Final: One Game, Twelve Decisions

One game from champion roll to nexus, stopped twelve times — decide at each stop, then write the post-game analysis

One game, stopped twelve times. You decide at each stop with only what was knowable then, and afterwards you argue about which of your decisions actually mattered.

The game is issued to you. You do not choose it, and it is not a game I selected because it makes a tidy point.

How it works#

The case arrives as twelve sequential states. You see each one only after you have committed to the previous. This is the whole design of the assessment: it is easy to describe good decision-making with the ending in front of you, and it is worth almost nothing.

The twelve stops follow the game’s own order:

1. champion assignment
2. bench and reroll
3. the first augment
4. the first purchase
5. the composition, once all ten are known
6. the first extended contact
7. a reset decision
8. a threat you had discounted
9. a build revision
10. a disadvantage state
11. the late commitment
12. the position at the end

At each stop: your decision, the alternative it beat, and the assumption it rests on. Two hundred words is usually enough for one stop and three hundred is usually too many.

The post-game analysis#

The second half of the submission, and the half that carries the marks.

Having seen how the game ended, argue which single decision most moved the probability of winning, and defend the choice. It will very often not be the decision that preceded the most visible event. The fight you lost at stop six may have been decided at stop four; the death at stop seven may have been the best thing you did.

Three requirements:

Keep decision quality and outcome apart. Some of your decisions were good and lost. Say so, and defend them anyway. Some were bad and worked. Say that too — it is the harder half and it is where the marks are.

Name at least one decision you would now make differently, and one you would not. The second is not padding. A submission that revises everything in hindsight has not analysed anything; it has apologised.

Do not rewrite your earlier reasoning. The stops are a record. If stop three now looks foolish, the analysis is the place to say why, not the place to quietly improve what you wrote before you knew.

Roughly 3,000 words in total across both halves.

How it is marked#

The same four criteria, and the same principle as the rest of the course: the argument is marked, not the result. A submission whose player lost the game can receive full marks. A submission that won and cannot say why is a weak submission.

The final is where the course’s own claim gets tested on you. If you don’t need better rolls, you need better decisions is true, then a student who was issued a difficult case should be able to score as well as one who was issued an easy one. That is the intent, it is why the cases are issued rather than chosen, and you are welcome to argue in your analysis that it does not hold.

Practical#

Submitted through the course page by the deadline. Cite the patch named in your case. Where you use data, source it. Where a claim is yours rather than the data’s, say which it is.