Decision 08 · Week 8 · 22 September 2026
Most Teamfights Are Lost Before They Start
The conditions that decide a fight before anyone commits to it
The claim
Fight outcomes are usually settled by conditions set before the first ability is cast.
The decision: the fight you cannot decline
Constraints this week is about
- forced contact
You cannot decline the fight. Two teams occupy one bridge and disengaging buys seconds rather than minutes, so the question should we fight is not really available. Contact is coming.
What is available is what will be true when it arrives. That is the entire lever this week, and it is larger than it sounds.
The state that decides it#
A fight is not decided by execution as often as players believe. It is decided by a state, and the state is legible in advance. Six variables:
| variable | the question it answers |
|---|---|
| ultimates | how many of theirs are up, how many of ours |
| key cooldowns | whether the tool that makes their engage work is available |
| health | who can be opened on, and who cannot participate |
| summoner spells | whether the target you plan to kill can leave |
| numbers | whether anybody is dead or walking back |
| ordering | week 7’s question, evaluated right now |
A team that fights with four of these in its favour usually wins regardless of mechanical quality. A team that fights with four against usually loses regardless. The band in which execution decides the outcome is narrower than its share of post-game conversation.
Two clocks#
The reason this is a decision at all, rather than an observation, is that the six variables do not move together.
Some regenerate. Cooldowns, ultimates, health, summoner spells — all return on their own, at rates you can estimate.
Some do not. Numbers and ordering only change because somebody does something.
Forced contact means the fight will happen. It does not mean it happens now. Whoever is better at running the clock on the regenerating variables arrives at contact with more of them favourable, and this is most of what strong play in this mode consists of: not winning fights, but arranging to be the team whose ultimates are up.
A case#
Your Riven and Zilean against their Neeko and Bard.
Their whole engage runs through Neeko’s ultimate. It has a long cooldown and it is visible when it is spent. From the moment it goes, there is a window in which the enemy team’s ability to open a fight is materially reduced, and in which they know it.
Two teams read this window differently. One notices and does nothing, because the fight had not started and nobody thought to start one. The other treats the window as the only period in which contact is on favourable terms, and moves before it closes — accepting a fight it would refuse ten seconds later.
- BeforeBoth engage tools availableContact on their terms is possible at any moment, so neither team can commit.
- TriggerTheir engage ultimate is spentIt is visible when it goes, and both teams see the same thing.
- WindowTheir ability to open a fight is materially reducedThe only period in which contact is on favourable terms. Nothing about it is announced.
- OutcomeOne team moves; the other waits for a fight to arriveNeither played better. They fought in different states.
- ResetThe cooldown returns and the window closesOnce it returns, the same fight is one the second team would now refuse.
Nothing in the second team’s play was mechanically better. They fought in a different state.
Zilean’s presence changes the calculation again, because a resurrection is a regenerating variable with an unusually long clock. A fight taken with it available and a fight taken without it are different fights, and the difference is knowable to both teams.
Why teams get this wrong#
Three reasons worth naming.
The regenerating variables are invisible unless you look. Nothing displays the enemy’s ultimate availability. It has to be tracked, and tracking is effort spent during a period when nothing appears to be happening.
Doing nothing feels like not deciding. Waiting through a bad window is a decision, and it is an unrewarding one, because the payoff arrives later and is never attributed to it.
Post-game evidence points the other way. The fight that was lost looks like a fight that was misplayed, because the misplay is the visible part. The condition that doomed it was set thirty seconds earlier by nobody in particular.
The counterargument#
Waiting for a good state is not free, and the strongest objection is that it is not always available to wait. A team that is losing the wave, or being sieged, does not get to choose the moment; the alternative to a bad fight is sometimes a worse position. Perfect condition-management is a luxury of teams that are not under pressure.
That is correct, and it sharpens the claim rather than refuting it. The point is not that you can always fight on good terms. It is that most teams do not know what terms they are fighting on, and a team that at least knows can tell the difference between a fight it chose badly and a fight it had to take.
Takeaway#
You cannot refuse contact. You can arrive at it with your ultimates up, and that is the same decision wearing different clothes.
Teaching team
- Spike — Convener — Game Systems & Decision-Making Lecturer