Dragon Quest IV – 1989 Developer Interview
In this lengthy interview from the 11/89 issue of Famicom Tsuushin, Dragon Quest creators Yuji Horii and Koichi Nakamura talk about the challenges of making Dragon Quest IV, with a particular emphasis on the AI system. Although the later ports of DQIV would add a manual control option, for its time, the party AI system was an impressive new addition to console RPGs.
Yuji Horii – Director/Scenario Writer
Koichi Nakamura – Programmer
Horii: Ah, we’re finally finished.
Nakamura: It took long enough!
—When did you start the development of DQIV?
Horii: I believe we completed DQIII around October or November of 1987. Then we started working on DQIV about two months after that, so January of last year (1988), I think? From the start, we knew we wanted to do a story with 5 chapters, and AI battles.
Nakamura: We started the development as we always do, with our traditional “staff retreat”,1 but I remember there were no restaurants or food nearby at the first lodge we stayed at.
Horii: Yeah, that was a failure.
Nakamura: It was so cold, I remember having to pace back and forth while eating dinner, just to stay warm. Those were the auspicious circumstances in which the Dragon Quest IV development began… (laughs)
Horii: I think we figured out most of the basic elements of DQIV during that retreat. Then later, in March, I took a trip to Spain to see the castles there and get ideas for the game. By this point, we’d settled on almost everything except the finer details of the AI system.
Nakamura: Right. From that point, until about June, every single meeting we had was focused on the AI.
Horii: Our plan, from the beginning, had always been an AI combat system… but until things got more solidified in June, I was honestly very worried whether we’d be able to pull it off. We even talked about creating a manual combat mode as a backup plan, if worse came to worse. (laughs) Anyway, once we had something workable in June, then we started on all the rest of the work, in parallel: the scenario, the monster graphics, stats, the maps, etc.
Nakamura: During that time at Chunsoft, we continued to work on programming the battle system. That reminds me, compared to DQIII, the maps in DQIV really took a lot of time, didn’t they?
Horii: Previously I had drawn all the dungeon and town maps myself, but this time I had the new staff members draw them first, after which I would look at them and make corrections and revisions. There was a lot of back and forth there. For a single map, on the low-end there might be 2 or 3 passes between us, but on the high-end, it could take 6 or 7 revisions. So yeah, creating the maps was way more time consuming than DQIII.
Nakamura: Yeah, and once the maps got sent to us and we actually programmed them in, it was a real pain to then have to go back and make revisions. It felt like it was double the work for everyone. (laughs)
Horii: The maps were 90% complete by the end of last year. Actually, though, they were mostly complete by October, including the placement of NPCs, but… I took a big chance and decided we would re-drew all of them. I didn’t think they were up to snuff. So the maps weren’t 100% complete (with NPC placement and everything) until January of this year.
Nakamura: The AI system also gave us a lot of problems, and it wasn’t finished until June of this year. Ultimately we ended up changing a number of things from our initial conception at the lodge, too. The AI alone took an entire year (6/88 to 6/89).
—It sounds like the AI system was very difficult to create. What made you decide to undertake this challenge?
Horii: Up to DQIII, in a certain sense, we had been modeling Dragon Quest on earlier RPGs, trying to take the very best of them and present them in an accessible way for players. However, with DQIV, there was no model. We were faced with the task of creating something entirely new on our own. That’s how we got the idea of implementing an AI battle system.
Nakamura: DQIII had the most fully developed battle system of all the games so far. Keeping that system the same, and just changing the story, was one option. But no one at Enix or Chunsoft wanted to do that. I can’t deny that we were very anxious about doing something completely new, though.
Horii: We were very worried about that. Not anymore, though.
Nakamura: Well, maybe a little, still. (laughs)
Horii: We were worried about changing the manual control battle system which players had become used to in the previous Dragon Quest. But I believe the AI system we created came out very well.
Nakamura: I do have some regrets about the finished AI system. The very first thing I was excited about doing with this system, when we came up with it at the lodge—I kind of felt like we didn’t execute it perfectly. If only we had a little more time….
—And what was that, specifically?
Nakamura: I think we didn’t quite capture that feeling of being a “general” and strategically directing the combat.
Horii: There wasn’t a lot of time for trial-and-error once we had completed AI, unfortunately. Only when it was finished did we have a good picture of the strengths and weaknesses of the system, but by then, it wasn’t easy to make radical revisions.
Nakamura: The other thing I felt was lacking, was the wagon. You can have up to 10 characters in the wagon, but unfortunately, the game never really requires you to swap them out in any strategic way. I guess it’s ok though, since this way players can assemble the party they like and use the same characters throughout their journey.
Horii: You’re meant to swap out party members when someone’s HP or MP drops too low, but the problem is, there aren’t many “life-threatening” battles like that, so you almost never need to. The battles are relatively easy in DQIV. I think if you stuff an RPG with too many intense battles, the experience becomes too stressful for players. The mid-bosses in DQIV have a lot of bite to them, though.
Nakamura: So yeah, although I have my complaints, looking at it overall, I think the AI system was a success. In fact, I think precisely because this was a Dragon Quest game we were dealing with, we were able to craft an intelligible AI system. Creating so many different algorithms and routines for an RPG combat system is a massive undertaking of time and energy, but I think Dragon Quest has a certain simplicity that made it feasible. Would other RPGs be able to implement an AI combat system like this… now that I don’t know.
Horii: That is to say, in DQIV, it’s a complete AI system: there’s no manual control at all, except of the hero. Even the final boss is fought with AI. Non-damage magic spells, items, everything is handled and decided on by the AI alone. It’s a crazy amount of logic routines we had to program in. I think creating something like that from scratch, without the pre-existing framework of the Dragon Quest combat system, would be very difficult to say the least.
Nakamura: How many times did we overhaul the AI system programming… 12 times? That’s why it took over a year!
Horii: The AI wasn’t very smart at first. It wasn’t that the programming was bad, it was the algorithms.
Nakamura: Yeah. It wasn’t bugs, it was our basic logic that was the problem. In the end we were able to create a better AI that would choose the best option—but if the AI always chooses the best option, it makes the strategy meaningless. We learned that an AI that’s too smart is also problematic.
Horii: In the beginning, we did talk about the possibility of adding a manual control mode. We were sure that if we did that though, everyone would use it instead of the AI… so we took a chance and removed the manual mode option entirely. In order to make the AI fulfilling to use, we added the strategy commands (Offensive, Defensive, Use No MP, etc). We wanted you to feel like you were directing the battles. But ultimately, I think we made the AI far too smart, and I do wonder sometimes if it would have been better to leave the manual mode in… (laughs)
—So you were planning to leave a manual combat option in, at first?
Horii: No, no, we never added one. We made that decision very early in the planning. The final version of the AI system closely matched the initial vision we had, but the way we got there—the algorithms we actually coded—were very different from what we had first imagined.
—Do you think the AI system will be too confusing for some players, then?
Nakamura: No, it’s easy to understand once you try it. It’s very simple. And compared with DQIII, it’s far easier to control. I feel like it really shortens the amount of time you feel like you’re grinding. I really felt how convenient the AI system was once I’d got all the characters, and they’d reached the level where they knew both gira (fireball) and begirama (firebane), and then seeing the AI decide which spell to use in the right situation—that was amazing to me.
Horii: Yeah, in the latter half of the game, a human player normally wouldn’t use a weak spell like gira (fireball), but the AI knows how to make efficient use of your MP. It actually learns the Max HP of the enemies as you fight, too.
—Does the AI ever make mistakes, or do things you don’t want it to?
Nakamura: Even a human player, when you first encounter new enemies, will sometimes make mistakes. So yes, your allies might do weird stuff at first, but as you fight that enemy again and again, they’ll get smarter. Eventually, they shouldn’t make any more numbskull moves.
Horii: Even within a single battle, your allies learn with each passing turn, in real-time. And let’s say they cast stopspell, but you’re unlucky and it doesn’t work; that doesn’t mean they’ll never use stopspell on that enemy again. They’re smart enough to try more than once.
—Do your AI allies have “personalities” that they fight with?
Horii: Well, in the sense that fighter characters will mainly fight, and healers will mainly play backup/support, then yes, they have those personalities. But they don’t fight according to their in-game personality, if that’s what you mean: for example, Alena, who is headstrong and does her own thing, won’t act that way in battle. And just because Brey and Clift support Alena in the plot, that doesn’t mean they’ll only do that in battle. We thought about adding those kind of things, but we weren’t sure it would ultimately make for a satisfying game. We also thought about having monsters join the party, but we felt the AI system was enough for this game.
—Aside from the AI, what other parts of DQIV are you proud of?
Nakamura: We had a lot of great ideas for the dungeons. There’s one that takes place in a tree, lots of fantastic stuff.
Horii: Almost all of the dungeons have something visually interesting going on, I think. Every dungeon has some kind of trick or trap going on. Typically, dungeons exist just for one single event at the end, but we really loaded these with unique things. It kind of feels like we went a little too far, even.
—Are the dungeons small in size?
Nakamura: No, not particularly. There’s some huge ones in there too, right?
Horii: Yeah, there are. But I dislike dungeons where it’s just a succession of floors and stairways that all look exactly the same. It may be common practice in RPGs, but we were determined not to repeat that.
—How difficult are the dungeons in DQIV?
Nakamura: I wouldn’t say they’re easier.
Horii: They are vital to the game balance, as well. We decided early in the development that when you’re walking around the overworld map, the party members in the wagon would also gain experience. That was to prevent characters from being permanently under-leveled and thus unusable. However, that alone would have been boring, so for variation, we made it so you have to leave the wagon behind when you enter a dungeon. We wanted to heighten the sense of fear and tension for the player that way.
The thing is though, when you think about it closely, most of the leveling in Dragon Quest games happens inside dungeons. Dungeons are where the stronger monsters are too. If you always had to leave the wagon behind in dungeons, then, the gap in level between characters would grow too large. To solve this, we created a number of dungeons where you can bring the wagon in, too.
Nakamura: You don’t get the wagon until the 5th chapter. From there, I think about 40% of the remaining dungeons allow you to bring the wagon inside.
—I thought the “Bank” in DQIII, where you could store items and gold, was super convenient. Will it be making a return?
Horii: Hmm, will it, indeed? The thing about conveniences like that is, once you introduce them to your series, you can’t take them out. If we removed the Return spell now, for example, people would get angry.
—Where did the idea for small medals come from?
Horii: DQIV is a multi-chapter story, and you ultimately end up gathering eight party members together, so we thought adding a bunch of story items you had to collect on top of that would be too much, and annoying. But collecting things like “orbs” and “crests” is undoubtedly one of the appeals of Dragon Quest, so we wanted to give players something to collect.
“Ah hah! Coins!” — that was our first idea, and small medals were originally “gold coins”. But that name was too easily confused with the coins at the casino. Then we remembered the idea we had for “small medals” for Dragon Quest III,2 and that’s how they got their name. (laughs)
—Dragon Quest IV takes place in a totally different setting and timeline from III.
Horii: Yes, it’s entirely different, but the first and second chapters are callbacks to Dragon Quest and Dragon Quest II: you begin your adventure alone, and next you get party members. We wanted to give brand new players to Dragon Quest a taste of the fun of those previous games.
Nakamura: The first chapter is very short. For us, it doesn’t take more than two hours to clear. Even new players should be able to finish it in about 3 hours.
Horii: The first chapter sort of resembles Aliahan from DQIII: one castle, one town, a dungeon, etc. The second chapter is much longer, compared to the first.
Nakamura: People seem to really like Chapter III, with Torneko. It’s unique.
Horii: Yeah. There are some people who really hated it, having to raise all that money… but it seems the people who liked it really loved it.
Nakamura: Horii, I imagine you find scenarios like that, which go in some different direction, the most interesting to write?
Horii: Not really. Dragon Quest, you know, has a certain framework, which you can’t really stray too far outside of. I’m more inclined to make small changes to the scenarios to distinguish them from each other.
—Chapter 4 features the dancing sisters… being adult women and all, I was hoping for a little erotic content…
Horii: Sorry, there’s none of that!
Nakamura: By the way, Horii, what is your favorite chapter?
Horii: I’d have to say Chapter 5, I think. It’s where the AI and the wagon come into play, and it connects all the chapters preceding it. In terms of specific parts, I especially like the opening scene to Chapter 5.
Nakamura: (laughs) Definitely, that opening is really cool. Was it your idea?
Horii: It was something that came out in our staff brainstorming sessions.
Nakamura: Yeah, Chapter 5… you can really enjoy the AI in that one, since all the party members are together. But I also feel its kind of a shame, that in the latter half of the game the AI really comes into its own, but then before you know it the game is over. I wish we had been able to let people enjoy the system a little more.
Horii: It’s a dilemma. The true value of the AI only comes into play once you get all 8 members. But making the battles longer, so players can enjoy the AI more, would probably be frustrating and tiring.
Nakamura: The dialogue took you a long to write this time, didn’t it.
Horii: Over 7 months. To be honest, I underestimated how long it would take. I figured Chapter 5 would take about 1/2 of the amount of time that Chapters 1-4, in total, took… but that wasn’t the case. I had to change all the dialogue in Chapter 5 (including for locations you’d already visited), so it actually took more time than Chapters 1-4 together. For awhile, I seriously thought it would never end. (laughs)
—In Dragon Quest III, if you select a female Hero, in the beginning your Mom says, “I’ve raised you as a boy”… I remember thinking, wow, what a crazy home. (laughs)
Horii: In Dragon Quest IV, the dialogue doesn’t change that much whether you pick a male or female character. People did complain about that in DQIII (that there weren’t enough differences), so we had intended to include more places where it differed… but I just never got around to it. (laughs)
Nakamura: I got the sense that you simply forgot there even was the option, for male/female heroes. (laughs)
Horii: I was only aware of it for a moment. Midway through I remembered it again, and it was like… well, there’s nothing we can do now. (laughs)
Nakamura: DQIV has twice the memory of DQIII, so of course that meant having twice the dialogue. Actually, it’s more than twice as much dialogue, isn’t it?
Horii: The dialogue alone is 3x as much as DQIII, yeah. Even then, there were still lines we had to cut by the end.
Nakamura: Yeah, there were a lot of lines I was sad to see go. Lines that gave a lot of extra flavor.
Horii: In DQIII, it was even worse: we had so little memory we had to cut whole events… event after event. For DQIV, we only had to cut some of the dialogue to make it all fit.
Nakamura: Yeah, overall, there were very few events we had to remove. Or to put it more accurately—if we thought something wouldn’t fit, Horii wouldn’t bother creating it to begin with. (laughs)
—Were any of the cut scenes or events from DQIII included in DQIV?
Horii: There’s actually lots of things we’ve been wanting to include since DQII, that we still haven’t been able to add… even in IV.
Nakamura: Yeah, there’s monster graphics we’ve cut. And some of Sugiyama’s music he composed still has never seen the light of the day. We actually used a couple of those older compositions for DQIV. I think the music is used really well in DQIV. Overall, it has a more “classical music” feel than DQIII.
—Do you have Sugiyama compose everything first, then you match it to the scenes, or…?
Horii: No, it’s different. First we ask Sugiyama to create the songs that we know for sure that we’ll need. After that, we ask him for other songs, say for a village or scene we where we think it might feel nice with some different music. By doing it this way, we don’t end up with any unused music.
Nakamura: Of the DQIV songs, I really love the requiem Elegy. I think it’s a masterpiece, personally. Of all the reqiuems we’ve had in Dragon Quest, it’s the best yet. I like it so much, I hope someone adds lyrics and vocals to it. (laughs)
—Did the balancing take a long time for DQIV?
Horii: Actually, we didn’t have much time for balancing at all.
Nakamura: Yeah, in that regard, we did a better job in DQIII. I think DQIII had the best balance of all the games.
Horii: We didn’t have much time to balance DQII either and it ended up being too hard. And the amount of time we had to balance for DQIV was similarly lacking.
Nakamura: Still, I think DQIV is better balanced than DQII.
Horii: Well, it’s strange, but human beings seem to remember difficult things. There are people who think DQII was the best Dragon Quest game precisely because it was so difficult. Our struggles lodge themselves deeply in our memories. In contrast, DQIII was very fairly balanced, so when it’s finished you feel refreshed. (laughs) DQIV is a little more difficult than DQIII. But there are no really cruel moments where the dungeons are too long or there’s enemies every two steps, nothing like that.
Dragon Quest IV – 1989/90 Interview
sourced from the GSLA
In DQIV, the NPCs in town are not simply messengers who exist to convey some information to the player. Each one is living their own life, with their own personal drama. Also, while the protagonists of the previous Dragon Quest games were all essentially avatars of the player, in DQIV they have their own individual stories and drama. In earlier games it was fun progressing towards the final goal. That is still here in DQIV, but the human relationships that are interwoven into the main quest are now a large part of the game’s appeal, too.
The day/night system from DQIII was very popular, so we naturally kept it in for IV. RPGs already involve a lot of talking with NPCs. Just talking to them one by one takes a lot of time. So in DQIII, we didn’t try to add dialogue that changed according to what you did in the game world. Actually, in some limited, special circumstances we did, where we didn’t think it would be too burdensome for the player.
In fact, we had plans for one such special scene in DQIII, in a town called Portopia.3 It happens when the hero stays at the Inn. During the day it’s just your normal Inn, but at night the Hero hears a rustling sound in the dark… and upon waking and checking his belongings, he discovers that the Final Key, which he had only just obtained after a huge ordeal, has vanished! Panicking, he starts to gather information from everyone in town: “now that you mention it, I did hear something strange last night.” Following the trail of clues, the hero makes a startling discovery. But this took way too much memory! So unfortunately we had to cut it.
The era when RPGs competed over who has the biggest map is already. Today its the content that matters: quality over quantity. So the map for DQIV is about the same size as DQIII. The numbers of towns and castles has increased, of course. But DQIII had around 70 dungeons, and there was no way we were going to double that with something like 140 dungeons! Please consider the expanded 4MB size to be for increased, deeper content.
Also, as for dungeons, I think players are getting bored of just navigating mazes. For DQIV, we removed the complex mazes. Instead we added various traps and contrivances to entertain the players.
The monsters are always a feature of Dragon Quest, and Toriyama does that work. The illustrations he gave us for DQIV were really great, and they inspired us to come up with all kinds of special attack patterns. There were also a lot of funny-looking monsters that made us burst out laughing. I think there’s even more silly ones than there were in DQIII.
Menu and Interface Revisions
One issue we faced was clunkiness with the system for buying items, and specifying the quantity you want to buy. For example, when you buy a Medical Herb the shopkeeper asks how many you’d like to purchase. But he also asks the same thing when you buy the Copper Sword–and who is buying more than one of those?! It also seemed weird to only have the merchant ask “how many?” when you buy items like medical herbs, though. Updating even that aspect of the shop menus proved to be difficult. But I think that with each game we’re gradually ironing out the frustrating parts. At meetings everyone airs their grievances about such things.
Memory Limitations and the Notebook command
You could call it the fate of every RPG, but when it comes to solving puzzles and mysteries, if enough time passes the player will forget the clues he’s heard. In DQIII we addressed this by having most of the information you’d hear that day relate to puzzles in the same area. But that gives the events of the game a very segmented, disconnected feeling. People weren’t very satisfied with that, and we wanted to emphasize the story elements more for DQIV, so we increased the amount of foreshadowing throughout the game.
But this introduces its own dilemma. Previously we had talked about adding a “Notebook” command to Dragon Quest, for players who couldn’t remember everything they’d heard. It would record what people in the various towns said. However, recording one entry took about 5 bytes of space. If you record 50 people’s conversations, that’s 250 bytes. There was no way we could afford that much memory to be eaten up, so we had to abandon the idea. The cart SRAM needed to be used for many things other than Notebook recordings; to make-up for the scant 2kb of RAM on the actual Famicom system, we had to use the cart SRAM for battle AI, map storage, and more. To add more SRAM would have raised the price of the game by about 2000 yen (~20$), and DQIV already had two SRAM chips installed.
On Dragon Quest sequels
If we had written another story, we could have continued the Loto4 saga. But I thought we could probably make a better game by starting a whole new story. I’m thinking that DQIV’s story will be the start of a new trilogy, too.
I think there are two ways to make a sequel. You can keep the game system exactly the same and simply change the story, or you can evolve the game system. I think Dragon Quest’s has been successful precisely because we always update the game system with each sequel. But if you make too many radical changes, I think you run the risk of alienating new players. So for DQIV we made less changes to the game system than in previous games.
If we had made DQIV on the Super Famicom,5 the biggest change would be the graphics. And maybe the increased graphics capabilities themselves would spur us to create new events and scenarios. But I think that having more memory is what will really increase our possibilities. Every time we make a Dragon Quest game, we have to cut out about 1/3 of what we’ve planned… if only we had more memory, we’d have so much more freedom. Just having more space for side quests would really increase the quality of the game, I think.
Ideally, I think it would be amazing to have a game where fighting monsters isn’t necessarily the main goal. You could fight them, or maybe you just want to do something else. One problem that would come up there, I think, is the so-called multi-ending. If a game has 5 endings and the player only finishes one, wouldn’t he be left without a sense of closure? Wouldn’t it be a pain to then have to play the whole game again to see the other endings? On the other hand, if you know there’s 5 endings from the beginning, its like you’re playing 5 different games, not one. Instead of having 5 games in one title, it would be better to just make one solid, good game… anyway, it’s a difficult matter.
If you've enjoyed reading this interview and would like to be able to vote each month on what I translate, please consider supporting me on Patreon! I can't do it without your help!
Basically a corporate vacation where key members of the development team would get together at a lodge/resort and brainstorm ideas for a new project.↩
Small medals were originally planned for the Famicom version of DQIII, but were cut due to space requirements. They were left in the ROM data, however.↩
Interestingly, the name “Portopia” comes from Yuji Horii’s first game, a crime mystery.↩
Erdrick in the US↩
At the time of this interview in 1990, the Super Famicom was not yet released. They may have begun planning for DQV on it, however.↩