apterous.org

Sean D I'm sure we've discussed this in the past, but I don't understand using margin of victory as a tiebreak in this tournament. As all games are against Bots you're double-punishing the player who was unlucky enough to play a high scoring bot.

Robin M Oh dear

Matthew Tassier Or I'm rewarding the player who has performed better against the common opponent.
Also, the easier the game, the more likely your bot will have posted a high score as they are more likely to spot solutions in rounds that can be solved in fewer steps.

Chris Hare "... [The bots] are more likely to spot solutions in rounds that can be solved in fewer steps."

Is this true, or is it just something that feels as though it should be true?

Adam Latchford Bot is most likely to solve a numbers when the numbers has one incredibly difficult solution, that bot is apterous nude and they're doing it just to spoil things (nude will then miss 100x8+5)

Sean D Interesting. I wouldn't have thought the bots were that finely programmed, it looks more like random 'fails' to me, moreso than 'realistic' fails in particular rounds, but I see your rationale

Mike Lee It must be random fails. Case study, why would Prime fail R17 of Brendan’s game for example or R14 of TCap’s?

Tal Lessner I'm quite sure that the random bit is how many operation at most the bot can make in a given round.
Once that is determined, the bot will get the best solution possible for this number of operations.

Bots difficulty is just the distribution of chances for number of ops. For example in Jr, Rex will have 100% for 4 operations, and 0% for the rest, prune 100% for 0 operations. The rest, somewhere in between (like Prime perhaps around 75% for 4 operations, 10% for 3, etc.), so on said Berndan's R17, it's the best solve with one operation.

So Tass is right and bots are more likely to solve rounds with fewer operations required.

Tal Lessner I'm quite sure that the random bit is how many steps at most the bot can make in a given round.
Once that is determined, the bot will get the best solution possible for this number of operations.

Bots difficulty is just the distribution of chances for number of steps. For example in Jr, Rex will have 100% for 4 steps, and 0% for the rest, prune 100% for 0 operations. The rest, somewhere in between (like Prime perhaps around 75% for 4, 10% for 3, etc.), so on said Brendan's R17, it's the best solve in one step.

So Tass is right and bots are more likely to solve rounds with fewer steps required.

Tal Lessner I swear I deleted the first comment.

Matthew Tassier Cheers Tal, that's what I figured, and it feels like it fits with experience of numbers rounds. I'll have a little investigation when I've got some time to see if I can confirm this empirically.

Maria Chandler Lol at my round 10: 2 2 1 1 4 → 99

Dan Southwell Could be the end of the road.... :( but at least I beat the Bot! :D

Fi Thorne I've just saved you Dan!