Re: Henchmen! (Summoner and Thought Invoker so Far)
Posted: 23 Oct 2015, 13:50They get 1 extra life for each worldmagic you have.lujo wrote:Henchmen are the guys who chase you down, have a bunch of (scaling I think someone said) life
Collectible Card Game Headquarters
High Quality Resources for Collectible Card Games and Home of the CCGHQ Team
https://www.slightlymagic.net/forum/
Improved Shandalar General Playtesting & Feedback Thread
https://www.slightlymagic.net/forum/viewtopic.php?f=76&t=17721
Page 4 of 14
Posted: 23 Oct 2015, 13:50Patch notes! XD
Posted: 23 Oct 2015, 21:40Ty! That sounds like something that would nicely apply to monsters with less than 20 life in general, capped off at 20.
---
Anywho, I've made a big ole' updating and revising and tweaking update "patch" as I went through all the decks (and I've kept making notes as I sorted through the pool and spammed testing matches all the time in Manalink).
What I can provide you with right now is just the "patch notes" - I need to put all this stuff into actual Shandalar decks and adjust sideboards and stuff before it can be used, but Moba players dig this sort of thing so I thought you might like it too
You can also yell at me like Dota folks yell at Icefraud when he nerfs their favorite stuff
Oh, and there's the new reworked Paladin list in there too, if the Shandalar AI doesn't play Shifting Sky right, that ought to be looked into to make it do that better. And that "flag which makes the AI stop prioritizing enchantments with non-stacking effects like Shifting Sky and Drop of Honey and a million others" for the AI is kind of necessary.
---
And I've also done Beastmaster, an important deck/version to try out the power level and opressiveness factor of some things, but I can't write it up now, no time.
---
Right, so here's Beastmaster, some thoughts first, then the deck:
---
And here's what I got for Nether Fiend, also somewhat experimental (the OG one was working pretty well lately from what I've seen):
---
It seems like the only enemy at the Vampire Lord - Paladin - Forest Dragon - Sea Dragon - Goblin Lord tier left is Goblin Lord.
That's the most difficult Shandalar deck to tweak from my perspective. It's a long story why, but I've spent hours trying to not make it too brutal or too meh, and I think I'll actually leave that one for last.
---
In the end I'll go with this for now, it's a very small tweak of the OG list:
---
Anywho, I've made a big ole' updating and revising and tweaking update "patch" as I went through all the decks (and I've kept making notes as I sorted through the pool and spammed testing matches all the time in Manalink).
What I can provide you with right now is just the "patch notes" - I need to put all this stuff into actual Shandalar decks and adjust sideboards and stuff before it can be used, but Moba players dig this sort of thing so I thought you might like it too
You can also yell at me like Dota folks yell at Icefraud when he nerfs their favorite stuff Oh, and there's the new reworked Paladin list in there too, if the Shandalar AI doesn't play Shifting Sky right, that ought to be looked into to make it do that better. And that "flag which makes the AI stop prioritizing enchantments with non-stacking effects like Shifting Sky and Drop of Honey and a million others" for the AI is kind of necessary.
- Patch Notes | Open
---
And I've also done Beastmaster, an important deck/version to try out the power level and opressiveness factor of some things, but I can't write it up now, no time.
---
Right, so here's Beastmaster, some thoughts first, then the deck:
- Beasmaster Thoughts | Open
- Beastmaster Deck | Open
---
And here's what I got for Nether Fiend, also somewhat experimental (the OG one was working pretty well lately from what I've seen):
- Nether Fiend | Open
---
It seems like the only enemy at the Vampire Lord - Paladin - Forest Dragon - Sea Dragon - Goblin Lord tier left is Goblin Lord.
That's the most difficult Shandalar deck to tweak from my perspective. It's a long story why, but I've spent hours trying to not make it too brutal or too meh, and I think I'll actually leave that one for last.
---
In the end I'll go with this for now, it's a very small tweak of the OG list:
- Goblin Lord | Open
Tier 2 Enemies
Posted: 24 Oct 2015, 19:29After getting exhausted with the Goblin Lord, I decided I'd do the circle of tier two opponents next, instead of the two-color or tier 3 guys. I haven't done any, yet, and Merfolk Shaman needs a less crazy-ass version (although I do believe the fellow who made it when he says he could've made it way more insane).
Here's Undead Knight, a somewhat flavorful take on stuffing your guys full of Auras, tweaks can be applied as needed in any direction:
---
Priestess
Right, so here's the temporary priestess, it seems to actually be quite decent despite the AI blundering around quite a few things (most of which I've reported after a few hours worth of testing, and there's a couple more I need to see crop up more often to be sure)
---
Sorcerer
Making red decks that don't consistently murder you 20 to 0 in fiveish turns max, don't suck in obvious ways, or both, is a bit of a nightmare.
Anyway after way too much time I have this for the Sorcerer. It might still be unplayable against and the Manalink AI seems to be able to sacrifice Frostling with no target for it, but I'm interested in how this works:
---
Elvish Magi
I don't agree with what Abe did to this deck - it was one of the stronger Shandalar decks, well made and not trying to be tribal at all. Which is good because that either ends up awful if there's not enough support, or too damned good because lords and below 20 formats don't mix too well. If I tried to make this guy tribal I'd get an elfball, and there's no way in hell I'm giving a 2nd tier green opponent anything resembling that, that's insane.
However, it being a deck with 4 Giant Spiders to lock up the place and Hurricanes to blow you out, along with rather unblockable 3/3 dudes to kicks your ass after the Spider sayes "you shall not pass" is also kind of inappropriate for a 2nd tier opponent. I couldn't help respect the power of the OG deck, but it was very annoying that Enchantress and Forest Dragon were disfunctional piles compared to it.
I got a lot to work with, so let's see what I can do later today.
Here's what I ended up with temporarily:
---
Merfolk Shaman
Well, at least this was relaxing. Blue felt a bit constrained when I was building Seer, and the the pool could use quite a large helping of cards that noone adding cards would add to the pool, but at least it's signature effects aren't necessarily too impactful. I toned down the deck that was being used to this:
I'd advise anyone who's in charge of the deck pool which comes with the updates to swap the current Merfolk Shaman deck with this one, even if the aura play isn't too stellar. The one in the pool is waaaaay above curve compared to the rest of the decks. I can't guarantee plopping most of my other decks in there would instantly make things better before I test it, but I think I can for this one.
Here's Undead Knight, a somewhat flavorful take on stuffing your guys full of Auras, tweaks can be applied as needed in any direction:
- Undead Knight | Open
---
Priestess
Right, so here's the temporary priestess, it seems to actually be quite decent despite the AI blundering around quite a few things (most of which I've reported after a few hours worth of testing, and there's a couple more I need to see crop up more often to be sure)
- Priestess | Open
---
Sorcerer
Making red decks that don't consistently murder you 20 to 0 in fiveish turns max, don't suck in obvious ways, or both, is a bit of a nightmare.
- Despair rant I wrote in several incarnations over the last few days but didn't wipe this one off | Open
Anyway after way too much time I have this for the Sorcerer. It might still be unplayable against and the Manalink AI seems to be able to sacrifice Frostling with no target for it, but I'm interested in how this works:
- Sorcerer | Open
---
Elvish Magi
I don't agree with what Abe did to this deck - it was one of the stronger Shandalar decks, well made and not trying to be tribal at all. Which is good because that either ends up awful if there's not enough support, or too damned good because lords and below 20 formats don't mix too well. If I tried to make this guy tribal I'd get an elfball, and there's no way in hell I'm giving a 2nd tier green opponent anything resembling that, that's insane.
However, it being a deck with 4 Giant Spiders to lock up the place and Hurricanes to blow you out, along with rather unblockable 3/3 dudes to kicks your ass after the Spider sayes "you shall not pass" is also kind of inappropriate for a 2nd tier opponent. I couldn't help respect the power of the OG deck, but it was very annoying that Enchantress and Forest Dragon were disfunctional piles compared to it.
I got a lot to work with, so let's see what I can do later today.
Here's what I ended up with temporarily:
- Elvish Magi 0076 | Open
---
Merfolk Shaman
Well, at least this was relaxing. Blue felt a bit constrained when I was building Seer, and the the pool could use quite a large helping of cards that noone adding cards would add to the pool, but at least it's signature effects aren't necessarily too impactful. I toned down the deck that was being used to this:
- Merfolk Shaman Rework | Open
I'd advise anyone who's in charge of the deck pool which comes with the updates to swap the current Merfolk Shaman deck with this one, even if the aura play isn't too stellar. The one in the pool is waaaaay above curve compared to the rest of the decks. I can't guarantee plopping most of my other decks in there would instantly make things better before I test it, but I think I can for this one.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 25 Oct 2015, 17:29Very nicely done, lujo. I keep enjoying your posts, but don't really have the time right now to let it all sink in (and many cards are completely new to me). So it's nice to be able to go back later.
As for the life of the henchmen, I never knew it depends on the number of purchased world magics! This means they max out at 28 life on Wizard difficulty (I've seen 27, but I never buy Tome of Enlightenment...).
And yes, if you beat them enough, they do try to run away from you (but won't ever pay tribute, I think).
As for the life of the henchmen, I never knew it depends on the number of purchased world magics! This means they max out at 28 life on Wizard difficulty (I've seen 27, but I never buy Tome of Enlightenment...).
And yes, if you beat them enough, they do try to run away from you (but won't ever pay tribute, I think).
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 26 Oct 2015, 05:49Ty! Well, it's a good thing you're not trying these out, I had a bit of power creep happen on me and I can only just now test them all out properly (and probably sneak in some cool new stuff too!)
The Merfolk Shaman, though, can most likely just default-replace the current one. The current one is brutal, the tweaked version is somewhat more sane. And quite different from the Sea Dragon, now
I've put up the missing Elvish Magi. Might be a tad too good, but, funnily enough, I managed to keep it the old Elvish Magi AND make it an elf ramp AND lose the hurricane and Giant Spiders to account for how they were a mite too good for a low tier opponent originally. I'm not sure I didn't make it a tad too good again, but it'll get it's testing.
I've also put up a priestess plan, looks sweet in my head and seems to be a good way to actually make a Tallowisp deck actually be low tier. Got to put it togather and run tests, but it ought to be quite an improvement. Not sure if potentially feeding folks Tallowisps is a good idea, but then again, this is temporary stuff.
---
Priestess in! The AI mishandles her in a lot of ways, but it'll get sorted out one way or another and it allowed me to catch some bugs/AI tings (possibly). She's quite playable I think, not the strongest deck in the universe but no slouch either. The deck is in the post with the other tier2 decks, with all the bugs and silly mishandlings noted. As I said I build them around the AI in multiple ways, so a casual observer focused on their own deck (or unfamiliar with particular cards and strategies) might not even notice, but if you take a closer look (and have the debug mode on so you can check the AI's hand and are familiar with the cards/deck/mechanics/intentions) you get to see memorable and eyebrow-raising stuff.
Got a ton of work on porting all the recently tweaked decks into manalink, then soaking up new cards, mapping stuff out, testing stuff in the proper engine finally, and making the rest of the decks. And equipment hit, gotta test that, too. And I can finally test decks at their proper lives (wizard difficulty). Tweak status update might be slower for a while, not enough hours in a day.
The Merfolk Shaman, though, can most likely just default-replace the current one. The current one is brutal, the tweaked version is somewhat more sane. And quite different from the Sea Dragon, now
I've put up the missing Elvish Magi. Might be a tad too good, but, funnily enough, I managed to keep it the old Elvish Magi AND make it an elf ramp AND lose the hurricane and Giant Spiders to account for how they were a mite too good for a low tier opponent originally. I'm not sure I didn't make it a tad too good again, but it'll get it's testing.
I've also put up a priestess plan, looks sweet in my head and seems to be a good way to actually make a Tallowisp deck actually be low tier. Got to put it togather and run tests, but it ought to be quite an improvement. Not sure if potentially feeding folks Tallowisps is a good idea, but then again, this is temporary stuff.
---
Priestess in! The AI mishandles her in a lot of ways, but it'll get sorted out one way or another and it allowed me to catch some bugs/AI tings (possibly). She's quite playable I think, not the strongest deck in the universe but no slouch either. The deck is in the post with the other tier2 decks, with all the bugs and silly mishandlings noted. As I said I build them around the AI in multiple ways, so a casual observer focused on their own deck (or unfamiliar with particular cards and strategies) might not even notice, but if you take a closer look (and have the debug mode on so you can check the AI's hand and are familiar with the cards/deck/mechanics/intentions) you get to see memorable and eyebrow-raising stuff.
Got a ton of work on porting all the recently tweaked decks into manalink, then soaking up new cards, mapping stuff out, testing stuff in the proper engine finally, and making the rest of the decks. And equipment hit, gotta test that, too. And I can finally test decks at their proper lives (wizard difficulty). Tweak status update might be slower for a while, not enough hours in a day.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 26 Oct 2015, 23:03Important note for anyone thinking of using my decks: My sorcerer needs an immediate rework - the Scry cards, and Spite of Mogis in particular are behaving very counter-productively.
I've reported the most silly aspect of this where it tends to completely missplay cards (which may or may not be scry related), but I also think the AI tends to scry lands away too much which makes the deck mana screwed as well (which also makes Pain Kami not work right). The way the deck misuses the concrete scry cards in it also causes it to misplay Frostling (It will cast an unreasonably small Spite of Mogis and then sacrifice Frostling to finish something off). I've also seen it make strangely pointless suicide blocks with solo Frostling, too.
I'm trying to figure out a way to properly report (anything! I try, fail and suck at reporting! but in particular) something about the way the AI makes blocking choices. That Sorcerer up there has a serious problem of choosing to chump block with Pyre Charger rather than Priest of Urabrask. It's evaluation system for blocking is all kinds of strange.
I'm very curious whether summoning sickness figures into any of the AI calculations about blocking choices and Aura targeting when choosing who to buff, because it that was true it could explain a lot of strange things and insane blocking. What I see a lot of, might be coincidence, is the AI madly blocking with things when they have summoning sickness which it doesn't otherwise madly block with. Pain Kami for example - it'll sit on it for ages if it doesn't have summoning sickness, but if it does and you attack with something it'll chump block like mad. Same with most things that have activated abilities (or possibly even all things). The Pyre Charger over Priest of Urabrask often happens like that, or when the AI doesn't have any untapped lands at the moment but behaves like it never will so maybe it thinks the priest is just strictly better or something.
If this were true, also, doing something about it could improve the AI play by leaps and bounds (simply because anyone treating everything with summoning sickness as if it was vanilla is bound to be making mistakes left and right).
The Aura targeting is something I suspect but haven't really kept a close eye on and it might not be anything, too, but it tends to prefer doing things with whatever is furthest to the left on the battlefield. Same with it's own buffs - they'll most likely end up on it's own thing that was in play the longest. I kinda noticed this (or got the notion of it even if was coincidence) whenever the AI was having problems with stacking auras or making spectacular removal missplays in general. Could be I'm way off with that, though, it's a bit difficult to dig into it on my end.
I've reported the most silly aspect of this where it tends to completely missplay cards (which may or may not be scry related), but I also think the AI tends to scry lands away too much which makes the deck mana screwed as well (which also makes Pain Kami not work right). The way the deck misuses the concrete scry cards in it also causes it to misplay Frostling (It will cast an unreasonably small Spite of Mogis and then sacrifice Frostling to finish something off). I've also seen it make strangely pointless suicide blocks with solo Frostling, too.
I'm trying to figure out a way to properly report (anything! I try, fail and suck at reporting! but in particular) something about the way the AI makes blocking choices. That Sorcerer up there has a serious problem of choosing to chump block with Pyre Charger rather than Priest of Urabrask. It's evaluation system for blocking is all kinds of strange.
I'm very curious whether summoning sickness figures into any of the AI calculations about blocking choices and Aura targeting when choosing who to buff, because it that was true it could explain a lot of strange things and insane blocking. What I see a lot of, might be coincidence, is the AI madly blocking with things when they have summoning sickness which it doesn't otherwise madly block with. Pain Kami for example - it'll sit on it for ages if it doesn't have summoning sickness, but if it does and you attack with something it'll chump block like mad. Same with most things that have activated abilities (or possibly even all things). The Pyre Charger over Priest of Urabrask often happens like that, or when the AI doesn't have any untapped lands at the moment but behaves like it never will so maybe it thinks the priest is just strictly better or something.
If this were true, also, doing something about it could improve the AI play by leaps and bounds (simply because anyone treating everything with summoning sickness as if it was vanilla is bound to be making mistakes left and right).
The Aura targeting is something I suspect but haven't really kept a close eye on and it might not be anything, too, but it tends to prefer doing things with whatever is furthest to the left on the battlefield. Same with it's own buffs - they'll most likely end up on it's own thing that was in play the longest. I kinda noticed this (or got the notion of it even if was coincidence) whenever the AI was having problems with stacking auras or making spectacular removal missplays in general. Could be I'm way off with that, though, it's a bit difficult to dig into it on my end.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 03:34Also one thing that must be taken into account but I completely forgot it in AI behavior : the AiDecisionTime setting in Shandalar.ini
It's worth a shoot now that testing is easier to quickly test the same duel with different settings (every patch I put it at 1040 something but I don't see much difference though)
It's worth a shoot now that testing is easier to quickly test the same duel with different settings (every patch I put it at 1040 something but I don't see much difference though)
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 06:34The new Sorceress tweak I've been trying out plays quite a bit better at double the thinking time - and the playspeed difference wasn't noticeable. For one thing I only got sensible Goblin Tunneler activations 5 matches in a row, whereas previously there'd be screwups 4 times out of 5. Thanks A LOT for this tip.
How far do you think it can be pushed before encountering massive slowdown, anyone tried that? Because this just changed an unworkable deck into a very workable one
---
!!!
---
The AI plays some things a bit better at 2160, however it still has serious issues with anything that has "Draw a card" or anything similar in addition to another effect. It can get quite crazy about casting those things at any possible time (except, funnily enough, at the right time), and does the exact opposite of what a player would in most situations - instead of ignoring the "Draw a card" part when determining how to use it, it ignores everything else XD
This is awful. "cantrips" are usually the best variants of any individual effect because drawing a card in addition to doing something you would do anyway is basically the most worth you could possibly tack onto a utility card. Basic cantrips are sorely missing in the pool - porting everything ever with "draw a card" and "draw a card at the beggining of the next turns upkeep", into manalink and into Shandalar (or bypassing Manalink), no matter how insignificant it looks at first glance would be huge. Zap, Afflict, Jolt, things like that would hugely but sneakily improve every low-to-mid tier deck you could think of. If the AI can't use them (or brutally misuses them), this is really bad.
And boy does it ever misuse them, it'll not even bother realizing that it could attack and kill something with Defiant Strike, I've just had it pass the attack phase and cast it in it's second main - and making itself unable to cast other things because of it.
Another thing that simply has to be fixed (inasmuch as anything "has to" be done in the world, ofc), with as much haste as possible, is the AI madly throwing removal as soon as it can target anything. Match after match I watch it Unsummon / Excomunicate a lonely insignificant threat instead of playing creatures. Empty board, no tempo to be gained, it sees a target - bam! If it actually had something to attack with on the board it would be a huge deal, if it waited a turn it would have a better target, heck if it played a creature the player would almost certainly supply it with a better target, but no - it has a target, it'll cast Seething Song to throw a 5cc 5 dmg spell to kill that LLanowar Elf. The game wasn't developed to be played like that, cards weren't made to work like that - I can't scale deck power by giving it more efficient spells because it's more likely to misuse them than benefit from them XD
Just think - if I give it a spell which does 5 damage to a creature for R, it will actually play worse than it would if I gave it a spell which does 5 damage to a creature for 3RR. Because the first one will get thrown at a grizzly bear while the second one has a chance of actually hitting a Serra Angel. The take-it-to-a-logical-conclusion-concequence, which I've also seen a bunch of is that the AI actually manages to play better if it gets mana screwed and misses a land drop than if it doesn't. Same spell, the 3RR 5 damage to a creature thing - the AI missed a land drop and didn't have 5 available on time to kill my Angelic Page, so when I played a Serra Angel and the AI hit the 5th land, it killed the big guy and went on to win. If it wasn't mana screwed I would've beaten it o.0
Whats more is that if it's making stupid plays at stupid times it's also wasting mana on it, which often means that if I give it the most efficient creature curve in the world the easiest way to make it mess it up is if I also give it efficient removal. But if I don't, if I give it removal/spells which it can use somewhat sensibly (read: overcosted), It will get stuck with hands where it can't play anything before turn 4-5 and will be eaten alive by then (and can't deal with a basic curve-out by the player because it can still only cast one of it's overcosted spells per turn).
Otherwise it'll spam Pacifism for 4 turns while 2 Mesa Enchantress sit in it's hand.
It's... difficult.
Oh, and AI blocking decisions are still mind boggling. It's easy to explain what's wrong with it's attacking theory in MtG terms - it hasn't got a way to decide what's a favorable trade for it and is unwilling to sacrifice almost any creature on attack (Even though you could, so to say, train a chimp to attack into trades which would result in two-for-one or three-for-one trades such as trading it's one creature for an enemy creature with an aura on it, for example.) This creates big trouble in terms of deck building because it forces me to give the AI blowout cards which just "win" on the spot (like Winter Blast or Icy Blast), especially in conjunction with how trigger happy the AI is with them.
This means that most decks can only really utilize one strategy - get stuff to be evasive or unblockable so that it would attack with it, which is bad and either uninteractive or easy to shut down for the player (somewhat dependent on the player color), or overload it with overly powerful (and very specific) removal that it's so powerful that you can't misplay it (like the Arc spells which you can't really missplay too hard). Or I have to just plain trick it into missplaying itself into moves like the Hornet Nest + Hail Storm interactions (which is also a variation on giving it removal and evasion).
But it's blocking logic is just strange.
TLDR:
- Nobody has to do anything they don't want to do, or accept anything they don't want to accept. Everyone involved in this does things of their own will and finds their own motivation for what ever they do. This isn't a rant, I'm not demanding anything and I don't want to kill anyone's motivation or make anyone feel bad about anything - I'm just saying this because I happen to be in a position to judge. So might other people, I don't know how good you guys are at MtG so I might just be stating the obvious, but I feel like I have to say this so we know where we are, in case we don't.
- The AI is VERY BAD - it does some stuff better than Manalink, but otherwise it is very lousy at handling some fundamentals, which cause issues. This is why I takes me ages to develop a deck - the way AI treats some stuff goes against how the game was designed to be played in fundamental ways. I suppose that the task of getting it to be better at it might be so daunting and work intensive that noone would even want to try, shrugging the current AI off as "passable". It's very far from passable if the purpose of adding more cards is anything other than having them there for the players to beat a very bad AI with. Specifically:
- AI related to cantrips (cards which have a "draw a card" clause in addition to a different effect like Defiant Strike or Repulse) needs serious fixing.
- AI related to removal and targeted spells in general is causing the AI to missplay the game so fundamentally as to give someone trying to make decks for it a brain stroke. It's waaaaaaay too trigger happy.
- AI doesn't display enough elementary capability to evaluate trades on offense. If you were teaching someone to play Magic, and the person showed complete inability to understand this concept, you would do well to give up and try another game with this person.
- Adding new cards, while cool if you want to play with them as a player, means very little at this point, because the AI can't really use anything but the bluntest, most simple things as is, and getting it to use less blunt stuff takes hours of tweaking decks so it stumbles into something resembling passable play or you make the deck unable to fail to do what you limit it to be able to do. The more cards are in, the more cards might need to be tweaked as the AI gets improved.
Anyway, sorry about the tone, length and all that. I don't mean any ill, I'm just banging my head against the wall here and feel like an idiot every time I try to report something related to basic fundamental things - I want to go through all these decks (and finish off the remaining ones), but just from the 4-5 I've already gone over I can see a AVALANCHE of bug reports. And for a bunch of them I'm not sure it's going to be easier for Korath to just shrug off as none of his business (and I wouldn't even blame him), or that everyone will just decide I'm immagining things or that I'm just a spam bot. And the guy worked his ass off for this update and here I am shoveling him with reports, I feel awful
Right, back to work, Sorceress down, Priest almost there, then Witch, check Druid, check Seer, and I might even have them in the "shandalar rework thread" by the end of the day ready to be enjoyed by everyone.
How far do you think it can be pushed before encountering massive slowdown, anyone tried that? Because this just changed an unworkable deck into a very workable one
---
!!!
Which one of these numbers is a typo!?;Relative time to allow AI to think. The default in the normal game is 5405;
;lower numbers mean quicker (and, to some extent, dumber) decisions.
;
;Corresponds to AiDecisionTime in config.txt for Manalink.
;
;Defaults AiDecisionTime = 540.
---
The AI plays some things a bit better at 2160, however it still has serious issues with anything that has "Draw a card" or anything similar in addition to another effect. It can get quite crazy about casting those things at any possible time (except, funnily enough, at the right time), and does the exact opposite of what a player would in most situations - instead of ignoring the "Draw a card" part when determining how to use it, it ignores everything else XD
This is awful. "cantrips" are usually the best variants of any individual effect because drawing a card in addition to doing something you would do anyway is basically the most worth you could possibly tack onto a utility card. Basic cantrips are sorely missing in the pool - porting everything ever with "draw a card" and "draw a card at the beggining of the next turns upkeep", into manalink and into Shandalar (or bypassing Manalink), no matter how insignificant it looks at first glance would be huge. Zap, Afflict, Jolt, things like that would hugely but sneakily improve every low-to-mid tier deck you could think of. If the AI can't use them (or brutally misuses them), this is really bad.
And boy does it ever misuse them, it'll not even bother realizing that it could attack and kill something with Defiant Strike, I've just had it pass the attack phase and cast it in it's second main - and making itself unable to cast other things because of it.
Another thing that simply has to be fixed (inasmuch as anything "has to" be done in the world, ofc), with as much haste as possible, is the AI madly throwing removal as soon as it can target anything. Match after match I watch it Unsummon / Excomunicate a lonely insignificant threat instead of playing creatures. Empty board, no tempo to be gained, it sees a target - bam! If it actually had something to attack with on the board it would be a huge deal, if it waited a turn it would have a better target, heck if it played a creature the player would almost certainly supply it with a better target, but no - it has a target, it'll cast Seething Song to throw a 5cc 5 dmg spell to kill that LLanowar Elf. The game wasn't developed to be played like that, cards weren't made to work like that - I can't scale deck power by giving it more efficient spells because it's more likely to misuse them than benefit from them XD
Just think - if I give it a spell which does 5 damage to a creature for R, it will actually play worse than it would if I gave it a spell which does 5 damage to a creature for 3RR. Because the first one will get thrown at a grizzly bear while the second one has a chance of actually hitting a Serra Angel. The take-it-to-a-logical-conclusion-concequence, which I've also seen a bunch of is that the AI actually manages to play better if it gets mana screwed and misses a land drop than if it doesn't. Same spell, the 3RR 5 damage to a creature thing - the AI missed a land drop and didn't have 5 available on time to kill my Angelic Page, so when I played a Serra Angel and the AI hit the 5th land, it killed the big guy and went on to win. If it wasn't mana screwed I would've beaten it o.0
Whats more is that if it's making stupid plays at stupid times it's also wasting mana on it, which often means that if I give it the most efficient creature curve in the world the easiest way to make it mess it up is if I also give it efficient removal. But if I don't, if I give it removal/spells which it can use somewhat sensibly (read: overcosted), It will get stuck with hands where it can't play anything before turn 4-5 and will be eaten alive by then (and can't deal with a basic curve-out by the player because it can still only cast one of it's overcosted spells per turn).
Otherwise it'll spam Pacifism for 4 turns while 2 Mesa Enchantress sit in it's hand.
It's... difficult.
Oh, and AI blocking decisions are still mind boggling. It's easy to explain what's wrong with it's attacking theory in MtG terms - it hasn't got a way to decide what's a favorable trade for it and is unwilling to sacrifice almost any creature on attack (Even though you could, so to say, train a chimp to attack into trades which would result in two-for-one or three-for-one trades such as trading it's one creature for an enemy creature with an aura on it, for example.) This creates big trouble in terms of deck building because it forces me to give the AI blowout cards which just "win" on the spot (like Winter Blast or Icy Blast), especially in conjunction with how trigger happy the AI is with them.
This means that most decks can only really utilize one strategy - get stuff to be evasive or unblockable so that it would attack with it, which is bad and either uninteractive or easy to shut down for the player (somewhat dependent on the player color), or overload it with overly powerful (and very specific) removal that it's so powerful that you can't misplay it (like the Arc spells which you can't really missplay too hard). Or I have to just plain trick it into missplaying itself into moves like the Hornet Nest + Hail Storm interactions (which is also a variation on giving it removal and evasion).
But it's blocking logic is just strange.
TLDR:
- Nobody has to do anything they don't want to do, or accept anything they don't want to accept. Everyone involved in this does things of their own will and finds their own motivation for what ever they do. This isn't a rant, I'm not demanding anything and I don't want to kill anyone's motivation or make anyone feel bad about anything - I'm just saying this because I happen to be in a position to judge. So might other people, I don't know how good you guys are at MtG so I might just be stating the obvious, but I feel like I have to say this so we know where we are, in case we don't.
- The AI is VERY BAD - it does some stuff better than Manalink, but otherwise it is very lousy at handling some fundamentals, which cause issues. This is why I takes me ages to develop a deck - the way AI treats some stuff goes against how the game was designed to be played in fundamental ways. I suppose that the task of getting it to be better at it might be so daunting and work intensive that noone would even want to try, shrugging the current AI off as "passable". It's very far from passable if the purpose of adding more cards is anything other than having them there for the players to beat a very bad AI with. Specifically:
- AI related to cantrips (cards which have a "draw a card" clause in addition to a different effect like Defiant Strike or Repulse) needs serious fixing.
- AI related to removal and targeted spells in general is causing the AI to missplay the game so fundamentally as to give someone trying to make decks for it a brain stroke. It's waaaaaaay too trigger happy.
- AI doesn't display enough elementary capability to evaluate trades on offense. If you were teaching someone to play Magic, and the person showed complete inability to understand this concept, you would do well to give up and try another game with this person.
- Adding new cards, while cool if you want to play with them as a player, means very little at this point, because the AI can't really use anything but the bluntest, most simple things as is, and getting it to use less blunt stuff takes hours of tweaking decks so it stumbles into something resembling passable play or you make the deck unable to fail to do what you limit it to be able to do. The more cards are in, the more cards might need to be tweaked as the AI gets improved.
Anyway, sorry about the tone, length and all that. I don't mean any ill, I'm just banging my head against the wall here and feel like an idiot every time I try to report something related to basic fundamental things - I want to go through all these decks (and finish off the remaining ones), but just from the 4-5 I've already gone over I can see a AVALANCHE of bug reports. And for a bunch of them I'm not sure it's going to be easier for Korath to just shrug off as none of his business (and I wouldn't even blame him), or that everyone will just decide I'm immagining things or that I'm just a spam bot. And the guy worked his ass off for this update and here I am shoveling him with reports, I feel awful
Right, back to work, Sorceress down, Priest almost there, then Witch, check Druid, check Seer, and I might even have them in the "shandalar rework thread" by the end of the day ready to be enjoyed by everyone.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 09:45The original game AIDecisionTime was 5405, and for long game it was taking lot of time to calculate play, so it was reduced to 540 for faster decision some month ago.
Edit: I would say even some years ago when it was introduced in Manalink
Edit: I would say even some years ago when it was introduced in Manalink
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 10:33
No wonder the AI is playing like a complete fool. It is too slow at 5400, though, but it's stupidity goes way beyond that. I wrote a big exhasperated rant, not angry at anyone but the universe in general it's in the post above do give it a look, guys, and please don't kill me. I have no social-intelligence- friendly way to express this as benignly as I mean it - but now you know what happens when you get a semi-pro mtg heavyweight to struggle with the current AI
I just want to do what I'm really good at, and I suppose you guys wouldn't be reading the thread if I wasn't, but it's getting to me. I actually admire Korath a lot and enjoy reading every lengthy explanation of whatever he does because I have a good idea of what HE is dealing with on a different front The AI makes me feel like I'm a parent to a child with very special needs which I love the most in the world but is it ever retarded and do I need to hit my fist against the wall about it occasionally
Off to get that priest done.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 12:41Don't be so mad with the poor AI, even with the default longest thinking delay it was still doing dumb choices 
(though after reading the latest gitub patch note this behavior can be partially explained).
With a decent computer (I3 or even good ol'dual core), there should be nothing noticeable at max delay now, and with your current enemy deck pool and unless the player has the best control/annoying/slow deck, it's usually kill or get killed at T5/6.
(though after reading the latest gitub patch note this behavior can be partially explained).With a decent computer (I3 or even good ol'dual core), there should be nothing noticeable at max delay now, and with your current enemy deck pool and unless the player has the best control/annoying/slow deck, it's usually kill or get killed at T5/6.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 27 Oct 2015, 15:31I'm mad because I wasn't trying to make a kill or be killed by turn 6 pool, but got forced by the AI into it. That's the thing, the AI isn't killing you by turn 6 because it's good - it's doing that because it's hopelessly unable to play anything but that. The pool is also a limiting factor because a lot of not-so-lethal themes and strategies were ported into either Manalink or Shandalar with parts missing - people were porting famous tourney stuff which is mostly all exceptional standouts and putting together decks made of those often can't result in anything else. Or they were porting themes and mechanics without having a good enough idea of what holds them togather or what cards that don't have "obviously exceptional" or "obviously cornerstone of mechanic X" written on them. I mean, I understand that Circular Logic is difficult to port, but the idea that you'd go port UG Madness or Psychatog into Manalink without that card is... it's ludicrous. Those decks would have never existed if it wasn't for that card. Odyssey and Wild Mongrel would've been remembered for Red Green Madness (or Frog-in-a-blender) or WG Madness, if Psychatog happened it would've been far less lethal as it wouldn't have had a hard counter for 1 mana (especially post Upheaval, as this means you have to wait a whole turn less), UG madness wouldn't be able to draw+discard for a counterspell with Merfolk Looter, or play Mongrel turn 3 with a hard counter to make sure it sticks on board. The idea of there even being UG Madness without Circular Logic is just 
EDIT:(The point here not being to badmouth Korath for not busting his ass to port Circular Logic, but whoever ported the rest of the damned thing into Manalink after realizing they couldn't do Circular Logic - it's a constructed format, wth. That deck can actually work without Wild Mongrel, and with either Roar of the Wurm or Arrogant Wurm, but not without Circular Logic). Decks don't exist in a vacuum. Or for the previous kind of problem - even if Circular Logic was in so that you could actually put UG Madness togather - it's a monstrous deck, you give that to an AI that can play it, you're asking for trouble. But if I wanted to give the AI a saner deck using the mechanic I can't, because there's no Aquamoeba - that would make the deck possible but it wouldn't be Wild Mongrel who's busted with or without madness. I wanted to make a Dwarf deck for the Sorceress - there's no Dwarven Soldier or anything resembling it, so whatever the 2 drop is can't get tapped for Dwarven Bloodboiler once it inevitably gets stuck (AI can't attack) and can't be pumped by Dwarven Liutenant who is the "pump enchantment" and not your proper 2 drop. So I have to drop all that for something else... Several "oh, damn the AI asses up with this" or "the hell, who ported X without Y, what were they thinking" and hours and hours spent sifting through the pool over and over what I inevitably end up with is most often something too damned good.
And don't get me started on mana bases, dear jesus. MtG would've died an ugly death were it not for mulligan changes, because, and as much as I love it, it has a huge flaw in that too damned many hands are auto-lose for the guy mana screwed or flooded. Previous mulligan changes, not the recent ones. I can't possibly put enough lands in decks to guarantee that he AI won't be mana screwed - the entire MtG world admitted that and just instated the mulligan rules, and then recently decided it didn't help enough and made them even more lax. When you combine that with the fact that the AI missplays removal and spells by being trigger happy with them (meaning I gotta give it expensive stuff in a weenie deck) and then inflate the mana base - they get mana flooded in decks archetypes which only looked the way the did because of the ability to mulligan agressively. A bunch of wholesale-ported decks wouldn't have existed without mulligans. Oh, and I can't give the AI all sorts of lands because players only put them in the deck because they can mulligan if that's the only land they got.
I also seriously need to nerf stuff across the board, but I was waiting for the update and am in the process of doing it going from bottom up. That's why these aren't in the "deck remake thread". I needed to test a bunch of mechanics, too, for evaluation purposes - prowess is cool as all hell to both play and just get your ass kicked by, but it's a bit new and I'm not yet sure where to draw the line.
The level of AI's stupidity forcing decks to be either murderous or harmless (or a completely different outlook when porting cards) is madning. IF you want it to attack you gotta give it pump, if you give it pump it draws several copies and kills you in two-three turns. Or you have to give it evasion, it draws more evasion than the player, it kills you in a few turns. You don't give it evasion it just sits there. You try to give it a mechanic - either it's a brutally simple mechanic and it rolls you over with it, or key cards are missing and it doesn't work. I'm unhappy with 60% of all non-land choices I make because they turn out to be forced or semi-forced by the AI and I go through 4-5 testing cards before caving in and just giving it the cheeze.
It can't bloody attack sensibly, so I have to give it things which win drafts if you have 1 copy in a pile of cards. Icy Blast, Winter Blast, whatever. I'm not saying you guys did a worse job of the AI or anything - I'm saying that the functional decks in the OG were held up with duct tape and that the AI was no measure of anything. The way it is currently, so many cards that have been added, comparatively uncomplicated ones, don't work for it.
---
Anyway, here's Sorceress and Cleric:
Noone asked me to do anything, but you know how Korath occasionally remarks snappily when something about Manalink coding turns out to be preposterous/godawful? I can't tell how bad it is, but I think I can understand the frustration, because all this ranting (and it's actual ranting now) is the very short versions of what can go on with the AI when I try to get it to do elementary stuff.
---
Incidentaly, are all the decks I've made really OP? Like kill-or-be-killed in 5-6 turns?
And now I feel even worse. I'll try to make up for it, sorry folks.
EDIT:(The point here not being to badmouth Korath for not busting his ass to port Circular Logic, but whoever ported the rest of the damned thing into Manalink after realizing they couldn't do Circular Logic - it's a constructed format, wth. That deck can actually work without Wild Mongrel, and with either Roar of the Wurm or Arrogant Wurm, but not without Circular Logic). Decks don't exist in a vacuum. Or for the previous kind of problem - even if Circular Logic was in so that you could actually put UG Madness togather - it's a monstrous deck, you give that to an AI that can play it, you're asking for trouble. But if I wanted to give the AI a saner deck using the mechanic I can't, because there's no Aquamoeba - that would make the deck possible but it wouldn't be Wild Mongrel who's busted with or without madness. I wanted to make a Dwarf deck for the Sorceress - there's no Dwarven Soldier or anything resembling it, so whatever the 2 drop is can't get tapped for Dwarven Bloodboiler once it inevitably gets stuck (AI can't attack) and can't be pumped by Dwarven Liutenant who is the "pump enchantment" and not your proper 2 drop. So I have to drop all that for something else... Several "oh, damn the AI asses up with this" or "the hell, who ported X without Y, what were they thinking" and hours and hours spent sifting through the pool over and over what I inevitably end up with is most often something too damned good.And don't get me started on mana bases, dear jesus. MtG would've died an ugly death were it not for mulligan changes, because, and as much as I love it, it has a huge flaw in that too damned many hands are auto-lose for the guy mana screwed or flooded. Previous mulligan changes, not the recent ones. I can't possibly put enough lands in decks to guarantee that he AI won't be mana screwed - the entire MtG world admitted that and just instated the mulligan rules, and then recently decided it didn't help enough and made them even more lax. When you combine that with the fact that the AI missplays removal and spells by being trigger happy with them (meaning I gotta give it expensive stuff in a weenie deck) and then inflate the mana base - they get mana flooded in decks archetypes which only looked the way the did because of the ability to mulligan agressively. A bunch of wholesale-ported decks wouldn't have existed without mulligans. Oh, and I can't give the AI all sorts of lands because players only put them in the deck because they can mulligan if that's the only land they got.
I also seriously need to nerf stuff across the board, but I was waiting for the update and am in the process of doing it going from bottom up. That's why these aren't in the "deck remake thread". I needed to test a bunch of mechanics, too, for evaluation purposes - prowess is cool as all hell to both play and just get your ass kicked by, but it's a bit new and I'm not yet sure where to draw the line.
The level of AI's stupidity forcing decks to be either murderous or harmless (or a completely different outlook when porting cards) is madning. IF you want it to attack you gotta give it pump, if you give it pump it draws several copies and kills you in two-three turns. Or you have to give it evasion, it draws more evasion than the player, it kills you in a few turns. You don't give it evasion it just sits there. You try to give it a mechanic - either it's a brutally simple mechanic and it rolls you over with it, or key cards are missing and it doesn't work. I'm unhappy with 60% of all non-land choices I make because they turn out to be forced or semi-forced by the AI and I go through 4-5 testing cards before caving in and just giving it the cheeze.
It can't bloody attack sensibly, so I have to give it things which win drafts if you have 1 copy in a pile of cards. Icy Blast, Winter Blast, whatever. I'm not saying you guys did a worse job of the AI or anything - I'm saying that the functional decks in the OG were held up with duct tape and that the AI was no measure of anything. The way it is currently, so many cards that have been added, comparatively uncomplicated ones, don't work for it.
---
Anyway, here's Sorceress and Cleric:
- Cleric | Open
- Sorceress | Open
Noone asked me to do anything, but you know how Korath occasionally remarks snappily when something about Manalink coding turns out to be preposterous/godawful? I can't tell how bad it is, but I think I can understand the frustration, because all this ranting (and it's actual ranting now) is the very short versions of what can go on with the AI when I try to get it to do elementary stuff.
---
Incidentaly, are all the decks I've made really OP? Like kill-or-be-killed in 5-6 turns?
And now I feel even worse. I'll try to make up for it, sorry folks.
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 29 Oct 2015, 12:34The discussion and writing up stuff got me somewhat calmer and even more determined, and writing up threshold for the mechanics thread (it's coming, it's a complex issue and a worthwhile read) got me reworking the Priest into a much less angry thing. It's still buggy as f**k, but most of it's issues are either individual card tweak things or global issues which just have to be adressed. This deck HAS to work evenutally, because if it doesn't that means too many basic things don't work at all, and even in this state it's still a nice ole deck that can spring cool moves on you if it manages to stumble into them. And, well, players won't be looking at it's hand to notice how badly it is missplaying at least some of the stuff, at least not all the time:
Now let's see where we are with other Tier 1 guys.
---
Right, so sorceress writeup, briefer and somewhat better handled by the AI, but with quite a few blunt concessions and compromises on my part (because most of what I tried was too buggy and mishandled in generic ways).
---
Seer needs feedback - really, there's a high number of quality cards here, which might be working or might be failing to work together. If she's beating you - that's not good, she might need nerfing, if she's too pushovery - I'd be curious about what deck's/colors you're using and what's going on in matches.
---
- Cleric, the "I believe in Korath" edition | Open
Now let's see where we are with other Tier 1 guys.
---
Right, so sorceress writeup, briefer and somewhat better handled by the AI, but with quite a few blunt concessions and compromises on my part (because most of what I tried was too buggy and mishandled in generic ways).
- Sorceress 0074.dck | Open
---
- The Seer 0069.dck | Open
Seer needs feedback - really, there's a high number of quality cards here, which might be working or might be failing to work together. If she's beating you - that's not good, she might need nerfing, if she's too pushovery - I'd be curious about what deck's/colors you're using and what's going on in matches.
---
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 30 Oct 2015, 08:58I like this a lot. Those original 'lucky charms' rarely helped. Are there equivalents for the other colours?lujo wrote:I'm also trying out Dragon's Claw in place of Iron Star. AI's handling of lucky charms is slowing it down and might be causing missplays due to how it uses mana. This might work out better in that regard, but is also obviously stronger in raw terms. That's actually somewhat welcome in Red, as red doesn't have lifegain and is likely to be facing red which is fast, nasty and burny. On the other hand, even having these things available to the player in any way can be a problem - you can actually sideboard these into any deck against, for example, red, and people have done this in serious competitive environments to such extent-success that it was known to chase red out of metagames until a core set with less good lucky charms rolled around. Still, we do have to test these things to see what happens and they're much better for the AI than the original lucky charms (I think)
The five castles have lucky charms as starting advantages on Apprentice difficulty. Replacing them with these artifacts would likely be an improvement (or would make beating the castles more annoying, I don't know...)
Re: Improved Shandalar General Playtesting & Feedback Thread
Posted: 30 Oct 2015, 12:37Original cycle: Ivory Cup, Crystal Rod, Throne of Bone, Iron Star, Wooden Sphere, followed by Urza's Chalice in Antiquities.
Darksteel cycle: Angel's Feather, Kraken's Eye, Demon's Horn, Dragon's Claw, Wurm's Tooth, followed by Golem's Heart in Scars of Mirrodin.
M2014 cycle: Staff of the Sun Magus, Staff of the Mind Magus, Staff of the Death Magus, Staff of the Flame Magus, Staff of the Wild Magus.
Paradise Plume is also relevant, though I don't know how well the AI would deal with it coming into play before any of its other cards.
By the time a player's attacking castles, I wouldn't think we'd need to be pulling punches at all. Same reason why I wouldn't have any problem with Dracur being a top-tier land destruction deck or Mandurang a hand destruction deck or so on: by the time those enemies start showing up, you can assume the player's able to put together a deck that can deal with them.
Or maybe I'm just biased because LD and HD and permission and so on were rampant when I started playing. During my brief stint on MODO during the Time Spiral block, I put together a janky Prodigal Sorcerer deck; when I'd play an Island on my first turn without anything else to cast, my opponents would assume I had countermagic and would just scoop. Every single time. Whenever someone gets upset that WOTC only seems to want to print "Durr, just turn all your creatures sidewise" cards these days, I want to shake them and say "It's because you all whined so much whenever anyone played anything else against you, you dolts!"
Darksteel cycle: Angel's Feather, Kraken's Eye, Demon's Horn, Dragon's Claw, Wurm's Tooth, followed by Golem's Heart in Scars of Mirrodin.
M2014 cycle: Staff of the Sun Magus, Staff of the Mind Magus, Staff of the Death Magus, Staff of the Flame Magus, Staff of the Wild Magus.
Paradise Plume is also relevant, though I don't know how well the AI would deal with it coming into play before any of its other cards.
By the time a player's attacking castles, I wouldn't think we'd need to be pulling punches at all. Same reason why I wouldn't have any problem with Dracur being a top-tier land destruction deck or Mandurang a hand destruction deck or so on: by the time those enemies start showing up, you can assume the player's able to put together a deck that can deal with them.
Or maybe I'm just biased because LD and HD and permission and so on were rampant when I started playing. During my brief stint on MODO during the Time Spiral block, I put together a janky Prodigal Sorcerer deck; when I'd play an Island on my first turn without anything else to cast, my opponents would assume I had countermagic and would just scoop. Every single time. Whenever someone gets upset that WOTC only seems to want to print "Durr, just turn all your creatures sidewise" cards these days, I want to shake them and say "It's because you all whined so much whenever anyone played anything else against you, you dolts!"