[confirmed]Goblin Test Pilot bias

by **Korath** » 30 Jul 2018, 15:50

Describe the Bug:
Goblin Test Pilot doesn't choose sufficiently randomly from among its legal targets - if both players have at least one legal target (including the players themselves), there's a 50% chance of choosing from each player, then an even chance from among that player's targets. For example, if you just have a Goblin Test Pilot, and your opponent has 49 creatures, there's a 25% chance of it choosing you, 25% chance of it choosing the Goblin, a 1% chance of it choosing your opponent, and a 1% chance of it choosing each of his creatures.

Which card did behave improperly ?
Goblin Test Pilot.

Which update are you using?(date,name)Which type(Duel,Gauntlet,Sealed Deck)
Dev master head, 6287802.

What exactly should be the correct behavior/interaction ?
Equal chance for each legal target. In the example above, it would be a 1 in 52 chance for each.

Are any other cards possibly affected by this bug ?
No.

by **Aswan jaguar** » 30 Jul 2018, 17:39

I don't think such cases can be tested with somewhat safe results if you don't have a high number of tests at least 100 or check the algorithm for it.
It does seem somewhat biased in 20 activations tested.

I noticed another problem that it doesn't deduct from the chosen targets correctly legal targets. AI had in play only an Autumn Willow and I guess that is what was chosen to be targeted but because it couldn't be targeted instead the AI's card with the least in game id was. This was a Cytoshape in AI's hand.