We put E1 52% winrate players in Master+ lobbies but a 76% wr master player gets into D1/low master lobby. Nothing wrong with the matchmaking
1
1
19
Replying to @detdert @drewlevin
There is nothing wrong with matchmaking and the ranked solves the issue
1
221
Replying to @Okabekin
There is nothing wrong with matchmaking system and the ladder reset won't just result in the same shitty games we had before the reset
2
139
Wonder how it works on coding/agentic tasks or real use cases, while topping benchmarks, we've seen Opus do that many times, yet still failing for real usage.
32
Gemini 3.1 Pro ranks dead last among frontier models on BridgeBench Reasoning.
Behind Grok 4.20, GPT 5.4, Claude Opus 4.6, Qwen 3.6 Plus, MiniMax M2.7, Claude Sonnet 4.6, and GLM 5.1.
Google's flagship model can't even beat a free Chinese model on grounded reasoning.
This is why I cancelled my $250/month Google AI Ultra subscription.
Gemini CLI was unreliable.
The model is mid.
The infrastructure is worse.
Google has the compute.
Google has the data.
They just can't ship a competitive coding model.
2
26
Replying to @RiotPhroxzon
@RiotPhroxzon @drewlevin The matchmaking is great, but I feel like there might be something wrong here. Could anyone tell me if my feeling is correct?
3
7












