How often do you need to win the river fight before contesting beats leaving for two waves of farm? At even gold that bar sits at 58.24%1. At a coin-flip fight the model still prefers leave by +1.94pp2.

The Patch 26.11+ objective term is 124.13g3: 90g cash plus an upper-bound plate-progress equivalent. The progress term is not guaranteed immediate gold.

Interactive figure unavailable: the current publication data did not pass validation.

Figure 1. Current-mechanics sensitivity by fight-win chance, leave-farm package, and pre-contest gold lead. The dashed line is modeled indifference at even gold.
Fight-win chanceContest valueLeave valueEdgeModel preference
15%-8.24pp+1.94pp-10.18ppLeave higher in model
20%-7.06pp+1.94pp-9.00ppLeave higher in model
25%-5.89pp+1.94pp-7.83ppLeave higher in model
30%-4.71pp+1.94pp-6.65ppLeave higher in model
35%-3.53pp+1.94pp-5.47ppLeave higher in model
40%-2.35pp+1.94pp-4.29ppLeave higher in model
45%-1.18pp+1.94pp-3.12ppLeave higher in model
50%0.00pp+1.94pp-1.94ppLeave higher in model
60%+2.35pp+1.94pp+0.41ppContest higher in model
70%+4.71pp+1.94pp+2.77ppContest higher in model
75%+5.89pp+1.94pp+3.95ppContest higher in model
Figure 2. Article curve table — contest value, leave value, and edge in percentage points of map win rate.

Limits on the estimate

The gold-at-10 conversion behind the article estimand is associational4— it converts the fixed gold scenarios into modeled map-win association. It does not make this comparison causal or estimate a live fight from champions, position, vision, cooldowns, or player form.

  • Gold-at-10 to map-win is an associational conversion, not a causal effect.
  • Fight-win probability is exogenous rather than estimated from champions, position, vision, cooldowns, or player form.
  • The 34.13g Touch term is an upper-bound plate-progress equivalent, not guaranteed immediate gold.
  • The comparison does not identify a live action policy and should not be used as a universal contest threshold.