HLTV editor NER0, also one of the designers of Rating 3.0, posted a lengthy response on an overseas community, addressing the controversy over w0nderful's "18 kills / 10 deaths, yet 0.99 Rating" and explaining why Rating 3.0 gave a 0.99 as well as his personal view on the rating.
w0nderful tweeted complaining about 0.99 Rating
NER0: Let me say this upfront: No matter how we improve the model going forward, an 18-10 K-D corresponding to -3.29% Swing will always look a bit off. Because when we watch matches, we naturally associate performance with K-D — whether it's glancing at the HUD from time to time or checking the scoreboard. And when watching the observer view instead of sticking to a specific player's first-person perspective, it's also harder to track exactly when each kill happens. We often notice the highlight moments, but fail to notice the bad ones.
But these numbers were not pulled out of thin air.
First, let me address a few issues to clarify some misconceptions about why his Rating is so low:
It was mainly Swing, along with damage stats to a certain extent (economy-adjusted ADR of 67), that pushed his Rating to 0.99.
This wasn't because he was a sniper. He only got 3 kills with the AWP in the whole match.
The main factor is: in rifle rounds where both sides are fully equipped, w0nderful's K-D is only 6-8.
Against full ecos, he went 5-0, but these "spec-ops" kills did little to help Swing; against half-buys, he went 2-0. So you can clearly see that once adjusted for economy, his K-D advantage shrinks rapidly.
Without these 7 ECO/half-buy kills, he's 11-10. That makes it easier to understand why his Swing went negative.
In the vast majority of rounds, he didn't dramatically change the game — see the orange bars here:
w0nderful impact visualization on HLTV page
He had 3 outstanding rounds and 4 very poor rounds. The rest basically cancelled each other out: 6 slightly positive rounds and 7 slightly negative rounds.
So overall, his positive/negative rounds are about 9-11.
HLTV editor Nohte's analysis of w0nderful's performance in this event
NER0: But, yeah, in a few spots, Swing was indeed a bit harsh on w0nderful. Compared to Nohte's round-by-round analysis, around -3% is indeed a bit brutal.
So we can take a closer look at a few rounds, and what insights these rounds can offer us, to improve Swing in the future.
Round 14: w0nderful completes 1v2 with Nova
First, since he is using Nova in this situation, the economy correction has already given him quite a lot of compensation, so there is no need to worry about that.
But Swing only considers the economic situation of both teams.
His first kill in this 1v2 retake did not significantly change the win probability. On Inferno, when it is FULL vs HALF economy, the T side has an 88% win rate in post-plant 1v1s.
By the second kill, iM had dealt very high assist damage — 80 damage. In the end, iM gained even more Swing than w0nderful: iM gained +32%, while w0nderful only gained +16%.
This obviously feels a bit off to us. In fact, iM benefits here from the timing of w0nderful's kill.
Regarding what proportion of the credit Swing should allocate to the killer versus the assister in this situation, this is an issue we are already aware of.
Obviously, the 80 damage dealt by iM itself did not increase the team's win probability by 36%.
This is an issue we need to research further.
Round 10: The 2v2 lost by w0nderful and b1t
On Inferno, the T-side win rate in post-plant 2v2s is 65.3%.
We all know what it feels like to get back on defense.
b1t died in Big Pit, but as w0nderful was also killed by the same player within 5 seconds, the system classified it as a failed trade.
So b1t will lose slightly less Swing, while w0nderful will bear part of the loss for this tradable death.
Obviously, we could define trading in a more nuanced way—and that's already on our to-do list. But for this round, I do think it counts as a genuine failed trade.
The post-plant on Inferno is so favorable for the T side that even with b1t already dead, w0nderful still has a 35.5% chance of winning this 1v2.
Therefore, when he was killed in action, this part of Swing was completely lost.
This is indeed a big chunk of Swing. But Inferno's post-plant situations are inherently very T-sided. A metric like Swing is characterized by punishing the other poorly played post-plant rounds even more harshly, namely the red bars on the website:
Round 17:
NAVI have a 3v2 advantage. w0nderful dies in Connector, then 3DMAX successfully plant the bomb.
This was a very costly death, which is hard to refute.
Round 18:
This is a bit harsh.
At that point it was basically already equivalent to a post-plant situation, and after w0nderful died at Graveyard, iM immediately got the trade kill.
Here w0nderful was given a -27% Swing, which is indeed a lot. Even though a 1-for-1 trade in this situation does increase the T side's win probability, this penalty still seems too harsh.
As I mentioned before, if we had a more granular trade-kill metric, then for instant trades like this, w0nderful should probably be given a bit more credit, while reducing the credit currently given to iM.
Round 22:
3v2 retake. w0nderful is the CT, Maka kills him with the AWP from the new box position.
He actually didn't have much room to operate, but this death was indeed the pivotal moment of the entire turn, so it will create a very large Swing.
Summary:
Yes, I think Swing is indeed a bit harsh on w0nderful.
His 1v2 with the Nova was undervalued because iM had previously dealt 80 points of assist damage.
And some of his deaths, like in Round 18, were actually not as bad as the numbers make them look.
But overall, one thing is very clear:
This game wasn't actually as impressive as the 18-10 K-D makes it look.
A considerable portion of his kills come from ECO rounds, and his deaths often occur at critical moments that clearly swing the round in favor of NAVI's opponents.
0.99 is indeed a bit harsh. For a performance that still has a +5 K-D after economy adjustment, this Rating may still be a bit low.
But I think, compared to giving him a 1.50 Rating just because of an 18-10 K-D, 0.99 is actually not that outrageous — because in this match, the times he negatively impacted rounds actually outnumbered the times he positively impacted them.
Of course, I also think that in many specific instances, the exact percentages given by Swing do seem problematic intuitively.