Change Request: Outcome Prediction

We’ve moved over to the paradox forums. Please come visit us there to discuss:
You can still read the collective wisdom - and lolz - of the community here, but posting is no longer possible.

Home Forums Age of Wonders 3 Discussions Change Request: Outcome Prediction

This topic contains 19 replies, has 11 voices, and was last updated by  NuMetal 7 years, 6 months ago.

Viewing 20 posts - 1 through 20 (of 20 total)
  • Author
    Posts
  • #205255

    NuMetal
    Member

    Hello there fellow forumites!

    This is an issue that never bothered me myself, however I have experienced a buddy of mine and my girlfriend both struggling with this individually and I think it currently is a newby trap and causes a lot of frustration:

    The Outcome Prediction sucks and any experienced player doesn’t take it into account at all and only relies on his own experience while a new player will trust it and get frustrated when losing all his units in a battle that was supposed to be a very likely victory.
    I think the biggest problems are that the Outcome Prediction doesn’t take into account resistances and battlefield enchantments.
    It would take a lot of tinkering to get the system to evaluate all situations correctly so I suggest something entirely different that should be a much easier solution and solve our problems:

    We remove the entire Outcome Prediction Algorithm and instead let the AI make a quick Autobattle!
    Even on the worst PCs an Autobattle only takes a second and the outcome would be much more telling than what the system currently does.

    So in the end it could look something like this:
    If the AI won the fight without losing more than 25% of its units it sais “very likely victory“,
    if it wins and loses more than 25% but less than 50% of its units it’s a “likely victory“,
    if it wins and loses more than 50% of its units but less than 75% of its units it’s a “probably victory“,
    if it wins and loses more than 75% of its units or if it loses and the opponent loses more than 75% of its units it’s “Evenly Matched“,
    if it loses and the opponent loses less than 75% but more than 50% of its units it sais “probable defeat“,
    if it loses and the opponent loses less than 50% but more than 25% of its units it sais “likely defeat
    and if it loses and the opponent loses less than 25% of its units it sais “very likely defeat“.

    Of course this is just an example and it could be made even simpler with simply stating “probable victory” if it wins and “probable defeat” if it loses. The details would be up for the devs to decide on anyways…

    Oh and to top it all off for King and Emperor there would be no Outcome Prediction at all!

    I hope my explanation was clear and I’d like to hear your guys’s opinions! 🙂

    #205285

    Gloweye
    Member

    Yeah, I like it. I never look at them, partly because they suck this hard. It also seems not to care about walls, as well as underestimating/not counting experience ranks of units(though it does something for heroes, but it doesn’t count items, only levels.)

    Oh and to top it all off for King and Emperor there would be no Outcome Prediction at all!

    Except for this part IMO. Maybe just give an option to turn it of(performance! I’ll never have it on…). There is currently NOTHING outside of the AI players themselves that’s influence by their difficulty, and IMHO it should stay that way. For example, what would happen if I assign 1 Lord and 1 King AI? And there might be an easy answer for that(like: Lowest AI in there, or Highest AI in there), but it’s only the start of the headaches – you remember the “replacement of loss” units you get on lower difficulties? They used to bug out in MP….

    #205291

    NuMetal
    Member

    Except for this part IMO.

    Yeah that was really just an afterthought. Not really important to me.
    I only want my less skilled friends not to experience so much unnecessary frustration anymore^^

    #205298

    Triscopic
    Member

    Possibly the simulation could work more like the auto battle in SM (that’s the simulated battle for the predicted result, not the auto battle itself), which, if I remember correctly, didn’t worry about movement, and let each unit hit with ranged or melee attacks each turn. This would be less accurate of course than doing the full auto battle, but might be faster to simulate and would leave a wider margin of chance (which I think is a good thing, as long as it’s reasonably truthful, which right now, it’s not at all)

    #205312

    NuMetal
    Member

    @triscopic:

    Your idea is nice, however I’d assume that taking an existing system would be much less work than implementing a new one.

    but might be faster to simulate

    Is the speed really an issue though?
    I really don’t know since I have a decent PC but I read in another thread that someone said they play on a potato and there the game “always freezes for a second when selecting Autocombat”. And if it’s only a second on a crappy pc I don’t think that speed would be an issue.

    a wider margin of chance (which I think is a good thing, as long as it’s reasonably truthful, which right now, it’s not at all)

    This is something I thought too at first but after thinking about it for a while I don’t see it as that important anymore.
    The thing is that you still won’t know if your hero/Champion unit/unit you love will survive the battle even if you win and also it’s mainly a feature for new players anyways so a smaller margin of chance isn’t that bad imo.

    #205315

    CSav10
    Member

    I do think these are often very innacurate ;P, and often I find certain outcomes incredably amusing when they can be so wrong ;). These do not bother me that much, because it has come to the point I don’t really pay much attention to them other than amusement purposes. Though I could definately agree it would help players that are newer to the game to know what they are going into, especially if they don’t know the units and what they are capable of. Either way I think your change would be a good one and in general helpful.

    #205392

    Triscopic
    Member

    @triscopic:

    Your idea is nice, however I’d assume that taking an existing system would be much less work than implementing a new one.

    but might be faster to simulate

    Is the speed really an issue though?
    I really don’t know since I have a decent PC but I read in another thread that someone said they play on a potato and there the game “always freezes for a second when selecting Autocombat”. And if it’s only a second on a crappy pc I don’t think that speed would be an issue.

    I play the game on two PCs, my gaming rig (~5 years old now) which freezes up very briefly when I auto-resolve multiple stacks (I auto-resolve most things, unless I expect heavy losses) and my laptop (which is quite light weight) which freezes for a good second on a small battle and a two or three seconds with multiple stacks. If that happened every single time a conflict was about to occur it would get annoying fast.

    As much as I dislike the existing system, I’d rather have it than have to put up with the extra freezes. (Although I suppose if the system cached the result it could at least auto-resolve instantly, which wouldn’t be terrible)

    #205404

    The way that this solution is presented here makes me think that using words like “probable” wouldn’t fit any more. Aside from “evenly matched” the winning party would already be known.

    Not that I have a problem with that. Even if the outcome of the autobattle was given with such accuracy before selecting whether or not to do manual combat, it wouldn’t affect my play. It would avoid those situations where a player loses a “likely victory” which I think would be an improvement for new players.

    I would rename the victory types something similar to the following if this system was used:

    decisive victory
    victory
    costly victory
    closely matched
    costly defeat
    defeat
    overwhelming defeat

    Also, removing the option to display the victory type might be possible for people playing campaigns on hard difficulty, but in individual scenarios difficulty is selected separately for each opponent. I guess it could be a game option “display autobattle results” but I’m not bothered either way.

    #205411

    Triscopic
    Member

    @mauvebutterfly: I agree entirely. I was thinking the same thing, but ran out typing energy.

    This is one of the reasons I originally shied away from using the existing auto resolve code, and suggested using a “third way”, even though that would cost more. (I think Triumph prefer to make the best change or none at all from what I’ve seen)

    Edit: That’s not to say my “third way” is “the best”, but rather that it’s worth considering all avenues to find the best, even if those avenues include a more costly possibility.

    #205444

    Joppsta
    Member

    I literally suggested this yesterday, more or less.

    http://ageofwonders.com/forums/topic/suggestionmultiplayer-and-auto-battles/

    It would make MP gameplay much more bearable as currently it’s ludicrous the amount of time wasted on stupid shit by my friends, through no fault of their own.

    If I wanted to play something that had risk involved, I’d go play Russian Roulette. As it stands now, auto battles throw up some of the purest smelling bullshit I’ve ever seen in a strategy game.

    #205465

    Triscopic
    Member

    @joppsta I may have misread your post yesterday but I thought your suggestion was based on a misunderstanding. Like I said I might have misunderstood it.

    This thread isn’t advocating any change to the existing auto battle code. Rather the problem is with the algorithm that works out whether you’re likely to win or not. That algorithm is ONLY used to determine what text to display in the battle summary screen. When you press the auto battle button the game simulates the battle in full using the same AI it would normally use against a human for both it’s forces and yours.

    The issue is that the probability to win text is often wildly off vs the result of the simulated battle.

    In my reading of your post I got the impression you beloved there was no full simulation, and that the game used the algorithm to decide to kill 50% of your troops just because it came out at “evenly matched”.

    Like I said, if I was wrong on that then I apologise.

    #205472

    Wintersend
    Member

    From what I can tell, the outcome predictor just does a relative tier comparison. E.g. I have 2 Frost witches, 3 harpooners, and a mammoth vs. 2 champion rank shadowstalkers. It will probably show me as likely wining since I have 2 t3s, a t2, and 3 t1s vs 2 t3s.

    In reality, I’m going to get countered hard discounting spells especially since I am out-medaled and I only have access to frost and physical damage, neither of which works against Shadow Stalkers very well.

    Medals especially seem to get discounted since I can send in groups with multiple champion ranks per unit into a battle against no medal units that it shows very likely defeat and come out without having lost a single guy. Heck, if I’ve got a necromancer hero, I might come out even better off than I was when I went in.

    So I agree, if nothing else it needs to take more factors into consideration when judging combat probability than the flat out tier comparison which is what I’ve observed.

    #205491

    Agent Fluff
    Member

    Haha, I know right. My city gets attacked and it says “very likely defeat”, and I go like: oh yeah? OH YEAH? Bam, wiped the floor with them. >_>

    #205502

    CSav10
    Member

    Personally for me I would be glad with more entertaining battle predictions to amuse me more.

    e.g

    The odds are your enemies will send you some carrot cake at the end.

    Your Warbreed will get infected with aids and die after getting poked in the eye with a harpoon.

    etc… It would actually make them worth reading :P.

    #205504

    Joppsta
    Member

    @joppsta I may have misread your post yesterday but I thought your suggestion was based on a misunderstanding. Like I said I might have misunderstood it.

    This thread isn’t advocating any change to the existing auto battle code. Rather the problem is with the algorithm that works out whether you’re likely to win or not. That algorithm is ONLY used to determine what text to display in the battle summary screen. When you press the auto battle button the game simulates the battle in full using the same AI it would normally use against a human for both it’s forces and yours.

    The issue is that the probability to win text is often wildly off vs the result of the simulated battle.

    In my reading of your post I got the impression you beloved there was no full simulation, and that the game used the algorithm to decide to kill 50% of your troops just because it came out at “evenly matched”.

    Like I said, if I was wrong on that then I apologise.

    Not sure what you mean by “beloved there was no full simulation” but to clarify I hate the current AB system due to the fact it punishes you for trying to be courteous to other players when playing MP. In my ideal world I’d be able to see what the end result will be and then be able to decide if I MB or take the result the RNG spits out.

    This would make MP much more playable for me.

    #205537

    Chaos-Knight
    Member

    This is actually genius, I’m 100% for it. As a relative newcomer auto battles screw me over big time a lot. I learned to game them a bit by now by adjusting my stacks around the enamy hex before I engage and other tricks like that, but honestly I’d much rather have a semi-accurate prediction than this black box.

    A perfect prediction entirely without randomness may have its downsides, but a simulated battle with slight outcome variability would beat the current prediction by leaps and bounds. It really is frustrating to lose half your army on a battle that according to prediction should be trivial.

    #205553

    BB Shockwave
    Member

    Well, to be frank, I very-very rarely use autobattle. I won’t use it to cheat against the AI, I won’t use it if it’s likely a weak unit of mine would die… And I don’t even use it battles where I dominate the battlefield, because I use these cases to only attack with those units who only need just a little more XP to level up, IE I ensure I level up my units every battle. Even those who won’t attack, I will try to improve by using Medidate, throwing nets or using entangle or web, any ability that adds XP. 🙂

    The only thing, for me, what the prediction is good for is that it lets me walk into a “probable defeat” with only a small 8-men army standing against a 3*6 force come to destroy my beacon, and walk out victorious without a single loss, feeling good about countering the prediction with a manual battle . 😀

    #205624

    Fenraellis
    Member

    I think the biggest problems are that the Outcome Prediction doesn’t take into account resistances and battlefield enchantments.

    It also ignores potential Leader/Hero spell casting.

    #205626

    Wintersend
    Member

    Like I said, it seems to just be a direct tier comparison as far as I can tell.

    Sometimes auto loses, badly, on a fight that shows very likely victory. e.g. Complete and utter defeat and kills fewer than half of the enemy.

    Others, it pulls off a resounding victory against a very likely defeat scenario. E.g. no to 20% casualties and total annihilation of the enemy force.

    These are especially common in city battles where the defender has a big advantage but sometimes it happens in the open field too, especially if there is a big difference in tier or counters. As shown in my previous example of a pair of Shadow Stalkers vs. a group of frostlings

    #205653

    NuMetal
    Member

    decisive victory
    victory
    costly victory
    closely matched
    costly defeat
    defeat
    overwhelming defeat

    Those are great!

    Just like Triscopic I had that same nagging feeling that the names wouldn’t be fitting anymore but I was also too lazy to come up with better ones.
    Yours sound perfect! 🙂

Viewing 20 posts - 1 through 20 (of 20 total)

You must be logged in to reply to this topic.