Clockwork Criteria: 6 Guidelines for Ideal Strategy Game Design

What are the criteria that make something a good “Clockwork Game”?

The Clockwork Game Design model is something I have been working on for the last five years or so. It is specifically an effort to figure out how to make the most elegant and effective strategy games possible. There are certainly practical reasons why you might not want a specific game to be a Clockwork game. But to the extent that you want your strategy game to be elegant, you should adopt as many of these principles as possible.

Above: my book

Below is a list of criteria that strategy games should strive for. I am sorting them by how controversial they are. In other words, I am putting the stuff people pretty much agree upon towards the bottom.

These are not ordered by priority. I am making no statements about which of these is more or less important; just that they are all something to strive for.


  1. Error-proof input. With a strategy game, we are trying to measure a player’s understanding of a system. Therefore, we should do our best to get an accurate reading. That means that there should be basically zero times that a player “mis-clicks” or performs any action that they do not intend to perform. To clarify: when a player looks at the list of inputs they made into the system, that list should 100% match the player’s intentded inputs. This means that high execution, or really any execution, is off the table, and means that probably most Clockwork games will be turn based. More here.
  2. Balanced information horizon. Games need hidden information. They shouldn’t be about “are you willing to do lots and lots of tedious calculation“. Instead, they should be a test of your understanding of the system. Then again, the input of a player needs to be tied to the final outcome as much as possible, so the answer can’t just be slap dice rolls between every player action and the result. The answer is to have a carefully chosen information horizon: enough hidden information to prevent calculation, but enough deterministic, public game state information to associate the end-state with the player’s input.
  3. A clear, binary, enforceable goal. If there is not a clear “player-has-achieved-the-goal” state, there is no way for the player to get the crucial end-of-match feedback that gives strategic meaning to the player’s input. It’s highly unfortunate that this has to go so high on this list, but the pattern of using “high score” as a model for goals has been plaguing systems that otherwise could be coming close to the Clockwork design model. Games like Civilization, X-Com and Rogue-likes come to mind as games which, but for their lack of a clear and enforceable goal, could have a lot of potential as at least frameworks for Clockwork strategy games.
  4. Balanced difficulty. The best way to get the most meaningful feedback possible out of your win/loss end condition is by setting things up so that the player has a 50% chance to win. In multiplayer situations, you can (and we already do) do this by implementing a matchmaking system. In a single player situation, you’ll have to do something like Auro’s single-player Elo system.
  5. Relies on systemic complexity, not content complexity. Content is stuff like characters, maps, items and so on. Systemic complexity is the amount of emergent potential from the game’s rules. Most current videogames rely almost entirely on content complexity. (There is sometimes grey area here, but the above examples are usually pretty obvious.)
  6. Has a timer. All games require a timer. Sort of the other side of the “error proof input” coin; on the one hand, you want players to be able to make inputs correctly with zero error. On the other hand, you want games to reach their conclusion in a timely fashion because of how important the win/loss feedback is to allowing players to explore a game’s depth.


The following are not criteria, but steps that you will probably have to take in order to achieve the criteria. It is my opinion that designers will have to do most or all of the following in order to create a solid Clockwork game.

Single Player. What we are trying to do with a strategy game is measure a person’s understanding of that system. The involvement of a second player (or more) causes there to be player-generated deformities in the randomness that they provide to the other player; the parameters of this randomness is, despite any attempts to guide players, slightly taken away from designers. Read more here.

Further, the system now has the bizarre requirement of “forcing test-answers to function as questions for another test-taker”. So we need to provide Player 1 with a system that gives them decisions and meaningful feedback while at the same time taking their inputs and having it set up interesting decisions for the other player. In practice, this obviously can and does work, but it is not an ideal setup.

Turn Based. As I said before, probably you’ll need to make your system turn-based in order to avoid input errors. But if you can achieve zero input errors in a real-time environment, that would be fine.

No “Metagame” (Asymmetry, customizability, etc). Since you need your game to be deep, accessible, and have a single core mechanism, any attempts to wrap a game in a larger metagame (such as choosing characters or customizing a character) will be, at best, a waste of time, and at worst, damaging to your system. Also, this tends to conflict with the “one core mechanism” concept, and the supporting structure that goes with that. Read more here.

A core mechanism. A clear, singular core mechanism is a basic action around which all interactivity revolves in a Clockwork game. This, like the other things on this list, is more of a “thing you’ll need to do to achieve the criteria” rather than a criterion itself.

Spend more than a year developing it. I think a reasonable time range where you can expect to end up with a solid, polished Clockwork game is probably between two and ten years. That doesn’t mean you can’t release the game after six months; you can release it at whatever point you like. But there needs to be serious post-release patching, probably for a few years, to get things close enough. The old idea of pumping out a new game every 3-6 months and then just moving on will not result in the achievement of the above criteria.

Depth & Accessibility

A couple of side notes: is accessibility a necessary property of a Clockwork game? I had it on an earlier draft of this list, but actually, I don’t think it is necessary. I do think if you follow all of these criteria, you’re likely to end up with something more accessible than if you hadn’t, but you could theoretically have something extremely inaccessible that’s still a Clockwork game.

And how about depth, actually? I had depth on here, too, earlier. But I actually think you can have a non-deep Clockwork game. It won’t necessarily be good, but I think you can follow this model perfectly and not have a deep game. But then again, depth is relative. I think that if you have an extremely simple Clockwork game, it will be deeper than a similarly-simple non-Clockwork game.

Depth and accessibility are really what we should all be striving for in game design, but I think they are outcomes, or byproducts of using the Clockwork Game Design model, and not “criteria” for it exactly, any more than “the game being good” is a criterion. Perhaps it’s somewhat like saying criteria for engine design is that it produces a lot of power efficiently.


  • Jake Forbes

    This is a handy list and a great summary of your model.

    The only point I have a hard time reconciling is #2, in that you explicitly state that games need hidden information. I get your frequent example of chess as an example of how no hidden information can lead to degenerate computation/memorization at top levels of play. Looking at contemporary board games more broadly, though, there are quite a few games with no hidden information that don’t seem to fall into that trap. At the simpler side, you have games like Blokus or Patchwork, or at the other end of complexity, Caverna, Terra Mystica or Puerto Rico. Maybe these only seem to work better than chess at avoiding the information horizon issue because they are still relatively young games where it’s unusual to play against opponents with dozens or hundreds of sessions under their belts? (That and house rules to prevent analysis paralysis)

    Board games with hidden information generally achieve this by introducing more randomness or “mind games” that would seem to make them less like CWGs. Codex, for example (accessibility aside) has the randomness of a shuffled deck. While it is a lean deck and players have a lot of control over its content, there’s still plenty of times when a game’s cadence will be determined by the luck of specific hands coming together at the right time. Hypothetically you could make a version of Codex with no shuffle and draw mechanism where deck and hand are synonymous. At that point, hidden information becomes a memory challenge for the opponent. Finally, you could just make the deck public information to all, which alleviates randomness and memory issues, but then the game is back to chess problem, only exploded by content complexity.

    I guess my question for you is, are physical games a good medium for CWG, or does the hidden information + single player implication mean they really should be digital?

  • I really just disagree about a game like Blokus – which, by the way, is one of BGG’s Clearclaw’s favorite games. That should tell you everything you need to know about Blokus and calculation. It’s *extremely* calculate-y.

    However, I think the reason you think it *isn’t* (same with Puerto Rico and probably the other games you listed) is that you’re probably picturing them in a multiplayer (3-5 player) setting. Especially when there’s 4 or 5 players, in a game like Blokus or Puerto Rico, there’s just SO much shit going on that you can’t really calculate stuff very well and people tend to just give up and kinda play off the cuff.

    But if you were playing even 4-5 player Blokus/PR competitively, it would absolutely be a calculation grind. It’s a lot like Go. Blokus is almost exactly like Go.

    The shuffled deck in codex is an example of hidden information. You say “randomness or mind games” but, it’s all just hidden information that can’t be calculated from a given position in the game.

    >are physical games a good medium for CWG, or does the hidden information + single player implication mean they really should be digital?

    I would say it’s possible to make CWG with physical components, yes. But it’s much more limiting, particularly when it comes to hidden information. Having a computer be able to intelligently track and process random information (while still keeping it fair and everything) is a huge advantage. And game design is exceedingly hard, so you kind of need all the advantages you can get. So I would say, if you can program, and if you don’t have some specific reason to make your game be physical, make it digital.

  • Jake Forbes

    Thanks for the reply. When you say “give up and play off the cuff,” I see that as a viable sweet spot for where games can be about “contests of understanding,” to riff off your earlier post. The chaos of 3-5 players can provide the same uncertainty as a system of hidden information.

  • Yeah in a practical sense that “works”, but it’s in the way that a lot of other bad random solutions work. Really “off the cuff” is not a hard fixed point, it’s VERY loose. If one player spends 30 more seconds thinking about their moves, they’ll win. So players are really not getting feedback for their performance – they’re getting a random feedback that’s based on “how long other players spent calculating their moves”.

    In a social/party situation, this is, you know, whatever, fine, but it’s also “you know whatever fine” to play something utterly random, like a party game, in that situation.

  • H0lyD4wg

    It seems to me that one corollary of your design philosophy is that
    games should be short. That is, they should be long enough to allow
    some complexity/depth, and to give the player enough time to make
    inputs without getting tripped up by execution barriers, but not
    longer. Shorter games have better information density, in the sense of
    giving the player more bits of win/loss feedback per minute spent

    And all else being equal, I also prefer short games. A large part of
    what makes games fun is the joy of learning and mastering a skill, so
    I really like it when, having learned something during the last match,
    I can immediately start another match and apply the lessons I’ve
    learned. When a game is too long, you simply can’t play that kind of
    followup match, because by the time you’re done playing a single match
    you’ve run out of gaming time and you have to go back to
    work/sleep/study and won’t get to play again until next week. And by
    that time you’ve already forgotten what it is that you wanted to try
    in the next match.

    Different games exercise and test different skills. Most of those
    skills can be tested in a short game in ways that are intuitive and
    natural. For example, your knowledge of enemy movement patterns in
    Auro allows you to set up impressive combos in the span of a few
    turns. Another example: if you’re playing a first person shooter, and
    you don’t know the map as well as your opponent does, it will take at
    most a minute for your opponent to punish you for your ignorance by
    sneaking up on you from somewhere you didn’t expect.

    But it seems to me that there are some skills that only come into play
    in long games. One particular skill that I have in mind is the ability
    to correctly assess the difficulty of a game level. Or, as an XCOM
    player would put it,

    God, grant me the sense to retreat, early and with minimal losses,
    from engagments I can’t win, the tactical savvy to win the ones that
    I can, and wisdom to know the difference.

    If you err on the side of recklessness, you will lose valuable troops,
    but if you play too timidly, you miss out on chances to gain resources
    and experience points and ultimately fall behind as the aliens become
    stronger and your soldiers don’t level up because they haven’t gotten
    any action.

    On one hand, I do enjoy that aspect of XCOM (and other games that have
    it, like Invisible Inc.), but on the other hand, XCOM campaigns are
    too long for my tastes. (I probably won’t be able to play more than
    six over a single ten-year period) It seems intuitive that a game that
    tests your skill in choosing which levels to play and which to skip
    necessarily has multiple levels, and is therefore long.

    So I wonder if it’s possible to design a game that does test a
    player’s skill in assessing the difficulty of a challenge (level,
    opponent, board configuration etc.) relative to their own power level
    (a combination of the player’s skill and in-game resources), such that
    the game is also reasonably short.

  • Conor

    I’m not sure if you have said it anywhere but why is it you like systems where you only win or lose as opposed to systems where there is a more granular ‘marking’ method like 1st, 2nd, 3rd, 4th or some form of scoring?

  • It’s not about what I “like”, it’s about what I think is the most effective model for accomplishing a certain thing. In my view, a binary goal is the most effective way to give players clear feedback on their understanding of a system. In a system that allows 1st, 2nd, 3rd place, that kind of thing, there is a distinct strategy for achieving all of these. You can “go for” 2nd place. So the problem with *getting* 2nd place is that it’s very unclear feedback – is 2nd place “success”? Did you do what you needed to do? What can you learn from that? I can see the argument that “I got 2nd place” is better than “the match ended and all players played equally well”, but a binary goal is best, by far. For more on this, listen to this podcast episode:

    Or watch 3 Minute Game Design on YouTube.

  • Asking “is second place success?” is analogous to asking “is second place a win or a loss?”. The reason that this works as a rhetorical question which cannot be answered is because there is information lost when we reduce other methods to a binary win/loss. 2nd place has a character all to itself, defined by its relationships with the other positions; 2nd is better than 3rd and worse than 1st. 2nd place is not winning or losing, it is 2nd place.