Does Penaltys 50% really filter half the calls?

Yes: 241 of 468 natural candidates admitted (51.5%); 227 suppressed. The 100% arm admitted all 442 of its candidates. No native game fix or changed ROM was needed.

Exact released ROM 20793fce15e640bf60d31c48e95bd0d284105fc07f8ae01ed08b1a579b37e99f.24 comparable controller-driven simulation cohorts per setting,180000 frames each; six mode profiles and four timing variations. Matches and RNG evolve naturally. These are not 24 identical regulation matches.

SettingCandidatesFilteredAdmittedCommitted penalties
100%4420442432
50%468227241236

50% cohort fractions varied 39.1–71.4%, from 7–33 candidates each. A short match cannot establish an exact half-rate. Native delayed decisions can be cancelled or remain pending, so commitments differ from admissions. Icing stayed almost identical: 662 versus 664 admitted events. Other stoppages are independent; stopped-clock transitions are not an audio-whistle counter.

What the percentage controls

One real native RNG draw per eligible new ordinary foul, both teams and CPU/human players. Injury-associated elbowing can become an injury-only report, so its stoppage remains. A new fight is filtered as one whole incident when Fighting is enabled. Already queued calls, penalty-shot variants, goals, icing, offside under Rules, goalkeeper/faceoff events and period/game endings retain their separate native paths.100% retains the original foul frequency; it does not penalize every hit.

Real menus and actual played-save reboots

Fresh default 50%; valid older saved 100% choices stay 100%. Full/Light/Mini locks survive actual completed matches and reboot. The three copied private saves below had four naturally completed matches each, then rejected real Left/Right attempts to change the locked 50% choice. No user save was read or altered. Still images show the selected value, not statistical behavior.

Fresh main menu: Penaltys 50% selected, with framed centred native controls.
Fresh main menu: Penaltys 50% selected, with framed centred native controls.
Full season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.
Full season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.
Light season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.
Light season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.
Mini season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.
Mini season private save after four naturally completed matches: 50% remains selected after a cold reboot and rejected Left/Right changes.

Results by mode

Mode profileSettingCandidatesFilteredAdmittedCommitted
Full-solo100%8008076
Light-coop100%6206262
Mini-coop100%8708787
Quickgame-1v1100%8108179
Quickgame-2v2100%6206258
Allstar-CPU100%7007070
Full-solo50%77393836
Light-coop50%93494442
Mini-coop50%90513939
Quickgame-1v150%75344141
Quickgame-2v250%62263635
Allstar-CPU50%71284343

Measurement correction and limits

The older test observer used an E900 ring that overlaps the current HUD. Its live counts and the enlarged patched-ROM trial were withdrawn. Final measurements use read-only host CPU observation with the released ROM completely unchanged.72 emulated-state controls match with tracing on/off; only compiler-typed host ASLR addresses are rebased, preserving pointer targets. No emulated RAM/cycles/RNG are changed. The official GenesisPlusGX source commit, frozen test hashes, raw-result hashes, uncertainty and controlled-test limitations are documented below. This is not a human hardware or multiplayer acceptance test.

Full method, behavior and limits · Structured results · Read-only observer controls · Actual save/reboot/lock checks · Back to the game