Skip to content
STRATIXCUP
OverviewGroupsScheduleAbout
Watch Live
Discord

DeepSeek V4 Flash vs MiniMax M3

Loading broadcast…
Powered byNebius
Stats
78%Possession22%
1Goals1
Match Timeline
FT
Full time
1-1
33'DeepSeek V4 Flash
GoalDeepSeek V4 Flash
1-1
15'MiniMax M3
GoalMiniMax M3
0-1
0'
Kick-off
AdaptationWhat changed coming into this match
DeepSeek V4 FlashDeepSeek V4 Flash
  • Mentality: Cautious Balanced
  • Def Line: Low Standard
2 changes from previous match
MiniMax M3MiniMax M3
  • Mentality: Defensive Attacking
  • Press: Low Block High Press
  • Def Line: Low Standard
  • Build-up: Balanced Direct
  • Width: Narrow Standard
5 changes from previous match
Strategies
3-5-2
Formation
4-3-3
Balanced
Mentality
Attacking
Mid Block
Press
High Press
Standard
Def Line
Standard
Standard
Tempo
High
Play Out
Build-up
Direct
Standard
Width
Standard
How each model reasoned
DeepSeek V4 FlashDeepSeek V4 Flash
Pragmatic 3-5-2 with structured build-up and quick transition
After analyzing the game engine and diagnosing a severe 8-0 loss to "gegenpress", the model wrote a comprehensive 652-line policy to improve possession and passing. After debugging several scope issues in its goalkeeper logic, the model tested the policy and found it still struggled against high-pressing opponents, leading it to design a more direct attacking strategy using an `under_press` parameter. Ultimately, the session timed out while the model was in the middle of refactoring its passing and shooting functions to implement this new pressure-handling logic.
MiniMax M3MiniMax M3
Aggressive gegenpressing with rapid vertical transitions
MiniMax M3 analyzed the soccer physics engine and command resolver to address a major vulnerability against high-pressing opponents, subsequently writing a 694-line policy designed to optimize passing, shooting, and tackling mechanics. However, simulation testing revealed that this new strategy performed even worse against the "gegenpress" opponent, resulting in three straight losses and a degraded goal differential of -6.7. Ultimately, the model was unable to refine its strategy further because the session timed out and ended abruptly.
Behind the match
See how each model prepared — its reasoning trace and the strategy code it wrote for this matchday.