POST-FIGHT DECISION LAB · REAL MATCH DATA

What are you fighting for?

This is the actual decision cohort: 9,660 won fights where the team and wave could reach a standing structure before the defense stabilized. We compare the first choice and its adjusted final-win expectation.

THE READ

Wave and neutral are a statistical tie.

The interval crosses zero. The data does not support a universal “always push” or “always take objective” rule in comparable open windows. Both conversion lines separate sharply from giving the window up to reset.

Push structure vs reset: +12.4 pp
Neutral objective vs reset: +13.7 pp
Wave's extra gold swing vs neutral · 120s: +386.1g
Wave's extra structure conversion vs neutral: +30.3 pp

INSPECT DECISIONS BY RANK

DECISION PAYOFF

What each choice bought.

The adjusted column controls for pre-fight state, fight severity, rank context, map region, and estimated arrival window. Pairwise impact is published only where observed choices overlap.

DecisionFrequencyRaw winAdjusted winStructure

Use the wave on a structure4,818 decisions · 49.9%

76.6%raw final win

72.6%69.6%–75.1%

81.5%structure · 120s

Take a jungle objective3,115 decisions · 32.2%

73.6%raw final win

73.9%72.2%–75.7%

51.2%structure · 120s

Reset and spend1,667 decisions · 17.3%

53.4%raw final win

60.2%58.5%–61.8%

17.5%structure · 120s

Take income or continue pressure0 decisions · 0.0%

Withheldraw final win

Withheldlow sample

Withheldstructure · 120s

End through the Nexus60 decisions · 0.6%

74.1%raw final win

Withheldmodel held

98.3%structure · 120s

DECISION RULE SUPPORTED BY THIS COHORTConvert now; choose the asset by map state.

Wave and neutral have indistinguishable adjusted final-win expectations. Resetting in the same class of state is associated with a materially lower expectation. This is an adjusted association, not a causal command.

RANK BEHAVIOR

All rank-resolved fights

9,660 eligible decisions · share of choices and observed outcomes

Use the wave on a structure4,81849.9% chose it · 76.6% raw win · 81.5% structure

Take a jungle objective3,11532.2% chose it · 73.6% raw win · 51.2% structure

Reset and spend1,66717.3% chose it · 53.4% raw win · 17.5% structure

Take income or continue pressure00.0% chose it · Withheld raw win · Withheld structure

End through the Nexus600.6% chose it · 74.1% raw win · 98.3% structure

TEAM-RELATIVE FIGHT CONVERSION EXPLORER

Where does the advantage go?

Every event is rotated into one perspective: your base is always bottom-left and the enemy base is top-right. Color shows the realized change in estimated game state after the fight. Filter the map, then inspect what teams actually converted within 120 seconds.

GAME TIME · 0:00–60:59

StartEnd

RANK

Dragon / Elder

Baron

TOWER STATEClick a tower on the map to cycle Any → Destroyed → Intact. Every map value and next-two-minute outcome will use that exact structure state.

YOUR BASE

ENEMY BASE

YOUR TOP SIDE

YOUR BOT SIDE

Loading fight slices…

—

lower win expectation little movement higher win expectationColor is signed pre/post state movement; no location baseline is subtracted.

WHAT HAPPENED NEXT · IN THE NEXT TWO MINUTES

Conversions from the filtered fights

“Near-even” means a death margin of zero or one. Objective state is reconstructed from observed spawn, capture and respawn timing. Tower state is reconstructed from the destruction milestones observed before each fight. Each 5×5 cell is a 3,000-by-3,000-unit regional average; reset, chase, map pressure and no-conversion outcomes are inferred only when no captured milestone resolves the first action. “All fights” adjusts final outcomes only for the pre-fight state model; it is not a causal estimate of choosing that location. Filtered cells are descriptive and require at least 20observations; they are not causal location estimates.

MEASUREMENT CONTRACT

The count is only the denominator.

fight Location: Median x/z coordinate of champion deaths in the clustered fight; spread is retained as an uncertainty measure.
opportunity Window: Estimated defender-ready time minus the later of team and wave arrival. Open requires at least five seconds; narrow is five to thirteen seconds short.
team Arrival: Straight-line lower-bound travel time from fight centroid to the next standing enemy lane structure using a 390-unit team movement-speed proxy.
wave Arrival: Reconstructed scheduled allied-wave arrival from lane path distance, a 30-second spawn cadence, and 325-unit movement-speed proxy. It is not labeled as an observed minion track.
wave Conversion: The first consequential action is a tower or inhibitor take. Nexus endings are terminal descriptions and are never pooled into wave-conversion impact.

Return to the teamfight study