Repetitive Strain

This is the eleventh part of The Ouroboros Sequence, a series on puzzle games.

Sooooo… I was trying to put together a few words building on some of the Ouroboros comment discussions. It started out as a short thought experiment but I kept expanding on it until I realised this was sort of a deep dive. Welcome to inside baseball, puzzle edition.

Do not think of this as My Magnum Opus Thesis of Puzzle Design but just a guy trying to get a handle on certain concepts. I’m happy to be shot down, have a contradiction identified or be told I’ve forgotten something.

Let’s talk about the “laboriousness” of turn-based logic puzzles, because if a puzzle feels like hard work, you’re more likely to throw in the towel rather than complete the thing. The idea of a puzzle as a chore keeps coming up. Compare the “laboriousness” of Sokoban to something like a contemporary laser reflection puzzle.

But what do we mean by laboriousness? What causes it?

Complexity vs Labour

I’ve previously floated the idea of “number of steps to solution” as a measure of complexity but it is pretty crude. It’s like describing Brexit as “just getting out of the EU” and you wouldn’t want to do that.

Let me demonstrate why. Earlier in Ouroboros, I mentioned I had fallen in love with Dissembler, a very distant cousin in the match-3 family tree. The player’s purpose is to “dissemble” the puzzle and every step takes more of the puzzle away; so the number of moves required is always bounded. If a procedurally-generated level of Dissembler requires over ten moves, that’s a sign a veteran Dissembler player will find it challenging. Yet ten moves in a basic Sokoban game is nothing.

Solution step count can only really function as a relative measure, comparing like with like, but still even then may not accurately represent the mental adversity posed by a puzzle.

Now “laboriousness” is obviously related to solution step count. If I have to carry out 50 moves to solve a puzzle, that’s a lot more work than a puzzle that requires 10 moves. Again, we must be cautious because it depends what those steps constitute and how those steps are decomposed into distinct, invigorating tasks. Give me a paragraph to explain.

Although I don’t like to indulge talk of The Room being a puzzle game, because most of the experience is “hunting hotspots”, it is a good example of how doing the exact same thing again and again does not feel like doing the exact same thing again and again. The Room doesn’t feel laborious, it feels glorious. Those clicks and whirrs and mechanical explosions never get old. Puzzle games built on simple templates can test player tenacity because, after a while, the banality of repetition might bleed through into the conscious mind. Raw Sokoban is very dry and moving a little box pushing fella around a maze of boxes can get old fast.

Nonetheless, I think step count is a much better measure of puzzle labour than it is of complexity. And going forward, let’s assume there is a single solution (aside from trivial alternatives) because that will just complicate our discussion for little gain.

It should go without saying that talking about the laboriousness of a score attack puzzler does not make sense because surviving as long as you can is the whole point. However, that doesn’t mean the concept is dead. Compare Threes to 2048: Threes is tight whereas the latter is bloated with unnecessary moves.

Sequencing

In Sokoban, the sequence of moves is important; swap around any pair of moves and there’s a good chance the solution no longer functions. We could call this sequence-dependent or step-dependent, but I’ll steal a term from mathematics: path-dependent.

Now consider a laser reflection game, where it does not matter in which order you put the pieces down. Provided everything is in the right place, the puzzle is solved. As the sequence of moves is unimportant, we can call this type of puzzle path-independent.

Algorithm games are path-independent. The algorithm can be constructed in any particular order. That includes Cosmic Express, Trainyard, SpaceChem and it is even possible to cast Blek, a puzzle featuring a living scribble, as an algorithm puzzle. We can fight about it in the comments.

(Aside: path-independence sometimes means a solution of the puzzle can be captured in a single screenshot. If I show you where all the mirrors and splitters are placed, that’s all you need to know to solve a laser reflection puzzle. In contrast, if I show you the position of the sausages at the end of a Stephen’s Sausage Roll level, it might be helpful but it does not tell you how to get there.)

The Known Unknowns

That’s not quite the full picture because sometimes puzzles don’t give you the full picture. There’s a curious island between our technical definition of path-independence and the player’s experience of it. Welcome to the beautiful holiday island resort known as hidden information.

I can show you the solution of a Minesweeper level and you can conceivably just go through clicking the squares one after another in any order you want. But, come on, let’s not be fooled. It’s grossly artificial; Minesweeper does not give you enough information in the initial state to solve the entire level. The player acquires new information from each correct deduction and advances step by step. From a player perspective, it is entirely path-dependent. We could broaden our definition of “path-dependence” to include this, but I’m going to call it synthetic path-dependence for today, which makes me feel like I’m writing a load of wank.

Hexcells and RYB that I described last week have synthetic path-dependence but not classic Picross. Picross is transparent from the get-go and there is no requirement to fill the shape in the correct order.

Omniscience

I cannot think of a path-independent puzzle where the player is given an avatar or group of avatars to control. It doesn’t need to be there and the player has an omniscient presence.

Path-dependent puzzles, however, often give the player an avatar. Lara Croft GO, Aaron Steed’s Ending, every Sokoban game in history. But it’s not a given. The player is omniscient in Dissembler, Tricone Lab and Fold.

Consequences

My thinking is path-dependent puzzles tend to be more laborious because of two things: avatars and unbounded sequences.

Path-dependent puzzles often assign the player control of avatars and these puzzles almost always require more action from the player; the solution step count is much higher than an “equivalent” omniscient puzzle.

However, the very dependence on sequence can increase labour. Towers of Hanoi is a puzzle in which a boring as shit pile of discs must be relocated but the player can only move one at a time and never place a bigger disc on a smaller one. It is easier for the player to get lost in a sequence, unsure whether they have actually made progress after fifty moves, or whether they are unwittingly dismantling progress. There’s no natural ceiling to the number of moves and I’ve spoken about the type of claustrophobic rut you can find yourself in, repeating the same loop of moves over and over. That feeling of Groundhog Day in path-dependent puzzlers, where you realise you’re actually repeating the same solution attempt as before, is an experience every player knows and abhors.

The holy grail of a non-laborious path-dependent game is represented by Dissembler. Not only is the puzzle omniscient but it is degenerative as pieces disappear with each step which prevents the solution from continuing to the end of time. (I was looking for an equivalent mathematical term like “path-dependence” but I wasn’t really able to find one. I assumed graph theory would have had a term describing the condition of revisiting a vertex/node, but I couldn’t find it.)

Remember, though, path-independent games with hidden information offer synthetic path-dependence. If we view RYB and Hexcells through this lens, notice these games are also degenerative like Dissembler. Further, the limited information available at each step effectively funnels the player to deductive conclusions.

However, the deeper into Hexcells you get the larger the puzzles become and the more obscure the revelations. Hexcells Plus is dominated with vast grids offering numbers that show you the number of lit cells within two hexes. For some, this brainfest is delightful, but I found it rather tedious – like adding grind to pad out an RPG – and it felt like I was being forced to scour increasingly larger areas for an information combo that would allow me to expand my hex empire.

Countering Labour

Puzzle degeneration is the first counter to laboriousness. The Witness’s panel puzzles are path-dependent but degenerative: each move reduces possible paths open to you as the line cannot cross itself. The player must undo actions to try different strategies.

However, adding in degeneration may fundamentally change design and thus be unacceptable. Some games offer a halfway house: an indicative step count, sometimes as a bronze/silver/gold achievement. Quell, a puzzle where you roll a marble (the game calls it a raindrop) around to pick up pearls, always shows the number of moves you’ve committed versus the perfect count.

For puzzles which cannot be solved computationally, designers may be worried that more efficient solutions exist and be coy about a step target for that reason. I’d also add this is no good if you only tell the player after the level is finished such in Cityglitch (the level on the map glows blue only if you completed it with the “perfect” move count). This is step count as a par value, an attempt to motivate players to replay.

As Hexcells demonstrates, keeping level size down is the best prevention for laboriousness. The downside, as explained in Claustrophobia, is that there are a limited number of distinct challenges that are contained with smaller level designs and once they’ve been mined the designer must either throw in new ideas or…  expand the level size. The latter is always cheaper.

Role of UI

If you want a path-independent game to feel really laborious, code it up with a grid cursor that moves around sloooowly. This is entirely due to the implementation and not the puzzle design. The point is that user interface design offers methods of mitigating laboriousness. After all, labour is expressed through UI: labour is what the player has to do.

That’s why modern Sokobanlikes offer undos and fast restarts. But it’s give and take: every time a designer alleviates the strain, it means the player can tolerate larger, more complex puzzles. Recently I played Fold+ on Android and was incensed to discover it offered just one undo! Having to restart the level each time I made a mistake became frustrating because it felt like unnecessary labour.

Even laboriousness of path-independent puzzles has been improved with GUI design. Going back to the laser reflection article, compare Chromatron which has you drag pieces in from a distant sidebar onto the board with Archaica: The Path of Light which has the pieces already on the arena. Which one is quicker to assemble into a solution?

Imagine having to drag rail pieces from a sidebar in Cosmic Express and Trainyard instead of just drawing them? SpaceChem has the player drag in pieces from a menu at the bottom but it advises you of shortcuts. Dragging is labour.

And let me end on one of my greatest UI puzzle discoveries in recent years. It’s not precisely about labour but about giving players control over their mistakes. One of my most favourite things about Stephen’s Sausage Roll is that you can undo a restart. This is brilliant.

The Twist

In the Electron Dance comments, Aaron Steed alluded to a particular puzzle design that encourages laboriousness, something he termed The Turn 1 Dick Move. It’s where the solution of a path-dependent puzzle involves an “unobvious” move at the start, maximising move rewinding – and puts the solution firmly out of reach if the player does not realise the problem is that far back.

The truth is puzzle developers speak to players through their puzzles. Much of the time, they are straight up trolling and plenty of players lap up this sort of abuse. For these players, spotting the Turn 1 Dick Move generates a wry smile and perhaps a muttered “oh fucking hell”. It just goes with the territory. Puzzle designer trolls players; film at 11. The art of the perfect troll is in the eye of the beholder. To see these precise clockwork mechanisms come together to troll the shit out of you can be a wonderful thing.

It’s similar to how I don’t quite click with much of Bennett Foddy’s work and I am never playing Getting Over It With Bennett Foddy. Despite this I can admire his work and know there’s an audience for it. Some people revel in this.

Friction is essential to challenge. To eliminate “laboriousness” is to eliminate the puzzle and instead you will have a toy that gives of itself willingly.

Appendix: From n To t

The recording puzzles I hated so much do not really exist in the same puzzle dimension as what we’ve been discussing: the realm of the turn-based logic puzzle. We can’t measure “labour” in the same way because instead of a discrete sequence of moves, the game exists within continuous time (or at least the computer’s approximation of it).

Obviously we could frame recording puzzles as path-dependent as what you record impacts future play, and I guess we could generalise labour from number of moves to solution duration… but this does not sit right at all. Puzzle labour doesn’t really work as a concept for dexterity puzzles, such as Tetris or action modes of games like Spelltower and Big Money.

What makes a good dexterity puzzle is a completely different ball game. And note that just because a puzzle is not turn-based it can be turn-like, in the sense that the player is plotting a sequence of operations. (The genius Recursed is turn-like.)

Update 27 Jul 2018: I originally wrote that players needed to complete Quell levels to learn the perfect move score. This was incorrect and this statement has been amended.