Decision 05 · Week 5 · 18 August 2026

The Recommended Build Is Often Wrong

Contextual itemisation under a purchasing window you do not control

The first purchase is made with less information than any other decision in the game and it is harder to undo than any other. That combination is the whole of this week.

What the recommendation knows#

A recommended build is not arbitrary. It is a compression of many games into an ordering, and as a starting point it is usually better than a novice’s intuition. Treating it as noise is as lazy as treating it as an instruction.

But be precise about what it was computed from. A recommendation is an average over games. It does not know your team, the enemy team, which augments you were offered, or that you are the only body between four enemies and your carry. It answers what has usually been good on this champion. The question in front of you is what is good now, and those questions have the same answer only when your game is close to average.

The gap between them is the course. It is the same gap as week 4’s, moved from augments to items.

The constraint that sharpens it#

On Summoner’s Rift a wrong first item is a mistake with a short half-life. You recall in ninety seconds, you buy the correction, the game continues.

Here you cannot. Purchasing is coupled to dying, and dying is expensive. A purchase you regret is a purchase you keep, sometimes for several fights.

This changes the criterion, and the change is not cosmetic:

On the Rift, buy the item with the highest expected value. Here, buy the item you are least likely to regret being stuck with.

Those select differently. The highest-expected-value item is often a specialised one that is excellent against the game you expect. The least-regrettable item is the one that remains useful across the games you might actually get. Under irreversibility, the second criterion is correct and the recommended tab is optimising the first.

A case where the tab is simply wrong#

You are given Garen. Your team is Garen, Sett, Nasus, Illaoi, Taric. The enemy is Lux, Varus, Nidalee, Karthus, Janna.

Every champion on your team is melee. Every champion on theirs is not.

The recommended build for Garen assumes a game in which Garen reaches somebody. It is a good build for that game. You are not in it. In this game the operative questions are whether you can survive the walk, whether anything you buy makes the walk shorter, and whether the team’s actual win condition is that you reach their back line at all — it may be that your job is to be the reason the wave does not kill your tower while four ranged champions fail to close a game they cannot end.

A player following the tab will build to win a fight that will not be offered. The tab is not malfunctioning. It was asked the wrong question.

The four questions#

Before the first purchase, in this order:

The counterargument#

Recommendations are a reasonable default for a player who lacks the knowledge to do better, and a course that teaches contempt for them will produce players who deviate confidently and badly. Deviation without a reason is worse than following the tab, because at least the tab is an average of things that worked.

So the position is narrower than the title: the recommendation is a prior. A prior is something you update, and updating requires evidence. No evidence, no deviation. The failure this week attacks is not following the tab — it is following it without having checked whether this game is the average one.

Takeaway#

Buy for the game you are in, and when you cannot tell what game that is, buy the thing that is least embarrassing to be holding when you find out.