Sign in to follow this  
Wavinator

How, exactly, do you design a strategy from scratch?

Recommended Posts

What exactly is a strategy? How do you distinguish a good strategy from a bad one? How do you determine when a game has a strategy versus when it does not? I feel like this one's so basic it's like trying to define the word "the." Sure, we all know what it means, but what does it really mean? I'd like to test people's definition with a small caveat-- the definition can't have anything to do with war, fighting or combat of any kind. A friend and I used a wonky made up game to help flesh out exactly what a strategy is and what distinguishes a good one from a bad one: Zombies On The Playground. Our objective was to analyze how to come up with good strategies a kid might use to escape a school infested with zombies. It was structured so as to move away from the well known arena of combat (which is so prevalent that I think it makes it hard to dig beneath the surface beyond offering examples of strategies). What we came up with was this: The Basics 1) In general, a strategy arises from at least two options which are different routes to one or more goals 2) A strategy generates different result / consequence once executed (or there would be no point in choosing one or the other). A good strategy generates a significantly different result / consequence. 3) Each option carries with it a different level and/or type of risk. 4) A good strategy is one which will not be effective in all situations (else it becomes a dominant strategy) nor ineffective in in most situations. Now The More Nebulous Parts What are the actual elements of a strategy?
  • The actual elements of the strategy are those actions / tactics which arise from the situation the player / avatar finds himself in and which would reach a desired goal. Daydreaming, for example, might be an action, but probably not one that would reach a desired goal when flesh eating monstrosities are chasing you.
  • More specifically: Strategies arise from a series of natural (and effective) responses to a situation. To this end, elements / actions that fall outside the situation make for strategies which do not resonate with the player. (Example of a bad strategy would be kids being able to lead zombies into quicksand, which you wouldn't expect to find in a school.)
  • The elements must arise from the execution and interplay of rules which the player understands in advance. If zombies are repelled by fire, for instance, then the use of fire becomes a strategy (contingent on other rules, such rain/water putting out fire, which makes the strategy effective only in certain situations). A good strategy keeps this consistent (zombies shouldn't suddenly and inexplicably lose their fear of fire just because of a level load.)
  • Each element/tactic can be considered a chain of moves / actions which, taken together, allow the player to achieve a certain goal.
  • Each element/tactic results in a change in game state or resources (???).
What do you think? I'd like to see if this can be extended or improved, or hear what you think. If possible, I'd like to keep this general enough to apply to a wide variety of game mechanics, which is another reason for the restriction on combat (violate it if you have to, but try to generalize it to noncombat games if you can).

Share this post


Link to post
Share on other sites
I think it's a good start, but not abstract enough.

I'd go towards Go. Turn based, with a large branching factor (one move impacts many later moves).

Abstractly, games are an initial state, a end condition (optional) and a series of 'interesting choices'. The choices change the state, which create new choices.

A strategy I'd consider to be a series of those choices that follow a pattern.


As for good/bad, depends on if you mean to the designer or the player. And the rules of the game. Most every game has strategy imo. The differences come in the nuance and variety that the rules allow.

Share this post


Link to post
Share on other sites
One way I visualize strategy is to think of a “decision” graph. For example, you are given choice A and B. If you execute on choice A now you have choice C and D and so on. This builds a tree that represents all of the choises someone has. Now, I think strategy and devising a strategy are made up of the following:

1 - Identifying what choices are available now
2 - Identifying what choices will become available upon picking a certain path down the tree, then doing the same thing with the next choice.
3 - Identifying the impact of a certain choice anywhere in the tree.
4 - Selecting a series of choices (a path through the tree) that will either maximize some positive, minimize some negative, or a combination.

In my option, #4 is really the strategy but #1-3 are important because we never formulate a strategy with complete knowledge of all choices, their impacts and what choices we'll have then.

Share this post


Link to post
Share on other sites
Re:

Quote:
What exactly is a strategy? How do you distinguish a good strategy from a bad one? How do you determine when a game has a strategy versus when it does not?


I would just like to point out that a strategy is an heuristic/plan/algorithm to solve a problem that a player devise. It is not a property of a game. A game does not 'have' a strategy.


A strategy consists of:

o At least one defined goal
o A list of action/reaction Laws


In the Playground Zombie example, this is a complete strategy:

o Strategic Goal: Scare the Zombies away
o Action Laws:
1) If Zombie is in throwing range and you pickedup something, throw it at the Zombie
2) If you have nothing to throw, scream.

I think that you are trying to discuss is what makes a game rich soil for the player to grow strategies, i.e. The elements of a strategic situation.

Share this post


Link to post
Share on other sites
Strategy can be as simple as patiently waiting for guard patrols or as complicated as systematically eliminating the enemies food supplies to destroy their army.

The only way you know when a strategy is good is when your thought process pays off and you succeed.

Share this post


Link to post
Share on other sites


I'd propose that a strategy boils down to a choice of tactics that the player choses to exploit while trying to reach their end goal.

First, the player has a choice of tactics that he/she can exploit.
I'd divide these into 4 categories: Time Burning, Weak, Strong, Exploit.
In general we should ignore the Exploit tactics other than considering ways to remove these tactics.
The others go in order of resource cost vs utility gain.

A Time Burning tactic, like daydreaming, may not cost any time or resources to begin, but in the end costs you a lot of time for "no gain".
You haven't killed any zombies, or got yourself, your friends, or your girlfriend any closer to safety. But this tactic can still be very
useful in an over arching strategy, as it can keep you in place (in danger or out of danger) long enough for another tactic to reach its
payoff.

A Weak tactic relies on known mechanics, but only exploits the direct cause-effect chain of that mechanic. Often a weak tactic can become
less effective. These tactics tend to provide moderate gain. With the "zombies fear fire" mechanic, weak tactics would mean lighting lots
of stuff on fire, or using very simple fire implements like a torch. The cost of a weak tactic may or may not be directly apparent. At a
glance, it is often cheap, as resources are easy to find and the time to implement is close to nill. At a second look, resources aren't
always going to be abundant, and the same resources could be put to more effective use.

A Strong tactic relies on known mechanics, situation, and a bit of insight. Often these tactics never become ineffective. The can provide
moderate to very high gain. With the "zombies fear fire" mechanic, strong tactics would mean making a mote with gasoline, or making a
Molotov and lighting one of many zombies in a crowd on fire. The cost of a strong tactic may or may not be resource intensive, but
requires insight into surroundings and expenditure of time. Though, they often use a scarce resource, or one commonly sought after for a
weak tactic.

Share this post


Link to post
Share on other sites
If you want to pursue the topic on a theoretical/abstract level, I suggest reading Artificial Intelligence, A Modern Approach [amazon].

They give a good abstract approach to defining AI, and terms like rational agent, utility and strategy, and touch topics like logic, search, planning, game theory, etc.

Roughly from the top of my head, they define a rational agent being an actor who acts in a world to maximize a given outcome (measured by some performance function, "goal"). A strategy is the set of rules that defines how an agent acts in each choice situation. It could be a dictionary that tells for each game state the appropriate action to take. Under this definition, naturally each game has at least one strategy (can't really put an actor to a world where the actor couldn't make decisions at all). A rational agent follows a strategy that is rational w.r.t. the goal.

So, we can distinguish one strategy being better than another by seeing that it gives better values for the performance function than the worse strategy. This could be really easy or perhaps just impossible. Also, proving that a strategy is rational might be impossible.

Talking at a level this abstract is really stupid if you're focused on a single application, but we might consider agents in different types of worlds, having traits like
- discrete vs continuous decision-making (go vs rts)
- deterministic vs stochastic events (chess vs backgammon)
- partially vs fully observable (poker vs tic-tac-toe)

Then it makes more sense.

Share this post


Link to post
Share on other sites
Go as an example seems a bit simplistic because it does not include randomness or time pressure, both of which are huge factors in many computer games. It might also be interesting to look at the difference between games which have a single, predetermined goal, and games which have multiple or more freeform goals.

I would start analyzing this problem by making 4 lists:

- What are all the basic categories of goals that occur in gameplay, whether game-determined or player-determined? (eliminate all enemy units, gain control of all territory, obtain the maximum amount of money in the minimum amount of time, do A points of damage in time less than B while taking less than C points of damage yourself, obtain N of resource X, obtain rank/title Y...

- What are all the elements which are used by a game to keep score? (e.g. time, money, units, territory, resources, hp, mana...)

- In what ways is the game 'smart enough' to recognize the player's input? Player actions are generally limited to multiple choice and context-sensitive pressing sequences of buttons over time. Joystics may add pressure-sensitivity, special types of controllers like the wiimote can add spatial sensitivity, some games even use microphone input, but really the ways in which a computer can understand human input are ridiculously limited and account for the main differences between game strategy and real-life strategy where stuff like 'sentence wording' and 'body language' are things we apply strategy to every day.

- Finally, how can a game's internal math act upon the player's input, and to what extent can the player understand this math and use that understanding to predict the results of their actions? In games which include randomness, the player's strategizing likely has a lot of probability and statistics involved. Turn-based games are typically about algebra, especially when these games feature wearable equipment or spendable stat points. Some game designs actively discourage strategic play by hiding numbers and being sensitive to units of time or degrees of pressure smaller than the player can sense, or by not giving the player any feedback on their actions unless the player happens to satisfy a major game goal. In these cases strategy is mainly composed of the player's cultural knowledge of what is usually good and bad to do in a particular context, and their subconscious 'training' by the game. Also they can do external research and/or hacking to try to discover the numbers and goal conditions hidden within the game.

Share this post


Link to post
Share on other sites
Eh, time pressure and randomness are factors in computer games, but not terribly relevant to the intrinsic abstraction.

Time pressure adds a choice regarding what to pay attention to, and a degree of difficulty towards getting things done. Randomness just influences the game state (random maps) or the likelihood of a result. Just more data to weigh for the choice. They're just rules for the game not so much changing the definition itself (imo).

Share this post


Link to post
Share on other sites
I agree most with Telastyn, but perhaps that's because he's pretty much the only one keeping his post short enough that I read his entire argument. :)
Here's my take on it:

A strategy is an action or series of actions designed to cause a later action or later create a state.


Well, I'd be surprised if anyone thought that was *too* specific. It's borderline useless, IMO, but still has some use. For instance, here is something that almost seems to fit in the definition, but shouldn't (IMO):

>You desire to move a mouse cursor, so you move a mouse.

Why not? Because it's not really "later." A strategy would be more like this:

>In the future, I will want to move a mouse cursor, so I should plug in a mouse now.

These tasks are no more complicated (moving a mouse vs. plugging in a mouse), roughly speaking, but one has immediate results, while the other will yield benefits only in the future.

I would, for example, argue that what Wai described is not a strategy. A tactic, perhaps, but not a strategy. The goal was to scare zombies, so he screams or throws objects and the zombies get scared. His current actions are not designed to cause a future benefit, they are done to get the immediate results. On the other hand, distracting the zombies by getting a friend to scream, and then using that to run around them and grab your lunch that you left on the swing set, because it is longer term.

Where is this dividing line between short term and long term? No clue, and it definately depends on what scale we're looking at. Perhaps it is that short term effects play out before we are given another chance to make a decision (eg: the zombie's becoming scared immediately after your scream) while long term ones don't take effect until after more decision opportunities (eg: a piece placed in Go, later letting you form an eye). By that definition, a game has strategy if (and only if) its results take longer than a player's ability to react.

And on a similar note, I would like to point out the difference between tactics and strategy. Tactics implies short-term results, while strategy implies long term results. Subtle, but necessary for this conversation, IMO.

Sorry that wasn't as short as Telastyn's posts. :(

Share this post


Link to post
Share on other sites
Quote:
Original post by Telastyn
Eh, time pressure and randomness are factors in computer games, but not terribly relevant to the intrinsic abstraction.

Time pressure adds a choice regarding what to pay attention to, and a degree of difficulty towards getting things done. Randomness just influences the game state (random maps) or the likelihood of a result. Just more data to weigh for the choice. They're just rules for the game not so much changing the definition itself (imo).


You're leaving the player psychology out of it though, both time pressure and randomness can have profound effects on that, and that's the reason I said they should be considered. Specifically randomness can interfere with a player's ability to deduce the internal math of the game, possibly negating their ability to predict at all what effects their actions might have. That results in totally different types of strategies, like choosing to ignore a goal and instead do whatever type of in-game activity seems the most fun at the moment, or a strategy of behaving randomly or experimentally trying every possible action once in hopes that some surprising and interesting result will come out; quite different from normal strategic behavior of deducing the most efficient behavior and performing it consistently.

Share this post


Link to post
Share on other sites
Ezbez, I don't think the distinction between whether the result is immediate or delayed is important, both are strategy; even if the result is instantaneous you are making a decision to do that action in that instant rather than all the other possible things you might be doing. And, since games by definition last more than an instant, even instantaneous decisions will contribute to or be influenced by long-term strategic decisions about how to play the game.

Share this post


Link to post
Share on other sites
I've never seen the difference between tactics and strategy that way. Strategy tends to be the pattern of choices (from my description), tactics are more of the individual choices.

In the Zombie example, strategy would be choosing between running from the zombies, or tricking the zombies, or slaughtering the zombies. Tactics becomes where to go, how to trick them, weather to use the axe or the cricket bat...


[edit: player psychology]

Enh. The randomness of a game is known. In the internet age, one guy will find the average results either via hack or experimentation then disseminate it. Not every player will know, but you shouldn't design a game around the ignorance of your player; not anymore anyways.

And that's ignoring the inherent suckiness that comes from trying to make a decision truly without knowledge of the result. It's not an interesting choice, it's a shot in the dark. Pick A or B. Meaningless. Experimenting with various combinations to see the result is tedium. No choice; it's akin to playing a slot machine.


Some players will always pick the most efficient strategy and perform it consistently. If you want something else, vary the state; vary the goals to make the strategy variable. Make different strategies viable.

But that diverges from the original topic I think...

[Edited by - Telastyn on November 10, 2008 7:27:49 PM]

Share this post


Link to post
Share on other sites
Re: Ezbez

For the sake of analysis I draw no distinction between tactics and strategies. The difference is in scope, which is defined outside the definition of strategy.



Re: Wavinator

Evaluations is done by declaring a set of quantifiable benefits, costs, and compute the expected ratio based on probablility. In the definition where a strategy consists of a set of action/reaction laws, the number can be obtained empirically by simulating the situation n times. At the end of each run, the benefit/cost ratio is computed. At the end of n runs, the average is computed.

Analytically, the Expected Effectiveness of a strategy is hard to computed because the related probabilities could be hard to compute. But some strategies can be evaluated by cancelling out shared probabilities, allowing you to compare them without knowing their absolute effectiveness.

In the Throw-stuff-and-scream strategy, the overall effectiveness is zero because zombies are not affected by screaming or stuff thrown by kids. The list of related probability is long, and includes:

1 The chance that something can be picked up and thrown
2 The chance that you can throw stuff after i throws
3 The chance that a thrown object hits a zombie
4 The chance that a zombie gives up after getting hit
5 The chance that a zombie gives up after hearing a scream
6 The chance that you can scream again after j screams

In this analysis, 1,2,6 are ammunitions that is related to how long you could defend the playground under if is a chance that a zombie would give up.

Suppose Zombies are scared by things being thrown at them. Then an improvement to the strategy would be to tie a string to an object before it is thrown, so that the thrown object could be retrieved and be thrown again. This becomes a better strategy because it contains an option to improve the chance of 1.




The reason I am confused about the topic is that this line in particular doesn't make a lot of sense to me because the word strategy is being referred to several things:

Quote:
The elements must arise from the execution and interplay of rules which the player understands in advance. If zombies are repelled by fire, for instance, then the use of fire becomes a strategy (contingent on other rules, such rain/water putting out fire, which makes the strategy effective only in certain situations). A good strategy keeps this consistent (zombies shouldn't suddenly and inexplicably lose their fear of fire just because of a level load.)


The last sentence seems uncompatible, because it seems that you are describing a situations where the game makes zombie's fear or fire inconsistent, thus making it difficult/frustrating to a player who is trying to devise a strategy. So I can't tell whether the topic is

How, exactly, do you devise a strategy, given a situation, from scratch?

or

How, exactly, do you design a strategic situation from scratch, given that you want the player to have fun strategizing?

Share this post


Link to post
Share on other sites
No one here has heard strategy and tactics used something like "Company X is very tactical, compared to the more strategic Company Y." Company X is investing in short-term paybacks while Company Y is slowly developing large-scale projects. Or have you heard someone saying that most RTSes are actually Real-Time Tactics games?

Well, let's look at the Merriam Webster definition of strategy: (this is only the military definition)

Quote:
1 a (1): the science and art of employing the political, economic, psychological, and military forces of a nation or group of nations to afford the maximum support to adopted policies in peace or war (2): the science and art of military command exercised to meet the enemy in combat under advantageous conditions b: a variety of or instance of the use of strategy


By this definition, strategies lead to favorable conditions at the start of combat. On the other hand, let's look at tactics: (Again, the military definition. Yhe other one is too vague)

Quote:
2 : a method of employing forces in combat


Are you starting to see the difference? Maneuvering squads is strategy. Withdrawing into the castle walls is strategy.


Now, that said, I am perfectly fine if we agree upon throwing out the difference, as Wai has done. However, since most of this thread is trying to define strategy, that definately needs to be explicitly decided upon.

@SunandShadow: I certainly agree with you around the issue of my "instantaneous result" definition. I merely proposed it, since it was the best I could think of. It works fairly well, IMO, for turn-based games, but has little use in real-time games (including much of real life).

Share this post


Link to post
Share on other sites
Re: Ezbez

I see the difference where you define strategy as a set of methods to prepare before an encounter, where tactics are methods during an encounter.

An encounter may be easier to define when there is combat (e.g. when you and the enemy are in each other's range and someone starts shooting, there is an encounter). How do you define an encounter when there is no combat and no fighting?

Share this post


Link to post
Share on other sites
I’d say strategy is made up of the following:

1 – Situational Assessment – Being able to analyze the current state of the world, and having an understanding of the rules and mechanic that govern the outcome of actions in that world.

For example knowing that zombies are scared of light, light bulbs generate light, so I can turn the lights on in a room to force the zombies into dark areas.

2 - Planning - Being able to plan out future actions, and make decision before hand as to what I have, want, need, and should avoid based upon point 1.

For example if there is another kid with me at start who is willing to come with me and trade his backpack for my baseball bat. Do I want to make the trade? Keep my baseball bat? Or beat him over the head and take his backpack knowing he won’t be able to help me later if I do.

3 – Position and Environment - Position and other environmental factors have a direct and meaningful impact on the outcome of actions. Which becomes a factor in point 2.

For example I can avoid zombies by keeping too well lit areas as well as use lights to herd zombies to where the school bully is hiding.

4 – Actions and Outcome – There is a domain of actions available with a range of outcomes that will contribute to my overall objective, hinder the opponent, provided me a benefit I can utilize for future actions, or hinder myself in some way.

For example if I take the risk of venturing into the basement to restart the school generator it will me the parts of the school will have power, which increases my options later on.

Share this post


Link to post
Share on other sites
I wrote a little doc called The Strategy Game Designer's Constitution and i included this bit (likely debatable):

Quote:
In designing strategy games then, we need to constantly be promoting certain aspects of game play. Strategy games are characterized by:

  • Long-range goals (besides simply winning). All games are meant to be won. That is what makes them a game at all. But for strategy games, there ought to be long-range objectives that need to be accomplished in order to win. Things like "capture the queen" or "fortify my current position". These goals are a level higher than manipulating specific game units or variables, but do not represent the entire game itself (winning).

  • Forethought to achieve those goals. Based on our description above, the classic strategy game Chess appears to be more of a game of tactics. Each move is critical, Each piece is valued. And yet, it is widely recognized as a game of deep strategy. Good Chess players are not playing individual pieces however. They are playing an overall position on the board and think several moves ahead in order to gain an advantage. Playing each move at face value is the mark of a novice player. Therefore, strategy games emphasize forethought.

  • Minimizing the impact of specific actions when compared to the overall strategy. That is, tactics should not win a strategy game. Strategy should win a strategy game. While a game may include giving orders or moving game units specifically, the game as a whole should not emphasize any particular action. Making a good or bad "move" should not automatically determine whether or not you have won or lost the game. Some games in particular make the player feel they must restart if they don't get a good "starting position". This could indicate a design flaw (or a player obsession)

  • Relative unimportance of individual units or functions as they relate to the whole. While losing a game unit or a conflict is always bad, its impact on a true strategy game is lessened. And this isn't to say that some key units or battles or tactics are unimportant, but in the larger scheme of things, these specifics should not win or lose the entire game. For instance, again in Chess, the Queen is a very important piece. But losing the Queen does not mean the game is lost or that that player is doomed. In fact, loss of this "unit" may even be part of the strategy.


Tactical games, on the other hand, emphasize the opposite. Moves are made at face value as the situation dictates. It focuses more on specific moves or specific units which really can win or lose the game. There are few long range goals and the game may be segmented into shorter sections ("acts", "chapters", "missions", and whatnot). Strategy games are generally played start to finish which allows for more variety in between, thus not restricting the player from a full range of actions.


There's lots of other goodies in there you might want to check out.

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

Sign in to follow this