You broke the base.
How do you finish?
Teams with an inhibitor advantage won 88.5% of the similar situations we analyzed. Here is what separated the clean endings from the 1,644 throws.
Commitment first. Resolution second.
The old analysis grouped teams by their next successful milestone. This version identifies the first observable commitment from positions and state changes, then asks whether that attempt converted, was denied, or was abandoned.
Siege attempt · no conversion
At least two attackers remained within 3,500 units of the enemy base, but the team secured no turret, inhibitor, or Nexus milestone in the following 90 seconds.
- Cases
- 198
- Eventual win
- 76.77%
- Resolution window
- 90 sec
- Evidence
- Inferred
The inhibitor is not one number.
Separate the team that arrived first from the state that got it there, the access event that opened the base, and whether five minutes of super-minion pressure survived contact with the opponent.
What happens after the inhibitor falls
These are retrospective paths, not forecasts available at the moment of destruction.
Judge the call when it starts.
Look at the first real commitment, not whichever objective the team eventually stumbled into.
Shorten the window.
Nexus continuations end fastest. Every extra rotation creates another chance for the defense to recover map access.
Can your team actually do it?
Waves, respawns, health, cooldowns, and teleport angles decide whether the next play is available.
Wins after a push are not the same as good decisions.
Late-game pushes create a trap. A team can attempt a siege, get denied, rotate to Baron, reset, and still finish minutes later. Grouping by the next successful milestone hides those failed or denied calls. The classifier tries to separate the first commitment from the later result, but timelines still miss intent, turret damage, wave health, and comms.
After the second inhibitor falls
Raw win rate describes the selected states. Adjusted probability uses grouped out-of-fold models with only pre-decision timeline state. Neither measures failed attempts that timelines cannot label.
| First successful continuation | Cases | Raw win | Adjusted win | 95% interval | Ends ≤3m | Ends ≤5m | Median time |
|---|---|---|---|---|---|---|---|
| Push NexusLive materialization | 3,172 | 96.5% | 95.1% | 92.8–97.4% | 84.0% | 92.4% | 13 sec |
| Take another inhibitorLive materialization | 988 | 96.2% | 96.0% | 94.4–97.5% | 74.8% | 89.8% | 19 sec |
| Take BaronLive materialization | 431 | 92.3% | 94.3% | 93.2–95.5% | 39.0% | 74.2% | 81 sec |
| Disengage / otherLive materialization | 1,511 | 84.9% | 87.7% | 85.8–89.9% | 65.5% | 79.4% | 180 sec |
Nexus wins the race.
How often each choice ended the game within three minutes after the second inhibitor fell.
The intervals do not pick a winner.
A positive value favors the first action. “Not estimable” means the sample is too thin or the win chance was already near certain.
| Comparison after two down | Adjusted difference | 95% interval | Remaining loss removed | Status |
|---|---|---|---|---|
| triple vs Baron | +1.7 PP | -0.1 to +3.6 | 30% | not identifiable |
| triple vs Nexus | +0.9 PP | -2.0 to +3.5 | 19% | local effect only |
| triple vs disengage | +8.4 PP | 5.6 to +11.1 | 68% | local effect only |
| Baron vs Nexus | -0.8 PP | -3.4 to +1.6 | -16% | not identifiable |
| Baron vs disengage | +6.6 PP | 4.0 to +9.0 | 54% | local effect only |
| Nexus vs disengage | +7.4 PP | 4.2 to +10.2 | 60% | local effect only |
The interval includes either action being better. The timeline also labels successful conversions, not the decision to try them. A failed Nexus rush can land in disengage / other, which is why unsupported comparisons stay unpublished.
The same pattern starts earlier.
14,302 team-game anchors. Nexus conversion again ends fastest; another inhibitor is the most common map-completion line.
| First continuation | Cases | Raw win | Adjusted win | Ends ≤3m | Ends ≤5m |
|---|---|---|---|---|---|
| Push Nexus | 6,357 | 94.7% | 93.9% | 79.6% | 89.0% |
| Take another inhibitor | 3,772 | 95.0% | 93.0% | 61.3% | 82.6% |
| Take Baron | 1,130 | 91.0% | 91.2% | 25.5% | 65.2% |
| Disengage / other | 3,043 | 66.6% | 72.2% | 20.2% | 39.6% |
A hierarchy, not a slogan.
- 1Can you kill the Nexus before the defense stabilizes?
If yes, end. The base is the only objective that cannot be traded back.
- 2Can you safely take the third inhibitor without losing tempo?
Take it when lethal is false and the third wave is already arriving. It buys maximum lane pressure, but the data does not show an extra final-win premium over Nexus.
- 3Are respawns, wave state, or low health closing the base window?
Turn to Baron when it is genuinely secure and useful. The observed Baron cohort ends much more slowly because it is the reset-and-return line.
- 4Is none of that clean?
Leave without donating shutdowns. “Continue” is not a strategy if the feasible action disappeared ten seconds ago.
Conversion analytics on what timelines can see.
21,996 ranked-solo matches and 6,102 two-inhibitor anchors with a pre-decision snapshot. Models adjust for the usual timeline state: gold, XP, CS, structures, objectives, items, side, patch, and region. Wave health, exact cooldowns, comms, and failed calls are still missing, and those decide whether Nexus was actually open.
When the end is open, take it. When it is not, this dataset does not prove the triple-inhibitor or Baron line was the wrong call.
Base continuation comments
Signed-in players can leave notes. Badges show Riot ID, ladder tier when known, and Manalytics champion score.
Loading comments…