Week 6 Tutorial: Take one away

Ban each of the specimen's seven upgrade cards in turn and measure what your simulated player loses.

1 April 2027

Before the tutorial#

Read Jaffe and colleagues up to the end of their definition of restricted play. Write down which card you think is worth the most. Seal it.

In the tutorial#

Baseline first: your delayed player turned down to 100 ms, leftmost card every time, 51 seeds. At 150 ms it wins 2 of 51, which leaves a ban almost nothing to take away. Then seven restricted runs. The only change each time is the card your player is forbidden to take.

Sort the table by wins lost. Open what you sealed.

Where it goes wrong#

• Fifteen runs. At 15 runs the baseline is 3 wins and the seven banned rows are 2, 2, 1, 0, 3, 2, 3. Every row is within three wins of every other and the table ranks nothing. At 51 runs the same rows are 7, 7, 3, 1, 9, 7, 10 against a baseline of 12, and pickup reach is plainly worth the most.
• A baseline from somewhere else. The baseline has to be the same player, on the same seeds, on the same day. A remembered 12 from last week, after you have edited the policy, is a different experiment.
• A ban that changes the player. When the banned card is offered, take the next one along. Do not skip the level-up: a player who takes nothing that round is weaker for a second reason, and you cannot separate the two.

What a finished answer looks like#

Baseline 12 of 51. Most costly ban: pickup reach, 1 of 51, 11 lost. Least: shockwave, 10 of 51, 2 lost. I had sealed “damage”. It lost 5.

Afterwards#

The table is this week’s quiz answer. Assignment 1 is due after the break; this table is allowed to appear in it.