Rob Zacny

Episode 249: Pressing Through Mud and Snow

Recommended Posts

Mohawk Games' Soren Johnson joins Troy and Bruce to talk about Drive on Moscow, the followup to last year's Battle of the Bulge. Bruce gives some insight from behind the scenes and Soren explains why writing a good AI is a large undertaking. 

 

Listen here.

Share this post


Link to post
Share on other sites

Bruce's point about the difference experience a player has between rolling some dice and having a computer generate a random number got me to think about how the experience of something like the Combat Commander series (I probably bring those games up way too much on these forums, so I apologize about that). That's another GMT card-driven war game that does have dice rolls, but the dice rolls are simulated via numbers on the cards, so you never actually roll any dice! Drawing a card to reveal a dice roll has a similar uncertainty, tension, and drama to a rolling actual dice, but it's also an interesting case where theoretically a person with a photographic memory could keep track of cards play, and calculate future probabilities based on how many cards are left in a deck. For some people that can be an advantage to cards because it can create the impression that you can only get screwed by the dice so much (although that's probably inaccurate since there can be so many unpredictable variables at work), but I also know that some people hate dice rolling games, and some people hate games that rely on cards. I think people's appetite for different forms of randomness is going to vary. Ultimately I agree with Soren's point though that it's the game designer's job to communicate some transparency to that randomness if you want to get players on board with what is happening.

Share this post


Link to post
Share on other sites

It's funny, the dice vs. pseudorandom number generator thing is something that always comes to mind whenever I hear unequivocal praise for a video game being designed as a board game first. It's fine if the board game has no randomness, but as the podcast points out it's *very* different once you're trusting the computer rather than rolling physical dice. Especially since many PRNGs have quirks; I remember running into one (I think it was stock rand() on the ps2?) that was mostly random, but every result that came out of it alternated between even and odd.  So it seemed mostly random, but if you (as it were) flipped a series of coins you got heads, tails, heads, tails, heads, tails...

 

I think most PRNGs (especially the ones used in video games) have quirks. More complex than the odd/even thing, but noticeable to the longer term player. Things like "it seems like if I miss twice, I'll often miss at least three more times before I hit again".

 

A game can make this better or worse with how it is designed. There was an early ps2 game (a launch title, I think?) called Ring of Red. It was a decent game, of the "tactical rpg" variety. They made one major design mistake, however. When two units were fighting, the game went into a "real-time" bit where the two units were marching towards each other shooting. You had the trigger for your unit's guns. As you approached the enemy, you had a gun sight view that sinewaved back and forth across the enemy unit, along with a %Hit counter counting up.

 

It *looked* like if you pulled the trigger at just the right time when the sight was passing across the enemy unit, you'd hit. Instead, whenever you pulled the trigger, it (hidden from you) rolled the dice against the hit chance and then told you you'd hit or missed.  So, it presented itself as a twitch game when it was actually a game of gambling with the odds. Inevitably that meant you'd fire too early and too often (since it looked like you had the enemy dead in your sights), and miss a lot with what looked like sure hits.  By the time you realize how it actually works, you're used to the die roll screwing you so much that the baseline assumption is that the game cheats.

 

(Related aside: I still can't believe anyone trusts online poker...)

Share this post


Link to post
Share on other sites

So this episode really made me wanna jump into a deep strategy game on mobile but I have an android tablet, vita, and 3ds. Is there something comparable on any of those platforms you guys would recommend? I'm searching around but just wanna see if you guys would have any recommendations on those platforms for games of this type. 

Share this post


Link to post
Share on other sites

I guess I'm one of those people Bruce mentioned might be a little less inclined to play the 3rd or 4th iteration of the Crisis in Command series, since I'll feel like I've "done that" already. I'm not really sure what anyone can expect from me though. I'm trying to keep up with 20 different genres, and my time is dwindling fast. I have to make some cuts somewhere. Some people might stick with what they know, but I'd rather leave what I know in a cherished place and try to find new amazing things.


I'll still buy El Alamein. The price and accessability is good enough that it's no big loss if I never get the time to dig into it.

Share this post


Link to post
Share on other sites

So this episode really made me wanna jump into a deep strategy game on mobile but I have an android tablet, vita, and 3ds. Is there something comparable on any of those platforms you guys would recommend? I'm searching around but just wanna see if you guys would have any recommendations on those platforms for games of this type. 

 

If its Wargaming on the Android, I have a short list of titles worth looking at, its well over a year old though and a number of major players in the PC wargame arena have started migrating titles to the 'droid (namely John Tiller and Battlefront's Combat Mission).

 

I try to maintain a list of games worth considering over at Playboard.me

 

Wargames worth considering

 

Boardgames worth considering

 

Strategy games worth considering.

 

Hope you find something in there.

Share this post


Link to post
Share on other sites

Fascinating news from Bruce about the AI: http://www.quartertothree.com/fp/2014/02/03/drive-moscow-drifts-back-onto-ipad-whole-new-ai/

 

If it's evaluating thousands of moves, I wonder if they have a chess AI going now? I think Soren mentioned there could be problems with that kind of thing in a wargame. I'm curious to see how it works out.

 

I was a little puzzled by the comments about AI in the games industry in general. I liked the theory that the people most knowledgeable about the game make the best AI designers. Naturally, people like that are in high demand for the next game in development, so they can't spend a year working on AI, just from a business perspective.

 

And yet, I think of someone like Creative Assembly. They've iterated all these times, and people still complain about AI generals throwing their lives away. The community helps -- but what's odd is they hired the guy who made the great Medieval AI mods, and the released games still have issues. That seems like just the type of inexpensive tester/tweaker that a big business ought to hire to avoid complaints about the AI. But it still doesn't help at release. I guess we just have to wait for patches.

Share this post


Link to post
Share on other sites

And yet, I think of someone like Creative Assembly. They've iterated all these times, and people still complain about AI generals throwing their lives away. The community helps -- but what's odd is they hired the guy who made the great Medieval AI mods, and the released games still have issues. That seems like just the type of inexpensive tester/tweaker that a big business ought to hire to avoid complaints about the AI. But it still doesn't help at release. I guess we just have to wait for patches.

Jack Lusted does community relations on the forums. They didn't hire him for his skills with the engine, they hired him for his popularity with the fans.

Share this post


Link to post
Share on other sites

From all I could find on Google after a minute, he was at least working as a unit design lead at some point. I can't find anything from when he was hired as an AI tester. I did find a QT3 thread on TMA episode 3 from 2009 where I mention it. So I either saw it somewhere at that point or I was mistaken. 

That thread also held a discussion about AI. Five years later, we're back where we always were. And I'm still forgetting that guy's name.

Share this post


Link to post
Share on other sites

From all I could find on Google after a minute, he was at least working as a unit design lead at some point. I can't find anything from when he was hired as an AI tester. I did find a QT3 thread on TMA episode 3 from 2009 where I mention it. So I either saw it somewhere at that point or I was mistaken. 

That thread also held a discussion about AI. Five years later, we're back where we always were. And I'm still forgetting that guy's name.

 

You obviously have researched this better than me. All I really know is from my own fitful participation in the CA forums, where Lusted's only visible post-modding roles were appearing in Let's Play-style promo vids and refuting any/all criticism of the AI. He was never able to speak to any of his specific duties on any particular game, which led me (as well as many others) to suspect that he didn't have many and that his hiring was a PR move.

 

Whatever the case, there was a lot of frustration that he went so quickly from "a handful of smart and simple tweaks can improve almost any AI" to "we're always looking for ways to improve the AI, but it's not easy" after getting the job. It's refreshing to hear Shenandoah and Soren Johnson talk a different talk.

Share this post


Link to post
Share on other sites

AI is a hard problem; I've done a little, and it really is just an ugly problem.

 

I've got a simple strategy game right now that's somewhat on the shelf because of the AI; right now it's just too good at playing the game, and will swarm over and crush the other players.  A human player can survive and possibly win, but only by adopting the same boring build-and-steamroll tactics that the AI uses. So, you can win, but only at the cost of the game being deadly boring.

 

That's for a game that has simple rules the AI can understand easily, and no hidden information. As the complexity of the AI's problem space increases, the AI falls behind human players because programming intuition is currently an art rather than an engineering discipline. When you have a mathematically solvable problem, your AI can play better than most people. When you have a vague probability landscape with semirational actors (ie: the players), the AI a babe in the woods.

 

Soren's solution for this is reasonable; design the game for the AI. If you can balance the math with the prediction and probability just right, you can have the AI be competitive without being overwhelming, and you can remove sticking spots in the game where the AI gets confused.

Share this post


Link to post
Share on other sites

Another solution Soren has mentioned is how asymmetrical opponents can be a designer's best friend. Asymmetry gives the designer the opportunity to compensate for whatever difficulties the AI might be facing, or at a bare minimum the AI's stupidity might at least be less apparent to the player.

 

The main problem of course is asymmetrical conflicts aren't always appropriate or desirable depending on the game being designed. Furthermore if there is the perception that the designer is cheating too much (alien squads "teleporting" in XCOM immediately springs to mind) you risk alienating players.

Share this post


Link to post
Share on other sites

why do people like series in board games and not on video games?

 

One reason is rules. If i buy a new SCS, OCS or C&C game i don't have to learn that many new rules. On a computer I don't have to learn the rules, but i do get to explore them in game. If the rules are the same in each game its less things to explore. In board games this cuts the effort buy in, on computer this is less game space for some.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now