Headlights

Adapted from an IFCOMP22 Review

This is a parser based game, but the implementation feels like it is fighting convention kind of just to fight it? The attempt is to get the user to type commands more naturally, in complete sentences. I can see why intellectually this may make some sense (and certainly as a hook for voice-based accessibility), but practically speaking it feels misapplied in a keyboard-based IF. The mainstream tradition and promise of IF is to ‘put the player in the story.’ Now, since this form predates VR goggles by decades, it needed to do this in text. Text is already a layer of disconnect though, right? You aren’t ‘walking north’ you are TYPING ‘walk north.’ The more you think about typing, the less immersion you have. (Excepting of course sly little games that integrate typing-at-a-keyboard into its conceit. Those minxes!) This is why one letter abbreviations are so prevalent in IF - ‘examine’ is so many letters to type, it drags immersion, especially when done repeatedly. The more you can do this the better. The power of parser IF of course is that it presents the illusion of limitless nouns and verbs. WE know that’s not true, but that is the tradeoff we’ve negotiated over several decades: type any word, but common ones give shortcuts. I mean, I’ve personally never finished an IF game thinking, “that was pretty cool, but I really wish I could have typed more.”

So Headlights wants to renegotiate that. Sure, why not, no sacred cows, right? I think I tried twice then said, “I’m typing way too much about this,” and reverted to more standard verb-noun and abbreviation conventions. To its credit, Headlights’ parser handled it. To its detriment it put what appeared to be debug messages after every command.

>OPEN DOOR
[I heard: ‘open the door’ → Say ‘as spoken’ to repeat exactly as you said.]
The door is open.
>N
[I heard: ‘go north’ → Say ‘as spoken’ to repeat exactly as you said.]

AFTER EVERY COMMAND. Other than actively berating me for not typing articles, I’m not sure how much MORE intrusive it could be. At a minimum there should be a command to shut that off. So I didn’t care for the new parser capabilities. Sure I could have made more effort to meet the parser on its own terms, but I think I would have chafed as much or more at the extra typing.

It had some issues re-implementing other parser features as well. Objects were sometimes listed via their code relationship, not necessarily their physical description. X GROUND in one spot yielded “Inside a meadow is a physical object, a place, a side, a thing, and an inside.” It aliased verbs inelegantly like when I TOUCH LIQUID, I got “you have petted the liquid.” I think my overall favorite was USE TOILET… “I don’t know what to do with a toilet.” Wow game, your parents REALLY let you down. These gaps were not as common as the debug messages, but still overwhelming.

Ok, so the parser implementation was Intrusive. How about the story? It was pretty bare bones. 4 or 5 chapters of 4-9 room exploration and minimal ‘get X from room2, use in room4’ kind of puzzles. The maps were all pretty linear, the descriptions pretty minimal. Usually a sentence of where you are, then a line by line list of objects in the room. That was useful at least, as there wasn’t a lot of searching. My favorite puzzle was (Spoiler - click to show)letting yourself get bit by a spider for extra strength, leading me to exclaim “I’m Spiderman now!” Honestly, it felt like a test drive for the parser more than a complete work of its own. There is a climax and payoff, but the stakes never really register as more than a dry IF puzzle.

Will be interesting to see where this parser implementation goes from here though.

Played: 11/12/22
Playtime: 40min, finished
Artistic/Technical rankings: Mechanical/Intrusive
Would Play Again? No, experience feels complete

Artistic scale: Bouncy, Mechanical, Sparks of Joy, Engaging, Transcendent
Technical scale: Unplayable, Intrusive, Notable (Bugginess), Mostly Seamless, Seamless

The first two games with the Perplexity engine, Kidney Kwest and Baby on Board, were ... well, a bit different from this. Those were quiet domestic affairs. And while taking your medication for kidney disease is important, the stakes are raised in Headlights. Here, you're out in the wilderness and injured. What are you doing here? And why?

You may be able to guess, especially with the clues the game gives. The detective work is more about just looking around and finding items. The world's a bit surreal. For instance, there's dark liquid dripping from the ceiling of a cave, and when you taste it, it's awful. Guessing the liquid provides a clue. There are also minor puzzles where you need to find a way to make light or gain strength. It feels like standard cartoon or comic book logic, which again is an effective indication you aren't in the real world. But for the most part, you look around and find things based on the room's description, and the verbs you have to guess are very standard.

So it felt technically smooth, much smoother than the previous games. They certainly had their charm, but you had to wait a long time for the next move. You can probably guess what has happened to the mangled deer. Everything's pretty tidy. Though I'm still not convinced that, as-is, the Perplexity engine has any special advantages over a standard Inform parser. I like the drop-down box that appears to fill in a command, e.g at one point, you may try to PUSH BOULDER, which fails, and once you think you can, you can autofill after typing P. That's not related to syntax parsing, and I'm still not big on the debug messages that correct your grammar if you type "PUSH BOULDER" instead of PUSH THE BOULDER. But the tutorial was neat and helpful and the engine appeared faster than I remembered from Kidney Kwest.

The writer does have a good concept of design, but unfortunately the dream world introduces a lot of puzzling for puzzling's sake. If you know the conventions, there's not much to worry about, but the problem is, without much to worry about, the big reveal doesn't have a lot of oomph. It feels like implementing Perplexity for text adventures has overall been positive, and it resulted in a clean, sensible game, but I can't help the chat-style interface worked better in Thanatophobia, and the creativity of both authors (Jordan White and Eric Zinda) would be better served using something that's already there. So far I even think all three of the games would look great in Adventuron (sadly absent from this year's comp.) But it's obvious that progress is being made with Perplexity as a text adventure platform

The action in Headlights consists of looking at everything to find items that open new locations.

It relies on an artificial sense of urgency, continually telling players to hurry, but the urgency isn't supported by any gameplay mechanics. Mostly, these reminders drew my attention to the lag between typing a command and receiving a response.

There’s no real “search” command with Perplexity — an object’s notable features are revealed when you look at it. The default state of objects involves less description, which led to an infuriating encounter with “a bush, a bush, and a tree.” (You couldn’t look directly at either bush, because the parser didn’t understand which bush you wanted to check, but “look at bushes” eventually revealed that each bush had its own identifying adjective.)

Overall, Headlights felt more focused than the last Perplexity game I encountered. Most of the experience here involved applying objects logically to move to the next location. I particularly enjoyed the puzzle that involved a surge of adrenaline, because it did an unusually good job of using a narrative to justify the following sequence of events.

Headlights worked smoothly when the parser and I stuck to clear language and simple concepts, which raises interesting questions about the future of Perplexity as a game design tool. Creations like Lost Pig, Vain Empires, and Zozzled are entertaining because they play with unconventional language and abstract concepts, and it may be difficult for Perplexity to enjoy similar success.

(This is a lightly-edited version of a review I posted to the IntFiction forums during 2022's IFComp. I beta tested this game, and didn't replay its final IFComp version so caveat lector).

I have a bit of history with the Perplexity engine that powers Headlights. A custom parser that aims to provide a natural-language approach to IF so that it can be played via voice (though I’ve admittedly never tried this out), I first encountered it in last Comp’s Kidney Kwest, an educational game aimed at helping kids with kidney disease manage their conditions; despite its humanitarian aims, I cold-bloodedly lambasted it for running slowly, requiring finicky syntax (you couldn’t even drop “the” when referring to objects without the parser complaining), and neglecting basic conveniences offered by mature IF languages (no pronouns, no UNDO, awkward disambiguation). Then this year’s Spring Thing boasted Baby on Board, a comedy about dropping a kid off at day-care, which I similarly found weighed down by an engine that made things way too hard, with few upsides to justify its idiosyncrasies.

So when I saw the author of a new game using Perplexity asking for testers on the forum, part of me groaned, but a fortunately-bigger part of me realized it’d probably be better to be inside the tent peeing out rather than continuing to stay outside peeing in, as LBJ used to say (well, in slightly saltier language). And I have to say, Headlights is a great improvement over what’s come before, at least for my playstyle. At a technical level, it runs notably faster, with barely any noticeable pauses on my machine, and while the game still accepts more complex sentence structures that mimic human speech, typical IF commands are catered to as well. And because the game also offers more traditional gameplay – use-object-A-on-object-B puzzle-solving, for the most part – I could actually see the advantage of some of Perplexity’s key features, like the ability to ask where you left certain items or otherwise interrogate the game about the state of the world.

The flip side of these moves towards the norm is that the scenario is also less novel than in the two previous Perplexity games – it’s a simple series of deserted, dreamlike environments setting up a twist you’ll see coming a mile away, with straightforward puzzles that help pace the experience appropriately but don’t have much inherent interest. And some of the parser’s remaining weirdness – like its tendency to expose ugly game-mechanical constructs at the slightest provocation when they’d better be kept discreetly out of sight – undercuts mimesis. I’m still waiting on the Perplexity game that wouldn’t be better off just being implemented in TADS or Inform, but I think Headlights shows a path towards getting there: firm up the fundamentals, and once the base is solid, lean into a design that takes advantage of the system’s idiosyncratic strengths.

This is the third game by Eric Zinda with the Perplexity engine. The first two games were intended to be played with voice, I believe, while this game didn't seem to have the voice option.

The Perplexity engine is still really rough, but each game has been better than the last one. I imagine there's a ton of backend work going on between games, but I think the front-facing part could use a tune-up.

In this game, you explore a bunch of surreal areas, usually involving nature, a deer, and traffic-related imagery.

While the game is a significant improvement over previous entries, it's still pretty rough.

Polish-wise, the game tends to form uncapitalized sentences when using automated descriptions. It is smart enough to answer the question WHERE IS THE _____? but not smart enough to make the output easily understandable. This version seems to understand most traditional IF commands and abbreviations (like X for LOOK AT and I for INVENTORY, which is a big relief.

Descriptiveness-wise, the game has many rooms with a cursory description followed by a list of visual objects, sometimes kind of confusing (like 'A bush, a bush, and a tree').

When it comes to interactivity, the game is mostly fair, but at least one point in the walkthrough asks you to interact with an object that is not visible and doesn't show up in the description of other objects (specifically the (Spoiler - click to show)branch in the mossy log area).

Emotionally, I liked the surreal theme and thought it was cool. The little clues were nice. The other issues made it harder to stay invested but I like the concept.

There's not a ton of replayability, but overall I wasn't sad I played.