Week 11 Tutorial: Three kinds of player

Build three simulated players of different skill and find out which of them the specimen is actually made for.

20 May 2027

Before the tutorial#

Read Roohi and colleagues as far as their population model. Think of three real people you know who would play this differently.

In the tutorial#

Turn the three people into three policies. Reaction time is the obvious dial. Aim error, how greedy they are for experience, and how early they back away from a boss are all fair as well.

Fill in the table from the lecture for your three. Then answer one question in writing: if you could only keep two of these players, which one are you willing to lose, and at which boss do you lose them?

Where it goes wrong#

• Three players who are one player. The course’s player at 100 ms wins 12 of 51. At 120 ms it wins 7. That looks like two people until you remember week 9: the same 100 ms player on a second set of seeds also won 7. Twenty milliseconds apart is inside the dice. Space your three so that the gaps between them are several times the noise.
• Turning two dials at once. A slow player who is also greedy for experience tells you about neither slowness nor greed. One dial per player, or a fourth player.
• Reading the wrong column. “Reaches the first boss” and “gets past it” are different claims. At 200 ms, 46 runs of 51 reach it and 18 get past it. The gate is doing the losing, and the road to it is not.

What a finished answer looks like#

50 ms, 150 ms, 250 ms. Past the first boss: 50, 38, 11 of 51. Past the second: 39, 5, 0. The specimen is built for the 50 ms player. I would lose the 250 ms player, and I lose them at the first boss.

Afterwards#

These three players are the ones your defence runs. Part two of the seven names them, and part three gives each of them before and after your change.