~Beyond the Border~ > Akyu's Arcade
Fire Emblem Heroes: (ed. NOT) Still Waiting for Tellius
Mеа:
Hm ok thanks guys!
In an attempt to better increase my tactics, I decided to try analyzing how the ai decides to make its moves. I took screenshots of every turn of every battle since then, deliberately remaining passive to see how the ai moves and using up a couple duel crests in the process.
At first it seemed easy, the ai initially 'assigns' a target for each unit and moves each unit independently to head towards the opposing unit that it could deal the most damage to.
But the real wtf quotient starts kicking in when the ai starts attacking you. The more I played, the less sense the underlying decision making process made. An ai marth would attack subaki instead of palla. Sometimes it would ignore a kill to do even less damage (and I'm not talking complex move patterns here). Sometimes it would ignore damage and color advantage (that could have orko'd) to instead attack an off-color unit. Sometimes it chooses to attack the color disadvantaged, bulkier unit instead of the frailer, color neutral one. I don't get it. I can only assume that there's an rng that sometimes causes the ai to make less than optimal decisions. Which I guess can be neat at times when you survive for no intelligible reason, but it makes things conversely hard to predict. Since the first 3-4 turns are what mostly determine the match, even though the first tidbit I guess is all I would need to know and be reassured of its apparentness, it is still a little bamboozling.
e: ah, I just realized, this means that it's very difficult to nearly impossible to separate two identical or nearly identical threats from each other (say, two takumis) from clumping together. You would really have to abuse the secant out of the terrain and initial unit starting positions. If the two takumis or whatever start next to each other, good luck I guess. But then again, y'all run stuff that aren't afraid of him anyway. The best way would be to bait one of them into attacking another unit while staying out of the second ones way. Then while you kite the first one away, the second will continue to go after its initial target (I think... as long as you stay out of its way).
OverlordChirei:
--- Quote from: Mеа on April 10, 2017, 09:33:28 AM ---But the real wtf quotient starts kicking in when the ai starts attacking you. The more I played, the less sense the underlying decision making process made. An ai marth would attack subaki instead of palla. Sometimes it would ignore a kill to do even less damage (and I'm not talking complex move patterns here). Sometimes it would ignore damage and color advantage (that could have orko'd) to instead attack an off-color unit. Sometimes it chooses to attack the color disadvantaged, bulkier unit instead of the frailer, color neutral one. I don't get it. I can only assume that there's an rng that sometimes causes the ai to make less than optimal decisions. Which I guess can be neat at times when you survive for no intelligible reason, but it makes things conversely hard to predict. Since the first 3-4 turns are what mostly determine the match, even though the first tidbit I guess is all I would need to know and be reassured of its apparentness, it is still a little bamboozling.
--- End quote ---
That seems strange. Usually the AI in every FE game can be predicted down to the digit. Having it randomly attack targets just seems very off from what they're known to do in the games.
Mеа:
--- Quote from: OverlordChirei on April 10, 2017, 09:46:19 AM ---That seems strange. Usually the AI in every FE game can be predicted down to the digit. Having it randomly attack targets just seems very off from what they're known to do in the games.
--- End quote ---
yeah... I swear I'm going insane or something. Will probably continue observing for a little longer
Forgive me for the double post, but I'll give the practical application of it here.
Since the ai will initially assign a target for each of its units to go after, ideally you will want to bait the enemy into sending a pair or so of units towards your pair or so of units that can effectively deal with the threats while keeping the bait out of reach.
Eg: the enemy has 2 red sword units and 2 mage units. You send the magically frailest unit off to bait away the mages while your second group contains, say, a blue lance unit and a physically frail the physically frailest green unit to bait the red sword units towards you. Swiftly clean up the reds and go the care of the remaining mages.
Ideally you'd have something like that, but it may not always be so perfectly possible what with the map (bridges, anyone?) or having your bait trap unit invite its own threat too because of its weakness. This will be all the more abusable once pre-battle positioning becomes a thing.
e: managed to get a no death full run of arena with all fliers (yayy for palla in the bonus characters) with this newfound knowledge. Couldn't even break past the 50k rank though. I should probably form an arena scoring team sometime too.
commandercool:
Alright, so you know that thing where my RNG is terrible unless it's getting me Camillas? I just did a gacha pull fishing for Beruka for theme purposes, or another Camilla to add some + to my current one. I uh got another 5* Camilla in one pull. :v
She's +Res, - Spd, which is frankly pretty awful. Could have kept her for redundancy, to try a different set, to inherit Brave Axe+ onto something, or for some kind of insane four-Camillas team, but I decided to merge her with Camilla Prime for +1 HP and +1 Spd.
OverlordChirei:
--- Quote from: commandercool on April 10, 2017, 12:40:47 PM ---Alright, so you know that thing where my RNG is terrible unless it's getting me Camillas? I just did a gacha pull fishing for Beruka for theme purposes, or another Camilla to add some + to my current one. I uh got another 5* Camilla in one pull. :v
She's +Res, - Spd, which is frankly pretty awful. Could have kept her for redundancy, to try a different set, to inherit Brave Axe+ onto something, or for some kind of insane four-Camillas team, but I decided to merge her with Camilla Prime for +1 HP and +1 Spd.
--- End quote ---
could have worked towards the 4camilla dream