| 1 |
[q]This line is something I don't understand. I have played this game for quite a bit, but I could never tell you what a unit on free-roam is going to do, or how a unit on fight-move will behave. Chase, skirm, low-priority target classes are changing without the player knowing about it, and even if they didn't it is very hard to follow the unit AI's decision making without looking at the code. I know Scorchers should automatically dive against some units and kite some others, but this is something I have been unable to predict. Then they also retreat from buildings when the building is about to explode and other special behavior. At that point I feel the scorcher dive/kite decision might as well be dictated by a neural net.[/q]
|
1 |
[q]This line is something I don't understand. I have played this game for quite a bit, but I could never tell you what a unit on free-roam is going to do, or how a unit on fight-move will behave. Chase, skirm, low-priority target classes are changing without the player knowing about it, and even if they didn't it is very hard to follow the unit AI's decision making without looking at the code. I know Scorchers should automatically dive against some units and kite some others, but this is something I have been unable to predict. Then they also retreat from buildings when the building is about to explode and other special behavior. At that point I feel the scorcher dive/kite decision might as well be dictated by a neural net.[/q]
|
| 2 |
I am not going to spend the time to attempt to communicate my full criteria. I tend to make up/discover answers on the fly based on my intuition of what makes ZK good. Neither did I claim that this "line" is a good guide for all cases or that all cases are perfectly implemented according to it. Some of what you're saying barely falls without the purview of the fuzzy idea I was attempting to express (units have to fire at something, for example) and other bits are ongoing problems.
|
2 |
I am not going to spend the time to attempt to communicate my full criteria. I tend to make up/discover answers on the fly based on my intuition of what makes ZK good. Neither did I claim that this "line" is a good guide for all cases or that all cases are perfectly implemented according to it. Some of what you're saying barely falls without the purview of the fuzzy idea I was attempting to express (units have to fire at something, for example) and other bits are ongoing problems.
|
| 3 |
\n
|
3 |
\n
|
| 4 |
The important thing is that we've reached this:
|
4 |
The important thing is that we've reached this:
|
| 5 |
[q]As a somewhat related, but non topic furthering point: I like the current level of automation at work. It's in the right sort of ballpark to allow the AI to handle things you want it to handle, whilst allowing the player to absolutely be the guiding force behind that AI. It's shepard. I feel as though more assistance than we have now just begins to erode the need for skill, experience and APM. [/q]
|
5 |
[q]As a somewhat related, but non topic furthering point: I like the current level of automation at work. It's in the right sort of ballpark to allow the AI to handle things you want it to handle, whilst allowing the player to absolutely be the guiding force behind that AI. It's shepard. I feel as though more assistance than we have now just begins to erode the need for skill, experience and APM. [/q]
|
| 6 |
\n
|
6 |
\n
|
| 7 |
[q]Could you elaborate on what makes a ZK player human, 'cyborg' or a bot? It really isn't obvious to me. Would a human player with today's interface still be a human player if he played 10 year old ZK with the modern interface? [/q]
|
7 |
[q]Could you elaborate on what makes a ZK player human, 'cyborg' or a bot? It really isn't obvious to me. Would a human player with today's interface still be a human player if he played 10 year old ZK with the modern interface? [/q]
|
| 8 |
The cyborg concept is that players are sets of commands sent the server and the server should not care in the slightest how those commands were computed. Players are merger of their abilities as a person as well as all the code that is enhancing them. Nobody "is" a cyborg, so such specific questions are incoherent.
|
8 |
The cyborg concept is that players are sets of commands sent the server and the server should not care in the slightest how those commands were computed. Players are merger of their abilities as a person as well as all the code that is enhancing them. Nobody "is" a cyborg, so such specific questions are incoherent.
|
| 9 |
\n
|
9 |
\n
|
| 10 |
[q]posted a screenshot of him using a whole bunch of widgets: https://zero-k.info/Forum/Post/219310#219310[/q]
|
10 |
[q]posted a screenshot of him using a whole bunch of widgets: https://zero-k.info/Forum/Post/219310#219310[/q]
|
| 11 |
That post is a complete mess and you should not expect anyone to look through it to find the thing you're referencing.
|
11 |
That post is a complete mess and you should not expect anyone to look through it to find the thing you're referencing.
|
| 12 |
\n
|
12 |
\n
|
| 13 |
[q]But I'm not convinced that's the real issue (I expect you'd probably accept PRs for most widgets provided they are disabled by default). [/q]
|
13 |
[q]But I'm not convinced that's the real issue (I expect you'd probably accept PRs for most widgets provided they are disabled by default). [/q]
|
| 14 |
As I said, a widget disabled by default barely existed for the purpose of this topic. Not everyone knows what the word 'widget' means to get the subtlety of what you were alluding to.
|
14 |
As I said, a widget disabled by default barely existed for the purpose of this topic. Not everyone knows what the word 'widget' means to get the subtlety of what you were alluding to.
|
| 15 |
\n
|
15 |
\n
|
| 16 |
[q]In Starcraft, AIs without APM limit play a completely different game than humans. A human would never be able to replicate what such an AI does. My goal for ZK would be that a human can always (at least theoretically) replicate what the AI does and use it for himself. This is not possible if some micro automation is inaccessible for the player.
|
16 |
[q]In Starcraft, AIs without APM limit play a completely different game than humans. A human would never be able to replicate what such an AI does. My goal for ZK would be that a human can always (at least theoretically) replicate what the AI does and use it for himself. This is not possible if some micro automation is inaccessible for the player.
|
| 17 |
\n
|
17 |
\n
|
| 18 |
This might seem like an impossible task, but ZK is fairly close to it currently. Depending on how we decide, this may in the future lead to AIs cheating by definition and void the "if you can't beat the AI, learn from the AI" slogan. [/q]Sure, and if we are fairly close currently why put up with the extra issues? See "scraping the bottom of the barrel".
|
18 |
This might seem like an impossible task, but ZK is fairly close to it currently. Depending on how we decide, this may in the future lead to AIs cheating by definition and void the "if you can't beat the AI, learn from the AI" slogan. [/q]Sure, and if we are fairly close currently why put up with the extra issues? See "scraping the bottom of the barrel".
|
| 19 |
\n
|
19 |
\n
|
| 20 |
@malric I think people mostly make widgets to make widgets. They want to play the coder sandbox or test their cyborg abilities. Winning can be a way to test a widget but I don't think people first set out to win at all costs and then decide to make widgets to do so. Other people obviously only complain about the widgets that are effective because they don't want to feel like they have lost to an exploited mechanic. If someone makes a widget to shoot themselves in the foot then everyone else is fine with them using it.
|
20 |
@malric I think people mostly make widgets to make widgets. They want to play the coder sandbox or test their cyborg abilities. Winning can be a way to test a widget but I don't think people first set out to win at all costs and then decide to make widgets to do so. Other people obviously only complain about the widgets that are effective because they don't want to feel like they have lost to an exploited mechanic. If someone makes a widget to shoot themselves in the foot then everyone else is fine with them using it.
|
| 21 |
\n
|
21 |
\n
|
| 22 |
[q]There are still lots of low hanging fruit. Just look at any unit that requires lots of babysitting which should be fixed.
|
22 |
[q]There are still lots of low hanging fruit. Just look at any unit that requires lots of babysitting which should be fixed.
|
| 23 |
For instance, Lances will fire at fleas rather than waiting for the enemy Lance that's just about to enter range, Recluse will wander into Stingers and a group of Scythes will just sit there getting murdered by a Glaive if discovered (though the AvoidanceAI I made helps here).[/q]
|
23 |
For instance, Lances will fire at fleas rather than waiting for the enemy Lance that's just about to enter range, Recluse will wander into Stingers and a group of Scythes will just sit there getting murdered by a Glaive if discovered (though the AvoidanceAI I made helps here).[/q]
|
| 24 |
Such Lance AI sounds fine. See the years-old ticket on it. @Dregs I think it should be obvious that @dyth68 is suggesting some sort of state toggle or more advanced configuration because not providing any sort of UI would make it a terrible idea. It goes back to the question of whether the behaviour resulting from the command is understandable.
|
24 |
Such Lance AI sounds fine. See the years-old ticket on it. @Dregs I think it should be obvious that @dyth68 is suggesting some sort of state toggle or more advanced configuration because not providing any sort of UI would make it a terrible idea. It goes back to the question of whether the behaviour resulting from the command is understandable.
|
| 25 |
\n
|
25 |
\n
|
|
|
26 |
@PRO_rANDY if someone wants to have their Lances not shoot at anything that costs less than 200, then I am fine with letting them do that without watching their Lances to make sure they shoot at things that cost more. Someone spamming Fleas at Lance should not be rewarded simply because Lance is finicky to control.
|
|
|
27 |
\n
|
| 26 |
What I will not do is accept any sort of Lance widget without anyone first making a serious attempt at extending the target priority and overkill prevention gadgets. This is another way that widgets don't "help development" that I failed to mention. There is actually very little low-hanging fruit, and most things that look like low-hanging fruit are bound to be expensive and suffer from latency issues if implemented as a widget. The lowest-hanging fruit for unit AI these days needs to be done as a gadget (a more powerful, game-only, type of widget). Gadgets have powerful abilities that make such unit AI much cheaper and requires them to be written in completely different ways. This is another reason that I view writing widget unit AI as a bit of a personally-indulgent waste of time that doesn't lead to any game improvement.
|
28 |
What I will not do is accept any sort of Lance widget without anyone first making a serious attempt at extending the target priority and overkill prevention gadgets. This is another way that widgets don't "help development" that I failed to mention. There is actually very little low-hanging fruit, and most things that look like low-hanging fruit are bound to be expensive and suffer from latency issues if implemented as a widget. The lowest-hanging fruit for unit AI these days needs to be done as a gadget (a more powerful, game-only, type of widget). Gadgets have powerful abilities that make such unit AI much cheaper and requires them to be written in completely different ways. This is another reason that I view writing widget unit AI as a bit of a personally-indulgent waste of time that doesn't lead to any game improvement.
|
| 27 |
\n
|
29 |
\n
|
| 28 |
Scythe is trickier but think about it in terms of decisions. I think it would be fine if Attack-Move cloaked units set to Hold Fire automatically avoided nearby enemy units. The player has chosen where to send the Scythe and that they want to avoid decloaking while they are trying to get there. Implementing this decision is mostly a matter of moving away from enemy units whenever they are near. That said, there can be some important decisions about pathfinding, but just as with the autoskirm AI, you have to accept simple predictable-ish behaviour when you use Attack Move.
|
30 |
Scythe is trickier but think about it in terms of decisions. I think it would be fine if Attack-Move cloaked units set to Hold Fire automatically avoided nearby enemy units. The player has chosen where to send the Scythe and that they want to avoid decloaking while they are trying to get there. Implementing this decision is mostly a matter of moving away from enemy units whenever they are near. That said, there can be some important decisions about pathfinding, but just as with the autoskirm AI, you have to accept simple predictable-ish behaviour when you use Attack Move.
|
Widgets vs Cheating