Advertisement
Piexplode

Untitled

Nov 15th, 2014
244
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 5.06 KB | None | 0 0
  1. <davidstone> In exchange, you take out the worst moves from TM's perspective
  2. <davidstone> Sleep Talk and Roar / Whirlwind / Dragon Tail, for instance
  3. <davidstone> Those are the moves that take the longest to evaluate
  4. <Piex> oh true
  5. <Piex> and RBY lacks all of those
  6. <Piex> mm maybe I was thinking more along the context about how
  7. <Piex> I find some time during the midgame I always need to
  8. <Piex> evaluate what's my plan
  9. <Piex> to have an advantageous engame
  10. <Piex> *endgame
  11. <Piex> AND
  12. <Piex> a possibly more important feature and what would probably make RBY a good testing ground for the 'prediction' elements
  13. <Piex> although GSC could be better arguably
  14. <Piex> is with respect to
  15. <Piex> in all games with very little exception
  16. <Piex> there become points where the player's habits and switching patterns become
  17. <Piex> very relevant to take advantage of
  18. <Piex> and in RBY versus even average, let alone top players, that is fairly profound
  19. <Piex> it just makes me curious how TM would adapt to that, how you would adapt TM to it
  20. <Piex> mm maybe I can find
  21. <Piex> I know I had just the situation yesterday where
  22. <Piex> 2 of my friends were battling and
  23. <Piex> one of them had hooked onto a particular play pattern over several games in a particular scenario
  24. <Piex> and he'd spotted it and turned it again into a secnario of his advantage
  25. <Piex> *scenario
  26. <Piex> mm it might be worth finding the log
  27. <Piex> okay I've found it
  28. <Piex> mm the explanation of why it was good isn't there so I'll do some
  29. <Piex> quick pm hunting tooi
  30. <davidstone> ok
  31. <Piex> http://pastebin.com/i2aEMA9j
  32. <Piex> that's got the first like 10 turns of the battle and then the pm conversation at the bottom
  33. <Piex> I think I'll review it myself to remind myself of
  34. <Piex> exactly what went on
  35. <Piex> ah okay so reviewing it seems that when raish has a weakened snorlax active that's in tauros's hyper beam range but not its body slam range, and he has a fully healthy exeggutor in the wings, Raish's usual play was to
  36. <Piex> bring in his exeggutor on the tauros's hyper beam and from there gain advantage
  37. <Piex> however in this scenario Raish has picked up on Marcoasd picking up this feature of his playing pattern
  38. <Piex> and used it against him to effectively remove his tauros
  39. <Piex> as it was his team had called for him to make that play (he was learning what a team without chansey is like in order to learn how to play better when chansey is removed or frozen)
  40. <Piex> (and as a consequence, exeggutor taking that body slam, with possible crit/paralysis, was too much for the team to bear)
  41. <Piex> whilst it's indeed a lot rarer and a bit more high level than a bot is likely to work at, the fundamental idea of a player being used to the way that the bot evaluates risk could allow a player to consistently abuse it
  42. <Piex> as much as indeed applying that when the opponent is a human being, as was showcased here
  43. <Piex> on a slightly more common level to consider is the use of sleeping pokemon, used by both players ina variety of ways (and to a lesser extent frozen pokemon too)
  44. <Piex> such as relieving specfalls (especially later on in the game)
  45. <Piex> or when a play bring in say a sleeping alakazam versus a chansey say turn 4 or 5 in, then the player owning the sleeping pokemon has the
  46. <Piex> option of staying and burning a sleep turn
  47. <Piex> or going into another pokemon such as exeggutor, lapras, or snorlax
  48. <Piex> meanwhile the opponent has a guarantee'd nice free turn where they can either stay in and attack/twave or something a predicted switch in, or prevent it from burning sleep turns for free
  49. <Piex> or they can switch to a similar threat with minimal cost
  50. <Piex> on both sides of that scenario with a couple of weeks of battling a player who is good at keeping and storing information on battlers would prrobably fairly easily be able to
  51. <Piex> exploit the playing patterns in those very common early game scenarios and use that to
  52. <Piex> gain more consistent victory over the computer
  53. <Piex> so whilst in a first couple of time match where you're unfamiliar with how the computer's decision making process and playing patterns work, it would be a perfectly fine opponent
  54. <Piex> in due course it would become much more predictable unless you introduce some other element that negates this and acts as a benefit rather than a hindrance
  55. <Piex> obviously the effect of this is much more potent in slower paced metagames, and for it to be a good quality bot in XY it probably wouldn't need these types of features to almost any extent whatsoever, but for it to play well in earlier generations, whilst there's more accurate teambuilding information (presumably - at least with context of partners, there's fewer choices)
  56. <Piex> the slower pace and increased importance of play pattern with respect to prediction
  57. <Piex> would make the bot less effective
  58. <Piex> :)
  59. <Piex> I guess when I made some comment on the predictive mechanisms before in which my comments were more or less dismissed, I think this was the sort of idea I was trying to specify, only this time I feel I've conveyed it with much greater clarity
  60. <Piex> ^.^
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement