Jon Shafer joins Rob, Troy, and Julian to talk about challenge in strategy games. What kind of challenges do we want from strategy games, and how does it get botched? Why are people still surprised when AI opponents aren’t very clever? Why are they so hesitant to take on multiplayer? What’s the difference between good scenario design and unfair scenario design? How amazing is Unity of Command? Seriously, you guys.
Register at Idle Thumbs then join us in the episode thread! (In coming weeks, all discussion will be over there and podcast posts will be closed to comments.)
Having written my own AI, I have to say that I think that developers should aspire for a perfect min/max capability and then strategically introduce flaws that allows players to win.
For example, you can simulate “concentration” and give the AI points that they can spend to think about problems. This prevents one major flaw with AI, the omnipotence and the ability to micromanage hundreds of units. Then you can let the player choose how many points the AI should get.
It makes sense to me… let the player choose the handicaps to place on the AI (or themselves).
Personally, I save and reload when I play games, I even use cheats. I am very comfortable modding games and creating the experience I want. I think players need to be more responsible and manage their own fun in gaming. If you know the AI sucks against a certain strategy, don’t use it… unless you are about to lose and want to buy some time to turn things around.
Also, one thing you didn’t mention was difficulty inflation. I feel like in the past it was insulting to be set on “easy” difficulty. However, I many gamers today want to play the game on “normal” and be able to win. Some developers purposely use obscure names for the difficulty levels to offset player expectations (or to just be creative). Some games give a multiple choice question: “how experienced are you with strategy games”
Should developers be wasting time making players feel better about themselves by rebranding the difficulty levels? Is it the end of the world if a player picks the wrong one? How many people buy a game like Civ 5 and only play once?
I am not a programmer so I probably don’t understand the complexity and limitations on making AI, but I have always wanted to see an AI that actually learns from previous experience and adapts or copies the human players. I’ve mostly thought about this in terms of Civilization and SMAC. I think SMAC is where I originally wanted to see it happen, because the city governors were supposed to “learn” or anticipate what you would build next in a city based on what you did in other cities. So I always thought, if I won the game, the computer knew what I did as far as cities were concerned and should just copy me.
Then I saw open transport tycoon delux and it’s (buggy) AI GRF’s that you can download and turn off and on. It seems to modularize the different aspects of AI from road building to vehicle maintenance. As a non-programmer I hear someone like Jon Shafer mention how he did the “diplomacy” AI and someone else did the city and improvements AI, I start to wonder if something system of adaptable modules of AI would work. Only the best combinations of AI modules matched to a given players play-style would live on. Like genetic algorithms or natural selection. But I suspect most people do not play out civ to the last turn when they are losing and it might be better to rate AI on how fun it was to play against.
But then you guys mentioned how multiplayer does not appeal to many strategy game players, and I thought, I would love to be able to write rules or parts of my strategy into some sort of AI script and upload it and see if it can beat my friends, without having to play them directly. Or ideally my play style would be automatically placed into my friends single player experience as say one particular civ. Like racing a ghost car in a racing game.
Is this do-able?
When it comes to strategy games (though to greater or lesser extents the same could apply to other types of games) what I’d like to see is AIs that behave, to use a term from the podcast, like rational actors. I don’t mind them having more hitpoints or resources than me or that they’ll tend to distrust me more than they do each other. That’s fine, they’re not (yet) as smart as me so tilting the odds is necessary to give me worthy opponents but they should act like they have a purpose other than being disposable cannon fodder to be hurled against me.
An example that has stuck in my mind dates back to TW: Medieval 2. As France I’d seized large chunks of the continent. Naturally England was against me and I’d therefore activated the Auld Alliance and been a solid friend of Scotland, propping them up with money to keep England busy. When I got round to invading England I conquered Wales and gave it to the Scots but within another turn the English siege in the North lost them their homeland leaving them with just that province. My plan was to give them Scotland back after I’d taken it but whilst I was regrouping the Welsh Scots ‘decided’ to attack me instead so it was goodbye to them.
Obviously this decision was ludicrous (and similar examples can be found in many other games), driven by the game engine’s desire to keep the player under pressure, though this particular play was a dismal failure in the regard. It’s also, in my opinion, counter-productive commercially. I gave Empire a try, found no improvement, and at that point gave up on the TW series. Game designers need to find ways of giving the AI sides a sporting chance without rigging the game so blatantly that suspension of disbelief is no longer possible.
Really enjoyed this episode :)
Reply to: Greg S
“Like genetic algorithms or natural selection”,
I think this could be achieved, still for the AI to work it would require lots and lots of data; the game would need an big player-base that constantly plays against a “learning system”. The restriction (I think), development time to invest in such an architechture, one that could evaluate tons of data from players against a considerable amount of rules.
“I would love to be able to write rules or parts of my strategy into some sort of AI script and upload it and see if it can beat my friends, without having to play them directly”,
Regarding this, Gratuitous Space Battles did a good job of configuring a pre-battle, submiting it online and then looking at results and comments from people willing to try it. Is not close to programming the AI of the game (wich in the case of GSB, has its problems in my opinion since the pre-battle and other systems may be “gamed”) , still the concept brought a lot of fun hours.
Will be listening to this later tonight after work – curious what Jon has to say about challenge in strategy games, considering there’s absolutely no challenge at all to civ 5. Has anyone tried the Fallen Enchantress beta yet? Curious how that turns out…
It didn’t take you guys very long to go off the deep end, but before you did I found the discussion about difficulty and what players want out of a strategy game interesting especially in Julian’s desire to be “bad ass” and Rob’s take down of it.
I know people are applying it to everything now but what struck me is the MtG design team’s take on this and their three types of players: Timmy, Johnny, and Spike. Not every play wants the same thing from a game and a designers job is either to make a good game that works for all three types of players or to design one that is great for a subset of the three (including only one type). Julian sounds like a Timmy he wants the cool story with the big bang end. Rob sounds like a Spike he wants to master the game and then to crush it…
I disagree with Rob that Julian and players like him and are part of the problem, they just want a different experience than he does.
I am deeply disappointed by Troy’s rant about “people exploited Civ 5, then complained that Civ 5 could be exploited because people are stupid” comment. The issue is that Civ 5 was so Very Very easily and obviously exploitable. Yes, there will always be someone who conquers the world as Yemen. I have absolutely no interest in being that guy, yet I had figured out how to become an unstoppable juggernaut by midway through my third game. For $50, I felt like I should have gotten more than 2 playthroughs before I (someone who thinks about this stuff a small to moderate amount, not Yemen guy) had completely solved the system.
Unity of Command is so good for everything the panel mentioned – (the intuitive interface and game mechanics, smart AI etc). And also because, for those same reasons, the player can wrap their heads around playing toward the endgame with a changing strategy.
This became especially pertinent to me as I was battling out the Stalingrad scenario while listening to the episode. In a game in which a scenario can be played out in half an hour or so, there’s a flowing back and forth in offensive and defensive strategy you end up with. (Something I just don’t often get in WitE as I loose sight of the holistic objective of the scenario).
From a strategic perspective, I find WitE, in comparison, a bit too cagey. UoC offers of a more fluid and, for me, enjoyable experience. Kind of like the Hegemony games by Longbow, I think UoC offers that holy grail of balance between scale, pacing, accessibility and thought.
Thanks Raul. I was just browsing Soren Johnson’s old page and found this: http://www-cs-students.stanford.edu/~soren/advanced-protection.html
Sadly I am out of nukes.
Although the reason why Rob always get’s the same conversation when he meets us is because we try to set up an office multiplayer game which runs for an hour or so every week. So when you only play for an hour a week the game will run for months. Still it is good for team building, bug hunting and general improvement of the game.
On a more general note I’d say that multiplayer for our games is such a long term time comittment that not many people can actually be able to invest the time to play consistently week after week (being paid to do so helps in my expierence).
Great show. Have you guys thought of doing an episode of the UI of strategy/wargames ? I know Bruce has thoughts on the topic as it was brought up back when the WITE show happened, but it would be interesting to here the group’s thoughts on what they think should be done to improve the UIs. For example, take the UI for Tiller’s games. What would be a way to improve the UI he uses in his games? Contextual menus?
Yes, there will be a UI show eventually. Trick is keeping it focused.
Sorry, Legate:
I didn’t find Civ all that easily exploitable – certainly no more than Civ 4 was – especially after the first round of patches. There is an elite level of players, I will admit, that could do it quite easily. I had to see the Sulla ICS spam posts to really believe it.
I would never say that a player has to hold themselves back if a game is obviously broken. Complain, fight, argue, DEMAND better games. We need that. I just found a lot of the Civ 5 complaints and revelations of exploits from the elite players to be a little excessive at times.
I haven’t listened to the whole episode yet because my dog didn’t crap enough in the yard to keep my busy for a whole episode. As far as the AI and providing a challenge and getting bonuses…
Obviously if a computer opponent can provide a challenge without cheating it is preferable to one needing to cheat. Beating Unity of Command was satisfying because the AI put up a good fight. When the AI needs bonuses to be competitive, beating it is less satisfying. It highlights how dumb the AI really is. Beating my 4 year old in a race isn’t satisfying even if I give him a big head start. Beating up on a dumb AI falls into the same category.
Also, giving the AI large bonuses can negate the fun that some of the systems in the game provides. In Civilization, I love building wonders. It’s my favorite part of the game. If I give the AI large bonuses, they get an unfair advantage and are able to build my wonders without giving me a chance. I’m not an elite Civ player. I can win a level or two above the even difficulty level, but that is it. I get cranky when the AI gets my wonder not because it played better, but because it cheated me out of it.
Like others have mentioned on the show, historical games can provide a challenge by letting you play the underdog. I think it hurts historical games when the game thinks it needs to balance the sides. If one had a big advantage in real life, that should be modeled. What I enjoy is when a game can score you to show whether you over or under performed compared to real life. You can play the underdog or the favorite, and get scored. Playing the favorite doesn’t have to make it easier to get a good score, because you must still outperform history. This can be applied to non historical games to with leader boards. I enjoy seeing what percentile I rank in games. The hard part is for the game designers making a scoring system that is fair and rewards good play without exploiting the system to get a good score.
One Guy, Too Much Dog Crap (or not enough to last an episode – dog needs more food)
Well, now it rained when I was mowing the lawn so I didn’t quite get through the episode. Did Jon join the Taliban and call in from a cave somewhere along the Pakistan – Iraq border?
I agree with the single vs multiplayer dichotomy. Whenever people on forums discuss how they wish there was a stronger AI, it is inevitable that another player will pipe in that if they want a challenge they should be playing multiplayer. This is really annoying because they undoubtedly say it in such a way to imply that real players play multiplayer and single player is only there to prepare you for multiplayer. Yeah, that makes me want to play multiplayer with them.
The thing I disagree with is that Diablo 3 will have the typical single / multplayer split. I played the open beta (single player) and have to say it was boring. I’m just not entertained by click-click-click anymore. Same with Torchlight. Have any of you guys played Diablo 3? Whetever I have asked on forums what people like about it they almost always say they like to get together with friends and play it.
Long time fan. Great show grappling with some core design issues that hit all games, not just strategy games. I’ve put a lot of design work into thinking about them and I would encourage you to read my posts that address it (why are there different play styles, how do we cater to them, etc):
http://dankline.wordpress.com/2012/04/15/the-rema-model-collected-links/
It’s a deep topic, but I hope you find the analysis as useful and clarifying as I enjoy the show.