project-navigation
Personal tools

Author Topic: Autoresolve Mechanics - Use Statistics  (Read 4252 times)

Offline narfanator

  • Cannon Fodder
  • **
  • Posts: 5
    • View Profile
Autoresolve Mechanics - Use Statistics
« on: September 17, 2011, 11:30:37 pm »
So, in late (mid? I ran out of research, but there's mention of stuff here that I haven't seen?) game, I started using auto-resolve almost all the time (honestly, got bored...), but I noticed that, well, it's not very accurate in my success rates. Terror/harvester missions, and even some crashed/landed UFO, I lose, but if I actually play them, I win with no deaths and usually no causalities at all.

Blah, blah, blah -

I'd like auto-resolve to take into account my success rates on similar missions, as a baseline. So - You have a data space of what each side brings to the fight, and you have the outcome of those fights. Look for the most similar fights that were fully played, make some guesses about the differences between this fight and those (more/better aliens/equipment), and use that to tweak the average outcome of those similar fights - more/better aliens, take more damage, worse/fewer, take less. If the system doesn't think it can make an accurate resolution, grey out the auto-resolve option, so that it can get a data point there. Added bonus - I know that any fight I can't auto-resolve has some new element.

Am I making sense?

Offline geever

  • Project Coder
  • PHALANX Commander
  • ***
  • Posts: 2561
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #1 on: September 18, 2011, 12:31:16 am »
The autocombat is being reimplemented.  We are going to take number of soldiers, their health, their equipment (and some randomness of course) into account when deciding who will win. I think using combat statistics would be wrong for several reasons, like:

* what to do when no statistics available? (no missions played manually yet)
* it would be abuseable (you can win almost any mission with 10000 retries)

-geever

Offline narfanator

  • Cannon Fodder
  • **
  • Posts: 5
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #2 on: September 18, 2011, 11:34:14 am »
If there are no (of insufficient) statistics available, either don't allow auto-resolve, or don't use the statistics to determine it.

Here's my issue: I love sniper rifles. I'm pretty sure I'm losing these missions because the game's valuation of them doesn't take into account how well I use them; that is, it just thinks "These are game-start gear, and so aren't up to par with mid/late game missions.". Maybe I'm wrong, but that's going to happen.

The idea behind taking statistics into account isn't to replace the classic designer-set valuation method - it's to adjust it. If I, as a player, suck or rock at using this or that equipment, have the game notice - the idea is that the auto-resolve resolves it as if *I* played the mission. I mean, how often have you been surprised by how a player uses or does a certain thing; how effective this or that is in their hands? Don't try to predict the effectiveness of everything ahead of time - take the data the player is giving you, and use that.

I don't see how your abuse case works, and I don't see how it's any less abusable than it is now. I save before every mission (takes about a second and a half, esc-save-click-enter-esc-continue), and if it goes really poorly, I just reload. Bam, ensure a win. If you mean, retry until whatever random element is in there gives me the number I need to win - any system with randomness has that issue, and, to be frank, I'm not suggesting any. I think you should have some, but it's not inherent to this strategy.

The generalized rule of game design I'm working off of is to never *require* a player to do a something they consistently succeed at. If they enjoy it, by all means, let them repeat to the end of time, but when they stop enjoying it? Don't require them to. When I've curb-stomped the same aliens (or near enough) on the same map for the fifth time, don't make me do it again. And if the auto-resolve thinks I'll lose, when I have always won, you're effectively making me do it again, or not at all.

TLDR: Take all those things you're talking about into account - soldier skills, equipment, and a bit of randomness, and then adjust based on how well that formula would have predicted the battles I've already fought. So let's say every time I walk into battle with what the formula thinks is winning, I lose, and every time I walk in with something it things should lose, I win, it should flip it's expectations. Make sense?

Offline H-Hour

  • Administrator
  • PHALANX Commander
  • *****
  • Posts: 1923
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #3 on: September 18, 2011, 01:01:13 pm »
Don't like the idea. It could reward the player for playing easy missions and auto-resolving difficult missions.

The only incentive to auto-resolve should be boredom/impatience. The player should perform worse on auto-resolve than they would playing the mission themselves.

Offline narfanator

  • Cannon Fodder
  • **
  • Posts: 5
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #4 on: September 19, 2011, 09:42:20 am »
I think I'm still failing to communicate this idea to you. How could this result in easy missions making auto-resolve think hard missions should succeed? You'd have different statistics for each.

The motivation is to make sure I never have to repeat a mission unless I want to or do not have consistent performance on it.

The result of the mechanism is that if I consistently always win a given mission with a given difficulty level with a given soldier/equipment combination, auto-resolve should result in a win when I go to play that mission again.

The mechanism is that the game adjusts the auto-resolve valuation of equipment and soldier combinations based upon actual real game data, as the game is being played.

I dig what you're saying that auto-resolve should result in a worse outcome, but don't make me play the same mission fifty times, especially if the outcome is always the same. If I want to, I want to, but don't make me.

Offline geever

  • Project Coder
  • PHALANX Commander
  • ***
  • Posts: 2561
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #5 on: September 19, 2011, 11:20:29 am »
We have considered your suggestion and rejected it. Thank You.

-geever

Offline H-Hour

  • Administrator
  • PHALANX Commander
  • *****
  • Posts: 1923
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #6 on: September 19, 2011, 12:33:57 pm »
The mechanism is that the game adjusts the auto-resolve valuation of equipment and soldier combinations based upon actual real game data, as the game is being played.

This is the mechanism that can be easily manipulated, which is why I don't support it.

If I want to, I want to, but don't make me.

No one is forcing you. There is an auto-resolve option. But you must accept that auto-completion will be less advantageous than manual-completion. That's why manual exists.

Offline narfanator

  • Cannon Fodder
  • **
  • Posts: 5
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #7 on: September 19, 2011, 09:19:35 pm »
I still don't see how it can be manipulated. If you play a bunch of easy missions, it would only affect the easy mission resolution, not the hard.

Offline DrewfusMaximus

  • Cannon Fodder
  • **
  • Posts: 9
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #8 on: September 19, 2011, 10:36:51 pm »
I hate autoresolving!  Only I have to do it, because there are bugs in the system. At least, that's my excuse!

I know that I can win, but I seem to fall into a lapse of reason when I autoresolve.  It's not bordom!  Quite the opposite.  I'm usually just tired, I wanna play, but I've only got, like, two soldiers left.  I feel irresponsible letting those Taman Bastards just run the planet, so, sometimes, I autoresolve.  It's just to show the world that I am still in the game.

What might be fun is if there were some questions during the autoresolve:

'Do you wan't to risk your Sergeant's life saving Hostages, or just let them die as you cower from cover?'

'Will you be using your own weapons, or the ones that you find on the dead Alien bodies?

'Standard aproach, or land on the Alien's wrecked craft?'

Really though, it's just a game, and if you don't feel like playing, you should get some sleep.

 ::)

Offline Mattn

  • Administrator
  • PHALANX Commander
  • *****
  • Posts: 4831
  • https://github.com/mgerhardy/vengi
    • View Profile
    • Vengi Voxel Tools
Re: Autoresolve Mechanics - Use Statistics
« Reply #9 on: September 20, 2011, 08:21:28 am »
i like the idea to bring some user defined variable into the auto resolve - but i'm not sure whether this would make it into 2.4. but if you open a feature request ticket for this, we will discuss it in the future and maybe you get your feature in the future.

Offline DrewfusMaximus

  • Cannon Fodder
  • **
  • Posts: 9
    • View Profile
Re: Autoresolve Mechanics - Use Statistics
« Reply #10 on: September 21, 2011, 08:16:52 pm »
Feature Ticket, hmm?

Okay, well, lemme ask some poignant questions directly, if I may...

What is the purpose of having autoresolve in the game, exactly?

When autoresolving, do I lose an opportunity to gain veterency for my soldiers?

Are kills assigned to my soldiers if I 'win the battle,' even thought I was never actually there?

Can I expect an unarmed, unarmored soldier to 'die with honor' on the battlefield, or must I actually put some thought into my assault team?

Does game engine use a calculation method based on my last encounter, with variables such as losses/vs/wins?

If I could get those questions answered, I could write a fairly serious proposal, and introduce it as a feature request.  Otherwise, please feel free to take my idea and run with it.  You have my blessing!