• Play
  • About
  • News
  • Forums
  • Yppedia
  • Help
Welcome Guest   | Login
  Index  | Recent Threads  | Register  | Search  | Help  | RSS feeds  | View Unanswered Threads  
  Search  


Quick Go »
Thread Status: Normal
Total posts in this thread: 33
Posts: 33   Pages: 2   [ First Page | 1 2 | Next Page | Last Page]
[Add To My Favorites] [Watch this Thread] [Post new Thread]
Author
Previous Thread This topic has been viewed 7644 times and has 32 replies Next Thread
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Revamp Pillaging AI Reply to this Post
Reply with Quote

I propose that the current AI found during pillaging is not fit for purpose, nor has it been since 2006 when BK blockades were introduced. Currently, battles are time consuming and involve being overly cautious due to the random moves the AI likes to throw out. This is due to the AI overhaul when BK blockades were released, as the old one did not account for friendly ships.

The old AI was punishing to mistakes, but rewarded navigators who had mastered the puzzle. As it is now, anyone with a good understanding of mechanics can win against even the hardest bots, albeit in lengthy fights.


Hopefully the old code exists, and the AI can be split how it should have been in the first place. If not, perhaps the community and devs can work together to create something that is fit for purpose.


I urge anyone who navigated, especially on hard routes, before the change, to give their input.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 10, 2017 12:14:01 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
patgangster

Member's Avatar


Joined: May 26, 2006
Posts: 576
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

If I remember it correctly, the old AI was awful, would go straight for you and try to grapple most of the time and one who learned their moves could 100% predict exactly what the AI was going to do. I'd rather not see that back in the game, the AI is supposed to feel more random and less predictable than that.

I'd be ok with another revamp to change it up again, but please don't go back to the old one.
----------------------------------------
TriplePat, Emerald.
[Jul 10, 2017 12:41:40 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
LJAmethyst

Member's Avatar


Joined: Jul 19, 2007
Posts: 4140
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Unlikely to happen. Whether it is technical hurdles or a philosophical objection, they have not reverted the AI even after there were tons of complaints at the time. They pretty much won't even discuss it.

It's a fine line - are those who are "rewarded by mastering the puzzle" the best puzzlers or are they exploiting a game mechanic unjustly? I think there's a great argument for the former, but others may see it differently.

Anyway, they're concentrating on Obsidian, where PVP is king. Delving into bits of the code that aren't related, for questionable profit, is not on the table right now.
----------------------------------------
Retired as of August 2015.
Sic transit gloria mundi.
[Jul 10, 2017 12:41:59 PM] Show Printable Version of Post        Send Private Message    http://www.newadvent.org/bible/jon001.htm [Link]  Go to top 
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

LJAmethyst wrote: 
Unlikely to happen. Whether it is technical hurdles or a philosophical objection, they have not reverted the AI even after there were tons of complaints at the time. They pretty much won't even discuss it.

It's a fine line - are those who are "rewarded by mastering the puzzle" the best puzzlers or are they exploiting a game mechanic unjustly? I think there's a great argument for the former, but others may see it differently.

Anyway, they're concentrating on Obsidian, where PVP is king. Delving into bits of the code that aren't related, for questionable profit, is not on the table right now.


I think this is exactly the type of change that will help retain players. Decisive, shorter battles would help keep people interested.

The devs have proven they will make changes when necessary. Hiding moves went from an insurmountable task to implemented in under 2 months. The recent blockade stat requirement changes were a direct response to player needs. Coding the ocean to have sinking and non-sinking zones will also have involved changes to the underlying code, as will many of the other changes made for Dark Seas. Let THEM respond to the players about what they are and aren't willing to do.


patgangster wrote: 
If I remember it correctly, the old AI was awful, would go straight for you and try to grapple most of the time and one who learned their moves could 100% predict exactly what the AI was going to do. I'd rather not see that back in the game, the AI is supposed to feel more random and less predictable than that.

I'd be ok with another revamp to change it up again, but please don't go back to the old one.


That was one element of it, but there was much more to it than that. The AI actively went after you, which would be expected from a ship that was larger and had more pirates on board. Why should a WF with 38-42 randomly run from a brig with 30? The main thing the AI needs is consistency. Being out tonight and watching the enemy run and do completely unjustified moves simply reinforced to me that it needs to change. Hopefully some other people who regularly ran pillages on difficult routes before the change will weigh in as well.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 10, 2017 2:04:49 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Scarpath

Member's Avatar


Joined: Sep 7, 2014
Posts: 246
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Would this also be the good spot to ask for a change to the Rumble difficulty to better match that of the battle navigation one? As it is now, I'm getting annoyed by all the harder route pillages set only to SF because once you get up to a yellow ring, the rumble is near impossible.
----------------------------------------
Scarpath on Emerald, Cerulean, and Meridian!

Also Scarpath on Obsidian, Defiant to the core.

Your grammar should at least be as good as mine, take the time to make it decent!
[Jul 10, 2017 2:16:57 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
LJAmethyst

Member's Avatar


Joined: Jul 19, 2007
Posts: 4140
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
Let THEM respond to the players about what they are and aren't willing to do.


That happens infrequently enough that I just felt I would save them some time and predict a likely answer. I never said I speak for GH, but I've been around the block a while, especially on Game Design, and I have a good sense for what is and isn't a priority to them; that said, once in a while I am surprised.

Consider the situation like Pac-Man and Ms. Pac-Man. When Pac-Man came out, gamers quickly realized that the ghosts were moving in determinate patterns, and devised counter-patterns that could be used to completely win the game, as long as no mistakes were made.

When Ms. Pac-Man came out, the ghost movement was modified to randomize it, with much more unpredictable movement, and the result was that nobody ever devised patterns to beat Ms. Pac-Man and an element of skill was re-introduced. Ms. Pac-Man was wildly popular, capitalizing and building on the previous game's success. I'm sure some of the hardcore gamers were dismayed that their patterns were busted, but most people were overjoyed to try their mettle against the bad, bad, ghosts in the machine.
----------------------------------------
Retired as of August 2015.
Sic transit gloria mundi.
----------------------------------------
[Edit 2 times, last edit by LJAmethyst at Jul 10, 2017 3:57:04 PM]
[Jul 10, 2017 3:50:22 PM] Show Printable Version of Post        Send Private Message    http://www.newadvent.org/bible/jon001.htm [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

I don't see a need to revert the AI to an older version. The older version was predictable. I think it's good that the current version throws in random moves and doesn't seem to always use the same strategy in battle.

 
Would this also be the good spot to ask for a change to the Rumble difficulty to better match that of the battle navigation one? As it is now, I'm getting annoyed by all the harder route pillages set only to SF because once you get up to a yellow ring, the rumble is near impossible.

IMO, the problem isn't with the NPPs, it's with the players not being good enough at rumble.
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
[Jul 10, 2017 6:21:22 PM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
EmpressTamar

Member's Avatar


Joined: May 7, 2012
Posts: 830
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.
----------------------------------------
Empresstamar of Emerald.
Yppedia page

Gorgeous avatar by by Cattrin <3
[Jul 10, 2017 8:26:14 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Scarpath

Member's Avatar


Joined: Sep 7, 2014
Posts: 246
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.


Yeah, really does seem take forever... As a jobber, or even when running a pilly, it can get pretty frustrating. My parents do a very good job of asking for help with stuff at random times, and I can't just walk away in the middle of a battle. However, the battle can take upwards of 15-20 mins just in the bnav puzzle, meaning my parents are done by the time I can even leave the computer without worrying about getting left on a do not job list.
----------------------------------------
Scarpath on Emerald, Cerulean, and Meridian!

Also Scarpath on Obsidian, Defiant to the core.

Your grammar should at least be as good as mine, take the time to make it decent!
[Jul 10, 2017 9:08:36 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.

That should tell you who not to job for.
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
[Jul 10, 2017 10:22:57 PM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
Shinito

Member's Avatar


Joined: Jan 3, 2006
Posts: 588
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

It would be interesting if there was a sort of "smuggler" type of pillage that required you to grapple within a certain time frame to get valuable cargo.
----------------------------------------
Thorkill on Malachite 2.0 Obsidian
[Jul 10, 2017 11:34:33 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

majestrate wrote: 
I don't see a need to revert the AI to an older version. The older version was predictable. I think it's good that the current version throws in random moves and doesn't seem to always use the same strategy in battle.


There is unpredictable, and plain stupid. The current version is the latter in many cases due to using the same multi-ship AI that governs things like BK blockades. I'm not going to push a hard line that the only solution is the old AI, but that would involve much less work than creating a new one.

Another point on the matter is that predictable does not necessarily equal easy. If that were the case, navigators would never have risen to the status they did. There was a huge amount of skill involved, not just waiting on a 1/3 chance the bot will actually do what you expect it to. Their chance to retreat is far too big.

EmpressTamar wrote: 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.


Scarpath wrote: 
Yeah, really does seem take forever... As a jobber, or even when running a pilly, it can get pretty frustrating. My parents do a very good job of asking for help with stuff at random times, and I can't just walk away in the middle of a battle. However, the battle can take upwards of 15-20 mins just in the bnav puzzle, meaning my parents are done by the time I can even leave the computer without worrying about getting left on a do not job list.


The is exactly the issue I'd like to see resolved. As it stands now, navigators have to make very safe moves, because the AI is not consistent. Yes, it used to turn towards your broadside if set up properly. Is it really better that it now runs instead at times? Not when we're concerned about time.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 10, 2017 11:58:23 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Snuby

Member's Avatar


Joined: Jan 30, 2006
Posts: 273
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Giving my input as bnaver

If all you engage is your own imperial spawns the AI is not random at all. If you engage non imperials prepare for some weird moves, mainly because the AI cant spawn stuff like 2 R moves in 1 turn so they'll use their triple left to get into position. If you engage something that aint your spawn, it is very likely that at some point they will run away.

A good bnaver will
-land 2-3 shots in the first 2-3 turns, which significantly decreases battle time
-take note if, for example, the AI used 2 rights during turn 1, meaning that they probably only have 1 right for turn 2, and play accordingly
-realise that whenever you change your moves, the AI also change theirs, meaning they have access to the moves they spawned meanwhile (which can be used to your advantage if you want them to move while they are sitting in a bad spot)
-setup next turn with advantageous positioning
-not move when you dont need to move
-never miss an oportunity to hit a shot
-never LSM 2 turns in a row
-and most important, realise that 90% of the time, the AI assumes you are gonna take the shortest path to grapple them with the moves you have available at the start of the turn

And if he cares about jobbers
-be ready to get shot once if it means you can shoot them 3 or 4 times
-grapple at 90% damage dealt if it means saving 6 turns of dancing around the AI to land the last shot without taking damage
-aim for 8-15 turns battles
-have a bloody dnaver aboard

Rumble against AI is easier than SF, issue is most jobbers are playing puzzle bubble and not rumble. For the most part, in any fray with a positive bnav score, the only way you lose is either single teamers (feeders) or a specific colour drought (since everyone gets the same pieces).

The only AI that needs fixing is HS/atlantis/flotilla that can easily be abused by backwalling, but I dont even feel like that should be a priority right now.
----------------------------------------
Snuby.
Current avatar by Phillite
[Jul 11, 2017 1:26:24 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
xelto

Member's Avatar


Joined: Mar 18, 2007
Posts: 5707
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles.

This is the part I would like to see fixed. If you disengage, you shouldn't be able to re-engage.

I would also like to see it changed so that if you're the one who engages in the fight, and you lose, you can re-engage. That would make auto-targeting more practical.
----------------------------------------
Gurndigarn on Emerald Ocean
"Oh, come on. You jobbed onto a ship called the Cursed Isle Raider and you expected *refined*?"
[Jul 11, 2017 4:32:51 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

penguinpaste wrote: 
majestrate wrote: 
I don't see a need to revert the AI to an older version. The older version was predictable. I think it's good that the current version throws in random moves and doesn't seem to always use the same strategy in battle.

There is unpredictable, and plain stupid. The current version is the latter in many cases due to using the same multi-ship AI that governs things like BK blockades. I'm not going to push a hard line that the only solution is the old AI, but that would involve much less work than creating a new one.

Another point on the matter is that predictable does not necessarily equal easy. If that were the case, navigators would never have risen to the status they did. There was a huge amount of skill involved, not just waiting on a 1/3 chance the bot will actually do what you expect it to. Their chance to retreat is far too big.

The only fallacy I see in the current AI is when the Sea Lords/Imps run into rock clusters/board edge multiple times in one or two turns. Like, they turned into the object, then the next turn they went F to ram it again.

Otherwise, I see no problems with the AI.

penguinpaste wrote: 
EmpressTamar wrote: 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.


Scarpath wrote: 
Yeah, really does seem take forever... As a jobber, or even when running a pilly, it can get pretty frustrating. My parents do a very good job of asking for help with stuff at random times, and I can't just walk away in the middle of a battle. However, the battle can take upwards of 15-20 mins just in the bnav puzzle, meaning my parents are done by the time I can even leave the computer without worrying about getting left on a do not job list.


The is exactly the issue I'd like to see resolved. As it stands now, navigators have to make very safe moves, because the AI is not consistent. Yes, it used to turn towards your broadside if set up properly. Is it really better that it now runs instead at times? Not when we're concerned about time.

Disagree. Having a predictable AI turns into what KH egging has become. People memorizing patterns and being considered "elite", which then results in less open pillages and more "elite" (closed) pillages.

Average/below average battle navigators should not be propped up so that we can save time. Maybe the better way to do it is that after so many turns, the NPPs disengage and cannot be re-engaged, even if they attacked. Maybe the only time you are allowed to re-engage NPPs is if you were attacked, there was a melee, and your ship lost the melee (similar to what xelto posted, but with a specific caveat; if changes are made, I think allowing 1 re-engage if you attacked an NPP and lost is a good idea)
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
----------------------------------------
[Edit 1 times, last edit by majestrate at Jul 11, 2017 11:49:15 AM]
[Jul 11, 2017 11:48:54 AM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

majestrate wrote: 
penguinpaste wrote: 
majestrate wrote: 
I don't see a need to revert the AI to an older version. The older version was predictable. I think it's good that the current version throws in random moves and doesn't seem to always use the same strategy in battle.

There is unpredictable, and plain stupid. The current version is the latter in many cases due to using the same multi-ship AI that governs things like BK blockades. I'm not going to push a hard line that the only solution is the old AI, but that would involve much less work than creating a new one.

Another point on the matter is that predictable does not necessarily equal easy. If that were the case, navigators would never have risen to the status they did. There was a huge amount of skill involved, not just waiting on a 1/3 chance the bot will actually do what you expect it to. Their chance to retreat is far too big.

The only fallacy I see in the current AI is when the Sea Lords/Imps run into rock clusters/board edge multiple times in one or two turns. Like, they turned into the object, then the next turn they went F to ram it again.

Otherwise, I see no problems with the AI.

penguinpaste wrote: 
EmpressTamar wrote: 
I don't nav pillages, but I do job on them a few times a week. My main issue with pillages is that they take too long per battle. With the way scoring works, navvers pursue perfect max-0 scores and will frequently disengage and re-engage imperfect battles. As a jobber, I'd like to see each battle take 5 minutes or less (not including fray). Perhaps the scoring could be adjusted to factor in battle speed instead of relying so heavily on accuracy. Another thing would be in the case of a dis-and-re to impact bnav stat at the dis, and the re would be unrated.


Scarpath wrote: 
Yeah, really does seem take forever... As a jobber, or even when running a pilly, it can get pretty frustrating. My parents do a very good job of asking for help with stuff at random times, and I can't just walk away in the middle of a battle. However, the battle can take upwards of 15-20 mins just in the bnav puzzle, meaning my parents are done by the time I can even leave the computer without worrying about getting left on a do not job list.


The is exactly the issue I'd like to see resolved. As it stands now, navigators have to make very safe moves, because the AI is not consistent. Yes, it used to turn towards your broadside if set up properly. Is it really better that it now runs instead at times? Not when we're concerned about time.

Disagree. Having a predictable AI turns into what KH egging has become. People memorizing patterns and being considered "elite", which then results in less open pillages and more "elite" (closed) pillages.

Average/below average battle navigators should not be propped up so that we can save time. Maybe the better way to do it is that after so many turns, the NPPs disengage and cannot be re-engaged, even if they attacked. Maybe the only time you are allowed to re-engage NPPs is if you were attacked, there was a melee, and your ship lost the melee (similar to what xelto posted, but with a specific caveat; if changes are made, I think allowing 1 re-engage if you attacked an NPP and lost is a good idea)


I wish I could do a side-by-side comparison of old and new, as it is hard to adequately explain how poor the current AI is for 1v1 fights. This isn't about propping anyone up, it is about allowing people to master the puzzle. The ridiculous chance to make a retreat move consistently sets the battle back by several turns, requiring setup moves to be done again.

I wonder if, assuming the old code exists, Ice could be used as a testing facility. I feel like dulled memories and second hand recollections are causing issues here.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 12, 2017 9:43:02 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

penguinpaste wrote: 
I wish I could do a side-by-side comparison of old and new, as it is hard to adequately explain how poor the current AI is for 1v1 fights. This isn't about propping anyone up, it is about allowing people to master the puzzle. The ridiculous chance to make a retreat move consistently sets the battle back by several turns, requiring setup moves to be done again.

I wonder if, assuming the old code exists, Ice could be used as a testing facility. I feel like dulled memories and second hand recollections are causing issues here.

Not all random moves are retreat moves. The AI seems to calculate probability of the player moves and then respond to those. But when it calculates probability, it doesn't always go with what it "believes" to be the perfect counter-moves/attack moves. Sometimes it will go with the 3rd best calculated option and input moves according to what that 3rd best option is.

I know your intent is not to prop up average/below average battle navigators, but none-the-less, IMO, that's what would happen if we reverted to an older AI where players could simply memorize what patterns the AI uses and input responding moves.

 
If all you engage is your own imperial spawns the AI is not random at all. If you engage non imperials prepare for some weird moves, mainly because the AI cant spawn stuff like 2 R moves in 1 turn so they'll use their triple left to get into position. If you engage something that aint your spawn, it is very likely that at some point they will run away.

Imperial spawns that engage me end up performing random moves during the battle. And I don't believe there's any real difference in the AI if you engage the NPP ship (your spawn) or if the NPP ship engages you. Yes, if you engage someone else's spawn instead of your own, the spawn will play along for a little bit, but eventually (before the 10 turn mark, I believe) it will run.
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
[Jul 12, 2017 11:00:39 AM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
Factor93

Member's Avatar


Joined: Oct 22, 2008
Posts: 74
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

It seems unlike most of you I don't go for max - 0 I actually go for for 0 - 0 with a grapple the first turn. If you've ever been on one of my pillies you'd know i grapple within the first one to three turns at least 50% of the time. I don't even do last second moves now. (Refer to Snuby's earlier post) With this technique I'm at about master - ren most of the time and obviously I don't consider myself a good bnavver. The stat and the AI is both broken in my opinion. How to fix this? I'm really not sure and I have no good ideas.

Maybe for stats it would be a good idea to take the time aspect of the battle away but obviously this might be bad for jobbers so feel free to criticise this.

Just my (fairly terribly informed on this subject) opinion.

Edit: Just mentioning if you want quick battles job for me. :p
----------------------------------------
Factors on every ocean - Active on Obsidian

Avatar by Dragonjayne <3
----------------------------------------
[Edit 1 times, last edit by Factor93 at Jul 12, 2017 11:12:55 AM]
[Jul 12, 2017 11:09:47 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Discflicker

Member's Avatar


Joined: Jul 4, 2006
Posts: 430
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

I think we should keep this thread's discussion to just the AI - the bnav rating system could get its own thread. Bnavving well (getting a max-0 or max-low in as few turns as possible) is not that well correlated with bnav stat IMO.

I believe that a full reversion would be bad (it was too easy and too predictable before the big bnav update). However, some of the moves these days are just incomprehensible and stupid. Like these two turns (which were at the very start of the battle).

I'd like a few more "hard rules" to make the battles a little more predictable. For example, in this clip - I believe that it used to be (at least on Meridian for the last several years) that the AI would always turn and throw the grapple on the outside. But for whatever reason (code change? difficulty curve?) they threw it on the inside here.

edit: Also, in case it's not clear, I agree with the OP's premise that the current AI makes battles too long by being too unpredictable. I don't think we need it to be fully predictable, but max-2 within 12 turns should be regularly achievable. I haven't done much hard route bnav on Obsidian, but probably logged 2000+ hours of it on Viridian/Meridian (enough to get illustrious and #1 bnav).
----------------------------------------
Not convinced I'll ever truly quit Puzzle Pirates forever if it sticks around.
----------------------------------------
[Edit 1 times, last edit by Discflicker at Jul 12, 2017 1:22:33 PM]
[Jul 12, 2017 1:15:57 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Discflicker wrote: 
I think we should keep this thread's discussion to just the AI - the bnav rating system could get its own thread. Bnavving well (getting a max-0 or max-low in as few turns as possible) is not that well correlated with bnav stat IMO.

I believe that a full reversion would be bad (it was too easy and too predictable before the big bnav update). However, some of the moves these days are just incomprehensible and stupid. Like these two turns (which were at the very start of the battle).

I'd like a few more "hard rules" to make the battles a little more predictable. For example, in this clip - I believe that it used to be (at least on Meridian for the last several years) that the AI would always turn and throw the grapple on the outside. But for whatever reason (code change? difficulty curve?) they threw it on the inside here.


Good clip, a perfect example of what is making things take so long. Ships randomly doing this in the middle of battle is incredibly frustrating. I do seem to recall that the inside move generally avoided grapples, but it wasn't 100% either.

I think that the issue here is either (or both) randomness is set too high, or the code that governs player move prediction is flawed.

Discflicker wrote: 
edit: Also, in case it's not clear, I agree with the OP's premise that the current AI makes battles too long by being too unpredictable. I don't think we need it to be fully predictable, but max-2 within 12 turns should be regularly achievable. I haven't done much hard route bnav on Obsidian, but probably logged 2000+ hours of it on Viridian/Meridian (enough to get illustrious and #1 bnav).


Sloop pillages used to be max-0 in <10 turns on difficult routes, unless the board was bad or the bots weren't cooperating. Full brig vs WF/GF took a bit longer, but it wasn't a massive difference.

I'd be perfectly happy with changes instead of reversion, however the former is likely to take a lot more time (assuming the old code is still accessible). At the very least, it should be split so changes can be made to brigand AI without altering that of other encounters. Something needs to change though.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 12, 2017 2:00:25 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Discflicker

Member's Avatar


Joined: Jul 4, 2006
Posts: 430
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Agree on all points with GreatBob, except I think that the formerly achievable max-0 in <10 turns (I think I remember some 4 turn battles) felt a little cheap. I think (and it's been literally about a decade) that before the change I was averaging 6-8 turns on a sloop going max-2 at worst against cutters.


Speculation on the code that governs player move prediction below:

It's possible that the player move prediction engine isn't necessarily inherently flawed and the Obsidian playerbase being more hardcore is ruining things (similar to how it's ruining things in other puzzles since the scoring curve is super harsh in duty puzzling and somehow I'm getting GM distilling without ever seeing an excellent). Then this would fix itself if/when we get an influx of greenies from a steam release.

Let's take the inside vs outside grapple for example. What if it's taking ocean-wide stats around how often inside vs outside works? Let's say that at the low levels (let's say you're fighting low sailors on the sailors-imperials scale), it throws it where it's less likely to succeed (let's say on Meridian, it's inside but on Obisidian, it's outside) to be easy. And the reason for this differing between oceans is that intuitively, players tend to turn to the outside to get away. However, Obsidian has all these elite players that have been playing since forever, so we all know to turn inside instead.

As it goes up in difficulty to high imperials, at some point it flips to take the move that's more likely to succeed (Meridian: outside, Obsidian: inside).
----------------------------------------
Not convinced I'll ever truly quit Puzzle Pirates forever if it sticks around.
[Jul 12, 2017 2:51:06 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
BobJanova

Member's Avatar


Joined: Oct 16, 2007
Posts: 5008
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

As I understand it the old AI was completely predictable and you just had to learn how the bots moved to be able to M-0 every time without doing anything more than a memory game. That's not enjoyable or fun.

The current AI is fine. The problem is more with the bnav puzzle itself, and we'll find that out a lot more with the increased PVP on Obsidian. If your opponent wants to run away, or trade shots, there's very little you can do to avoid that.

One big change would be for carping to reduce fray damage, as well as ship damage. Against hard enemies you need to finish M-0 or M-1 to win the fray, and that encourages negative navving because shots you take are never recoverable (unlike blockade/flotilla/Atlantis nav, where people will take risks to land shots because you can get fixed afterwards).
----------------------------------------
Bobjanova on Viridian and Malachite
Shops and stalls with fair and profitable wages for all: Jubilee, Napi, Chelydra
Stripped/Barely Dressed (Malachite)
Phantasm/Reign of Chaos (Viridian)
[Jul 12, 2017 4:06:54 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Snuby

Member's Avatar


Joined: Jan 30, 2006
Posts: 273
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Refering to my other post
Snuby wrote: 
and most important, realise that 90% of the time, the AI assumes you are gonna take the shortest path to grapple them with the moves you have available at the start of the turn


Discflicker wrote: 
However, some of the moves these days are just incomprehensible and stupid. Like these two turns (which were at the very start of the battle).


On your first clip, first turn, they assume you will go RL and try to grapple them (shortest path to do so), so they play accordingly. Because they are merchants, it means they LSM every turn (which allows them to use all the moves they spawned during the turn) and do the weird triple Left.
On 2nd turn, the shortest path to them would be a diagonal (LRL) and once again they move accordingly. Since they used 3 lefts last turn, the auto generated moves will be lefts, and the same thing happens again. It looks weird but it makes sense.

Discflicker wrote: 
For example, in this clip - I believe that it used to be (at least on Meridian for the last several years) that the AI would always turn and throw the grapple on the outside. But for whatever reason (code change? difficulty curve?) they threw it on the inside here.


On your second clip, yet again, the AI assumes you will take the shortest path to grapple (as in Right and grapple) so they move and throw their grapple accordingly.

Also as an aside, thanks for streaming the game, I watch the VODs to practise bnav and learn new things. Unfortunately I think we are in completely different time zones, so I can never interact live with the stream.
----------------------------------------
Snuby.
Current avatar by Phillite
[Jul 13, 2017 2:01:28 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
I think we should keep this thread's discussion to just the AI - the bnav rating system could get its own thread. Bnavving well (getting a max-0 or max-low in as few turns as possible) is not that well correlated with bnav stat IMO.

I believe that a full reversion would be bad (it was too easy and too predictable before the big bnav update). However, some of the moves these days are just incomprehensible and stupid. Like these two turns (which were at the very start of the battle).

Wait, what? How in the world is that a perfect example of anything.

First turn
Clearly the AI moved according to it thinking you were going into the pool and exiting with a left. Then it went 2 lefts, just in case you went "right, left" out of the pool. Ignoring the first shot it took and that last grapple, the moves were sound.

Second turn
It moved to try and set you up for a broadside. I see nothing wrong with what was done.

If anyone thinks that the example clip provided is "too random and causes long battles", then you need more practice in your PvE 1v1 battle navigation.

Edit 1: Missed a carriage return or 2
Edit 2: Should have read the full thread before replying, I agree with Snuby's reply
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
----------------------------------------
[Edit 2 times, last edit by majestrate at Jul 13, 2017 4:38:18 AM]
[Jul 13, 2017 4:35:27 AM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
Discflicker

Member's Avatar


Joined: Jul 4, 2006
Posts: 430
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Sorry, I should have been clearer because there's only 1 move in that first turn that I think is dumb.

I agree that the LSL on the first three moves of the first turn are fine by the merch - I should have clarified that the L after that is what I was really confused by. I don't think there's a possibility where that grapple can hit anything and there's not a lot of reason for their ship to steer in that direction AFAIK.

On the second turn, if they're assuming that i'm taking LRL they're believing that I'm only using 3 moves and not ending broadside on them. Or they're assuming I'm going LSRS for some reason, both of which would be really weird/bad human navigator moves. They took a LSLL, but if they took LLSR instead, they'd have a higher chance of grapple IMO since a human going LR and then sitting and shooting would be much more reasonable than LRL or LSRS.
----------------------------------------
Not convinced I'll ever truly quit Puzzle Pirates forever if it sticks around.
[Jul 13, 2017 11:26:24 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
GreatBob

Member's Avatar


Joined: Sep 3, 2005
Posts: 1465
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

BobJanova wrote: 
As I understand it the old AI was completely predictable and you just had to learn how the bots moved to be able to M-0 every time without doing anything more than a memory game. That's not enjoyable or fun.

The current AI is fine. The problem is more with the bnav puzzle itself, and we'll find that out a lot more with the increased PVP on Obsidian. If your opponent wants to run away, or trade shots, there's very little you can do to avoid that.

One big change would be for carping to reduce fray damage, as well as ship damage. Against hard enemies you need to finish M-0 or M-1 to win the fray, and that encourages negative navving because shots you take are never recoverable (unlike blockade/flotilla/Atlantis nav, where people will take risks to land shots because you can get fixed afterwards).


Completely predictable is a gross overstatement. It adhered to a set of rules that, with proper setup, provided predictable results. There was a massive amount of skill required to do this, as evidenced by the status successful navigators had.

I understand why people are struggling to remember properly, it was a long time ago. Also, the remembering of ease could be due to jobbing on ships with someone who tended to make it look easy. It is also difficult for those who have only known the current AI to accept that it is inadequate, as they have accepted it as the way the bots work.

Again, I wonder if Ice could be set to use the old AI or be tested with another, different one. I'd be happy to put hours into testing on there if it meant having a better system.
----------------------------------------
Penguinpaste, SO of Polaris, Obsidian. Dark side.


Avatar by Phaerie.
[Jul 13, 2017 11:52:02 AM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
marren362

Member's Avatar


Joined: Mar 11, 2006
Posts: 60
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

Having Bnavved during the old system I was one of the ones who was very happy that some change occurred. If anyone can recall back then things hit a point where the rank list only had Legendary. The top ten or however large it was back then was also down to who had the most experience as every Bnavver was capping their score as it was unlikely to have a bad fight once you learned the system. It was merely going through the motions.

I'd agree with you that a change to Bnav would be nice as it would be something new to learn. One of the aspects I enjoyed when they did finally change was a lot of players fell off, they lost rank because they were used to gaming the system and didn't like that it suddenly needed more patience or thought. I'll agree that the opponents moves can at times be ludicrous, having encountered the bots bumping themselves a high amount recently. But even so it's mostly to the detriment of the bot rather than yourself.

How long do you actually feel a battle should be taking? Considering the increased payouts a 10-12 minute battle on average doesn't seem that extreme to me. If it was always only 5 mins the pay for effort would be quite substantial.

Also not every battle needs to be Max-0. Of course it increases the likelihood of a win but the reality is that often you can win when it's Max-Max or Max-4 etc at least with sword fighting. Rumble I find is more difficult but I'm a pretty terrible rumbler so in theory I shouldn't be fighting the most difficult bots the game can produce. With that in mind I would say if people feel they can't win unless it's absolutely always Max-0 then they shouldn't be on the route they're pillaging. I'm not a massive jobber but in my experiences not to many repeatedly disengage and re-engage I know of some who abuse the mechanic but as has been said further up, don't job with them so they learn to adapt for jobbers.

A massive amount of skill? They were taught a system to replicate time and time again and anyone who had it explained would have been able to replicate the results after time. As was shown with the fact that the lists wouldn't move as so many players capped out and the game couldn't even give Ultimate anymore.


Interarch provides about 30k+ average per battle, with 30-50k TH, Forages, KB sometimes added in. If you're earning that on average 12 min per battle and you drop that down to 5 minutes. You're going to drastically alter the economy and make it unrealistic to do the other activities as pillaging would be everyone's sole focus.
----------------------------------------
Tevez
Captain of The Midnight Society
----------------------------------------
[Edit 1 times, last edit by marren362 at Jul 13, 2017 2:31:56 PM]
[Jul 13, 2017 2:31:15 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
Sorry, I should have been clearer because there's only 1 move in that first turn that I think is dumb.

I agree that the LSL on the first three moves of the first turn are fine by the merch - I should have clarified that the L after that is what I was really confused by. I don't think there's a possibility where that grapple can hit anything and there's not a lot of reason for their ship to steer in that direction AFAIK.

No reason for the grapple, but also no reason for the shot on the first move of the first turn.

As I said, that last left was an attempt to try and setup a broadside on the next turn, all based on the player ship going into the pool.
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
[Jul 13, 2017 5:36:44 PM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
majestrate

Member's Avatar


Joined: Feb 18, 2005
Posts: 3958
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

 
BobJanova wrote: 
As I understand it the old AI was completely predictable and you just had to learn how the bots moved to be able to M-0 every time without doing anything more than a memory game. That's not enjoyable or fun.

The current AI is fine. The problem is more with the bnav puzzle itself, and we'll find that out a lot more with the increased PVP on Obsidian. If your opponent wants to run away, or trade shots, there's very little you can do to avoid that.

One big change would be for carping to reduce fray damage, as well as ship damage. Against hard enemies you need to finish M-0 or M-1 to win the fray, and that encourages negative navving because shots you take are never recoverable (unlike blockade/flotilla/Atlantis nav, where people will take risks to land shots because you can get fixed afterwards).


Completely predictable is a gross overstatement. It adhered to a set of rules that, with proper setup, provided predictable results. There was a massive amount of skill required to do this, as evidenced by the status successful navigators had.

I understand why people are struggling to remember properly, it was a long time ago. Also, the remembering of ease could be due to jobbing on ships with someone who tended to make it look easy. It is also difficult for those who have only known the current AI to accept that it is inadequate, as they have accepted it as the way the bots work.

I thought that the AI change was done because of the "predictable results". You get the NPPs into certain situations and they would always perform the exact same thing. So instead of long battles where everyone is trying to cope with random moves (now), back then you had most players doing their best to get setup into the right position so that the bots would behave predictably (or run them out of moves). The better navigators would have that setup more quickly, but the result was the same. Average navigators going max-<low> against NPPs just because they caused the NPPs to be DitW by taking 15-20 turns instead of 5-10 turns.
----------------------------------------
#TeamEvil
Marto wrote: 
We can't rely on majestrate he yells at people


Avatar by the gracious and wonderful Phaerie <3
[Jul 13, 2017 5:42:33 PM] Show Printable Version of Post        Send Private Message    first_majestrate    majestrait [Link]  Go to top 
estonianguy

Member's Avatar


Joined: Feb 14, 2006
Posts: 93
Status: Offline
Re: Revamp Pillaging AI Reply to this Post
Reply with Quote

the AI is fine, the bigger problem is that frays are not winnable on equal damage because the "imperials" fray skills range from Able to hit bingo every breaker.
Also spawning ships that are 3x bigger than you, Junks and baglahs should not spawn for a sloop pilly
----------------------------------------
Xzomg
[Jul 13, 2017 10:52:43 PM] Show Printable Version of Post        Send Private Message [Link]  Go to top 
Posts: 33   Pages: 2   [ First Page | 1 2 | Next Page | Last Page]
[Show Printable Version of Thread] [Post new Thread]

Puzzle Pirates™ © 2001-2020 Grey Havens, LLC All Rights Reserved.   Terms · Privacy · Affiliates