Week 7: Vision and Information

Most of the map is invisible at any moment, so vision converts gold into knowledge, and knowledge into safer tempo. Builds directly on the jungle (Week 5) and tempo (Week 6) weeks, since both depend on what a team can and can't see.

5 April 2027

Guiding question#

What can you know when most of the map is invisible?

Vision is an information problem, not a placement guide#

This week is not about where to put wards. It’s about what a team can actually claim to know at a given moment, and what follows from admitting the rest is uncertain.

Five kinds of (non-)knowledge#

• directly observed — currently visible: ground truth, right now
• inferred — reasoned from indirect evidence; all of Week 5’s jungle tracking is this
• missing — genuinely unknown, with no direct or circumstantial evidence yet available
• stale — was observed, isn’t current, and degrades in reliability the longer it’s been since the observation was made
• information denial — actively preventing the opponent from directly observing you: clearing wards, pathing around vision, timing a play to outlast a ward’s duration

Uncertainty is what’s left over after sorting through all five. It isn’t a failure state to be eliminated — it’s the permanent condition decisions get made under.

What a ward actually buys#

Not “vision” as an abstract good. A ward changes which actions are safe (a gank threat stops mattering once its approach is visible), which threats are credible (an all-in only works if the target doesn’t see it coming), and which parts of the map a team can commit to with confidence rather than a guess. Week 6’s tempo only pays off if the information behind the decision to spend it hasn’t gone stale by the time it’s spent.

Builds directly on Weeks 5–6#

Week 5 practised inference without naming it as a category; this week gives it the same structure explicitly — observed, inferred, missing, stale — and applies it past the jungle to the whole map. And a tempo decision (Week 6) made on stale information is a bet, not a plan: the window it was aimed at may already be gone.

Map-state exercise#

You’re told:

• your team currently has vision on the enemy top laner and on the blue-side red-buff camp, both empty
• the enemy jungler has not appeared on any ward for the last ninety seconds
• your bot lane’s wave is slow-pushing toward the enemy’s own tower
• Dragon spawns in 40 seconds

Sort this into four buckets before deciding anything:

1. What do we know? Top lane’s state, directly observed. The red-buff camp is clear, directly observed.
2. What can we reasonably infer? The enemy jungler is plausibly dragon-side, given the timer and the incentive their team has to contest it — and your slow-pushing bot wave gives them an additional reason to be down there rather than elsewhere.
3. What do we not know? Whether they’re already in the river. Whether their team is grouping to join them.
4. What decision follows despite that? Send appropriate support or vision toward dragon-side before letting your bot carry commit alone to the pushed wave — the cost of being wrong about a pick under an unwarded bush is worse than the cost of contesting a wave slightly later.

This is exactly the shape of reasoning Assessment 2 asks for, at length, against a real partial game state.