Automation tools

I seem to have a different opinion than a lot of others here. I don't like automation tools. If one player has a useful widget that the other player does not have, that seems unfair to me. Furthermore, if a game has "redundant" tasks, they are there for the player to do. If players just use a widget to handle a redundant task, then why is that task even in the game?

Ultimately, I like the raw player vs. player aspect of RTS games, not player+widgets vs player+widgets. Granted, this might just be an appeal to a different kind of game than Zero-K wants to be. Look at Starcraft. There are redundant tasks that must be done, such as building supply structures and building units without a queue. These tasks are not automated, which introduces a new point of skill to the game. One player may be awesome at micro, while another player never gets supply blocked, and another player has impeccable macro. With everything automated, these possible areas to gain an edge on your opponent disappear. Of course, this is probably what Zero-K wants. Instead of worrying about supply and constant worker production, players have to focus on not excessing metal, not stalling energy, and reclaiming. In my opinion, widgets that handle these tasks for the player remove possible ways for a player to outplay their opponent, and I do not like it. Not one bit.

When someone thinks of an RTS game, there are certain traits that come to mind when wanting to improve. These include APM, scouting, expanding, micro, game sense, etc. Zero-K is the only game I can think of where coding is a valid route to improvement (not actually true, see: Robocode). This is not a bad thing in itself, but it is a very foreign concept to most gamers, and it will turn many players off.

I would argue that automation, and thereby reducing the number of points of skill, results in less choices along with less clicking. Each click is a choice, and the player must handle the hidden resource of time. With less automation, there are more choices to the player regarding where to spend their time. With automation, the player has fewer things to worry about, fewer things that demand their attention, and therefore, fewer "choices" to make with regard to their time. The remaining choices are relatively high-level and hopefully more interesting, and the player can focus on these choices more thanks to automation.

quote: Each click is a choice

"Having to order a drone every n seconds or you lose" is not a choice, it's a technical skill barrier.

It's just a matter of finding the right balance between having an AI playing the game for you and having to play qwop to keep every single unit moving.
I think the nice thing about widgets is that they allow players to help with this part of development (unit ai, spotter widgets, pre-game queue...) I think this helps to find that balance. It is allways easier to wrige a widget to automate boring and meaningless tasks than to write a widget that makes meaningful decisions for you or that is better at the fun unit micro.
I don't think that widgets are currently used to get a significant advantage in competitive play, and I don't think this will ever be the case. But ideally all such widgets should be made publicly available or even included in the game.

till you have your AI beat the other players AI up.

The thing I find with Zero-K is that while automation is a major thing, the game is designed around it. As a result, the clicks you do make [color=grey](and there can still be a lot)[/color] now are all doing something meaningful, rather than just rote [color=grey](such as moving workers to minerals or building supply buildings/units)[/color]. I find this makes both combat and base building more stressful, as line of sight, physical projectiles [color=grey](thus dodging and either the current lack of velocity inheritance (thus retreat micro) or a hypothetical game with projectile velocity inheritance (which would have its own micro tricks))[/color], and turret rotation speeds make a huge difference in the moment to moment combat; while line of sight for defenses and distances for units make a lot of difference for defensible base building.

It's very moment-to-moment, though, not as much that can be laid out as you would in a build order or unit micro guide. Still, because those interactions matter [color=grey](and the unit lethality and unit build times are both high enough until super late game for it to matter. Bearing in mind late game StarCraft (at least BW) involves entire armies getting rebuilt in between fights this pacing isn't that different)[/color], it becomes a hugely important skill, but because everything is highly contextual and situational, it's really hard to do well in a practiced way.

quote: 1: AI in RTS games still cannot (yet) compete with human players, unless the game makes it unduly hard for the humans to play with artificial APM drains like larva injection

Really? That's surprising actually. There're plenty examples of AI micro making perfect attrition vs. pro micro. I imagine perfect or near perfect logistics would also be fairly easy. Scouting might be difficult for an AI, as well as decision making, but I imagine that with a dedicated coder you could program even better inference than is common in pro-players. Stuff like '10 marines visible at 2:30 mark -> opponent has at least 3 barracks -> opponent cannot possibly be expanding', the list of such possible statements being very large and the application of this logic being done every few seconds or whatever. AI have the potential to have 'passive scouting' at a higher than humanly possible level.

AI have a lot of advantages and it's sort of confusing that there isn't a single one made in a single context that compares to pro players? Explain!

anyway.

we had an AI competition twice in the last year at work in a little different context.

a (simple) game board with defined rules, running in real time, defined timeframes for cpu cycles to calculate decisions and execute orders (if capped it will be discarded), an exposed API to program against in all major coding languages. i did not participate because of the lack of time, but it was really fun to watch and the replays are recorded. the graphical representation was in the browser.

it would really be fun to make a zero-k AI competition on a regular tournament basis. players code their AIs and let them clash each other,see what happens and iterate code in the next cycle.

these AIs could be released in the wild, to be triumphated by noobs in public games!

This discussion has conflated two similar topics:
1. Players coding their own UI.
2. Automation.
There is also the extra topic of RTS AI.


The first topic is a messy one. Coding your own UI has the potential to become an issue but it does not seem to be an issue at the moment. The "pure design philosophy" of ZK does not prevent local widgets but in practice issues could arise. If it became an issue we could block local widgets (or make a modoption). This would be circumventable but with some difficulty. Those people would know what they are doing so we could happily label them as cheaters. Local widgets are currently allowed for a few reasons:

• ZK is designed to not be trivialized by automation.
• Players effectively do developer work by writing useful widgets.
• Players tend to share useful widgets that they have written.
• Nobody seems to have exploited the current trust based system.
• Few people have complained and not loudly.

One fairly recent example is the Newton launcher widget. Someone made a widget which semi-automated Newton launchers, increased fire rate and made them reliable. They shared the widget and so, with slight usability improvements, I included it in ZK. Newtons now have a command which sets a fire-zone in which they will launch any unit which enters.

I understand the main argument against local widgets; that using different UIs introduces unfairness. For example the unit AI in ZK could be a lot better so there is potential for someone to develop a widget which gives them a significant advantage over other players. This is a good thing if they share their widget (because it would be included in ZK) but unfair if they keep it to themselves. It would still be annoying to lose against the widget while it is still under development (not yet released) so perhaps there could be blocked widget modes. More competitive matches could have blocked widgets while other matches allowed people to test and develop their widgets.

My current stance towards powerful user widgets is to include them. I am not worried that such a widget would break gameplay because I have not seen any sufficiently powerful user widget. Automation widgets tend to highlight areas of ZK which should be automated, they don't tend to remove choices and if they did I would consider it a flaw in ZK rather than a widget to be blocked.

I am not so sure that ZK and SCII are so different on the interface customization front. There seems to be a lot in SCII hotkeys and tricks to improve your play. Furthermore I don't see drilling your UI interaction in SCII as significantly different to customizing and learning to use the default UI of ZK. Watch this video explaining how to Larvae Inject as if you had a widget to do it:


This brings me to the second topic; automation itself. Automation, for me, is there to decrease a tasks Decision:Clicks ratio in order to free players from spending most of their time implementing simple decisions. This gives them time to consider a wider range of (hopefully more interesting) decisions. ZK contains so many decisions that the difficulty of making good decisions rapidly should keep the game from becoming trivial. In theory the aim is for ZK to be able to be carried, as a worthwhile game, by the decisions instead of the clicking.

The expected rate of decision making can be very high which I why I do not think combat could be automated (locally). For example say you have a group of Rockos dancing around an enemy army. At any point you could decide to dive the Rockos in and go for a commander kill. Whether you do so depends on factors outside the army; how much their commander is worth at the time, whether your relative armies (across the whole map) mean you can afford to lose the Rockos, what units they may have just outside LOS etc... all weighed up against how likely the dive is to succeed.

Decisions are required even faster during raiding. Should you lose a few raiders to secure a constructor kill? Where are they likely to have placed their raiders? Where are there defenses? Is it worth losing a raider for some extra scouting? Where are they likely to have a Tick? This leads into another aspect; the mind games of reading your opponent. I have no idea how you would communicate these decisions to an advanced raiding unit AI (short of mind reading). State toggles are too slow and lack nuance. The fastest way for a skilled player to implement these decisions is probably already in place; line move and Fight. So I do not think we could reach a point where 'fancy widgets' are playing the game for us.

With automation anyone can set Rocko to fight and have it behave and interact basically like it is supposed to. I like this for a few reasons:

• New players get to experience an approximation of the balance which high level players experience. There is a dampening on unit power variability with respect to skill level. I am allowed to make stuff like Rocko a core part of the balance because everyone can use it.
• Units can be balanced closer to what would be required if we had super-godly micro players.
• The game feels more 'physicsy' when units act closer to their 'true' behaviour and limitations.
• Multiple battlefronts are easier to manage and there is less unevenness between the penalties received for not paying close attention to different unit types. The concept of attacking where your opponents units are temporarily idiots annoys me.

quote: When someone thinks of an RTS game, there are certain traits that come to mind when wanting to improve. These include APM, scouting, expanding, micro, game sense, etc. Zero-K is the only game I can think of where coding is a valid route to improvement (not actually true, see: Robocode). This is not a bad thing in itself, but it is a very foreign concept to most gamers, and it will turn many players off. I would argue that automation, and thereby reducing the number of points of skill, results in less choices along with less clicking. Each click is a choice, and the player must handle the hidden resource of time. With less automation, there are more choices to the player regarding where to spend their time. With automation, the player has fewer things to worry about, fewer things that demand their attention, and therefore, fewer "choices" to make with regard to their time. The remaining choices are relatively high-level and hopefully more interesting, and the player can focus on these choices more thanks to automation.

As written in the first section I somewhat agree with the coding point. I disagree with most of the last paragraph.

Automation does not necessarily reduce the number of different skills required for a game. If micromanagement was the main defining skill of a game (and it is for non-top SCII players) then reducing that micromanagement lets other aspects of the game rise in importance. By "defining skill" I mean the one which would yield the most improvement if practiced. Also by "micromanagement" I include many tasks referred to as 'macro' in SCII. I also do not think it reduces the total required skill because there always seems to be higher level things to manage once the burden of lower level micro is eased.

I also do not think automation reduces choices. It does not even reduce choices about where to spend your time. For example consider you have a long list of choices about where to spend the next 10 seconds. One of the items is:

• Manage a battle. This would make the battle be a fairly even trade. Failing to manage the battle causes my army to be destroyed easily and an eventual loss of the game.

This option is quite conceivably much more urgent than everything else on the list. This choice makes the other choices on your list unviable. You have to manage the battle or lose. Now consider the case where the expected payoff of managing the battle is more in line with the other choices. Managing the battle might yield a bit of efficiency and focus fire on the unit types which you currently want your opponent to not have. This is some benefit but it is not an overwhelming payoff like "Don't lose the game". Suddenly there is no obvious place to spend your time so the choice becomes a lot harder.

I understand that this system works in SCII because players are supposed to spend (say) 90% of their attention managing the battle and have a very hard choice regarding their remaining 10% of attention. This system locks players out of this decision making until they have the skill to make the battle take less than 100% of their attention. Also, perhaps SCII needs this system because all the other choices would sum to less than 100% attention for a highly skilled player. Then the battle choice provides pressure to make the time choice meaningful (I don't know, just a possibility). Regardless, I do not see that automation reduces time choices in general because there may still be many things demanding your time. If automation makes the Attention:Urgency ratio closer to uniform then it should actually increase the difficulty of the choice.


I would like an AI competition. It should at least give us some more AIs for people to play in singleplayer. It also sounds fun to play.

<3 gogglefog