a lot of AWPers
Boombl4 is the biggest benefactor of the switch to Rating 3.0, gaining a whole 0.07 rating points. It's explained by the fact that he's a disciple of the CIS school of the jumping Mac-10, so while he doesn't get a lot of raw kills, they're more impactful on average since he's wielding an SMG against a rifle. Similarly, 3.0's emphasis on economy is an indirect "nerf" to AWP mains, since the AWP is the most valuable weapon in the game and kills with it (even if it's against a big rifle) are valued less than before. Then again, passive AWPers offset a lot of the impact by not giving away dumb deaths.
Rating 3.0 leaders for 2025: largely chalk across the board
As one can expect, the current leaders in Rating 2.1 are still the leaders in 3.0. Notably, sh1ro, despite a lot of buzz around him being the player nerfed most by the new system, actually stays almost exactly where he is (as in, he lost roughly as much rating as all the other primary AWPers) because, while the AWP kills are devalued, he doesn't die needlessly and he's one of the best clutchers in the scene, which will now be emphasized even more than before.
15 Comments
donk & zywoo on a galaxy of their own
Interesting.
Most players either won or lose around between +-0.04 but heavygod got boosted from 1.08 to 1.15 , that’s pretty insane
Comparing the two non-humans players, looks like ZywOo got nerfed more when it comes to ADR and donk more when it comes to the overall rating
i love these kind of posts
Always love to see improvements to better account for the context and soft factors
One of the best clutchers with 16-15 in 1×1 clutches this year. 13 1×2 and 4 1×3.
This narrative is really laughable tbh.
Literally no one really calls Zywoo “one of the best clutchers”. 23-13 1×1, 11 1×2 and 4 1×3.
Ropz 22-10 1×1, 6 1×2, 3 1×3.
Niko 23-11 1×1, 9 1×2, 3 1×3.
Frozen 16-10 1×1, 14 1×2, 4 1×3, 2 1×4.
Monesy 18-18 1×1, 19 1×2, 2 1×3.
Senzu 15-7 1×1, 12 1×2, 3 1×3.
Mezii 19-11 1×1, 9 1×2, 4 1×3.
Where is your shiro superiority in clutching quite literally?
He only wins as much 1×2 and 1×3 clutches because he is the very last to stay and his winrate in 1×1 for some top-5 of the world is quite literally pathetic, barely topping 50% (the only one equally impressive as being bad is monesy, everyone else wins from 60 to 70%). I guess this stat doesn’t calculate % of success in 1×2 or 1×3 otherwise it would’ve been laughable.
i wonder what tiers are assigned to autoshotty or nade kills. if i was in charge i’d probably get the primary weapon from the inventory of the grenader if that’s possible and assign the autoshotty/nova/whatever to the SMG tier.
Donk and zywoo playing in their own tier this year.
Didn’t expect xertion to go down that much. Apex was kind of expected though because he is one of the biggest eco fraggers.
Boombl4 getting the biggest boost is not what I expected.
Oh how broky has fallen. Hope he returns back to form.
Sh1ro having really good stats against top 20 is expected. I just wonder how good his stats will be against top 5, especially against mouz and vitality in playoffs. That’s where I expect sh1ro to receive a nerf.
M0nesy will probably still be top 3 this year, top 4 minimum, even though he is clearly in a slump.
Great post by the way.
frozen and 4 grandaps :<
wouldn’t be a hltv stat without donk at top 1
we’re edging closer and closer to quantifying the eye test
Thank you for compiling this bro
cool stuff
I went to watch the segment, and this this is all really interesting.
…but what the fuck was Machine trying to achieve? Jesus christ he was insufferable.
Am I wrong in feeling like this is just more closely mirroring how Leetify already measures impact?