Designing Rewards in Games
August 24, 2005
Assuming we have built a game with a core activity that the player enjoys, how do we keep them playing, and what makes them stop? It's all about the rewards the player experiences, and the manner in which those rewards are structured.
What sort of rewards can the player experience? Without attempting to define a taxonomy:
- Currency rewards: the acquisition of a game resource that can be spent represents a fairly universal reward system... giving the player shops to spend currency rewards can be effective, provided there is plenty in the shops to choose from. (Note that the shop can be a 'meta-shop' - it need not be a literal shop in the game world).
- Rank Rewards: like currency rewards, but ratcheted - the player gains benefits from acquiring points towards an eventual step up in rank. The classic example is level in a class and level RPG, although in video games, Elite's (entirely cosmetic) Rank system demonstrates that a Rank reward can motivate even without mechanical benefits. A draw for Type 1 Conqueror and Type 2 Manager players if expressed in verbal terms, but if the 'Rank up' is accompanied by sufficient fanfare its appeal can be more universal.
- Mechanical Rewards: such as increases in stats that the player can feel the effect of. Highly motivating for many players - but the mechanical increases must maintain relevance to the play. Effective for Type 2 Manager and Type 1 Conqueror players in particular.
- Narrative rewards: a little narrative exposition is effective for certain players as a reward. A cut scene can be a bigger reward than dialogue - when used well. But overlong or irrelevant cut scenes quickly become devalued. Effective for Type 3 Wanderer and Type 4 Participant players in particular.
- Emotional rewards: related to the above, but applicable when the player feels they have done something for someone in the game. Animal Crossing's present giving, for instance. A draw for Type 4 Participant players.
- New Toys: anything new that can be experimented with is a 'new toy'. Although primarily a mimicry reward, there may be mechanical benefits of well - a new weapon in an FPS is a new toy with mechanical rewards, for instance. Especially of value to Type 3 Wanderer players.
- New Places: like new toys, new places are a mimicry reward for players driven to explore (a common drive!). Especially of value to Type 3 Wanderer and Type 1 Conqueror players.
- Completeness: perhaps only a drive for the Type 1 Conqueror player (or the Rational player), achieving completeness (chasing 100% for instance) can be a reward in itself.
- Victory: defeating a challenging foe (or a boss) is purely agonistic reward, especially appealing to Type 1 Conqueror players.
The other aspect of how we design rewards is the way the delivery of the reward is structured. Psychologists call these structures contingencies or schedules. I first learned about this from John Hopson's excellent (but nervously delivered) GDC talk Behavioural Game Design, which is also available as an article at Gamasutra. Briefly, psychologists have observed a number of different reward schedules in animals (humans included):
- Fixed Ratio Schedules: these provide rewards after a fixed number of actions. They produce a high level of activity and are easy to understand, but after the reward is achieved, there is a pause. XP in CRPGs is an example - although the gearing of XP systems is exponential, the intent is that the player is constantly moving up to tougher foes, thus keeping a constant ratio of kills to level. Hugely addictive to Type 1 Conqueror players, they can work for any play style - if the rewards are right.
- Variable Ratio Schedules: these provide rewards after a random number of actions - like a slot machine. You keep putting in coins, because at some point it will pay out. These also produce a high rate of activity and interest, but they tend to block exploration - as the player will stick with the reward schedule until it is exhausted, or until they burn out on it. Effective with all play styles - but burnout is always a risk. They are inherently aleatory, and may appeal less to Type 1 Conqueror and Type 4 Participant players.
- Fixed Interval Schedule: that is, a reward is provided after a set amount of time. This provides better control over the rate of reward, and comes with the same post-reward pause as a fixed ratio schedule. Indeed, pauses are inherent to fixed schedules of all kinds. An example is the new items in the shop each day in Animal Crossing - the player comes back on future days to see what's new.
- Variable Interval Schedule: like the variable ratio schedule, this produces a steady rate of activity with no pauses - but its not as intense as the variable ratio schedule, because players quickly learn that their actions are independent of the reward. Good for encouraging a player to come back to certain places in a game, however, if a reward appears in certain places 'at random'. Again, they are aleatory and hence may not appeal to all Type 1 or Type 4 players.
In general, ratio schedules produce high rates of activity - "the more you do, the more you get". Variable schedules produce constant activity - "everything has a chance of reward". When these combine, (variable ratio schedule), the player will eventually burn out. Conversely, fixed schedules create a pause - which needn't be a negative matter. To keep a player's interest in a CRPG, the 'pause' after gaining a level frees the player up from the treadmill of leveling up to go and carry out other housekeeping activities in the game. (If the player levelled up with a variable ratio schedule, they could rapidly get burned out).
These elements - types of reward, and the schedules upon which they are delivered - form a framework which maintains a player's interest in the game they are playing. The more complex the game, the more different rewards and schedules for the delivery of those rewards are required to keep the player involved. A simple game can be built upon a single schedule.
All of this comes to a head in how and why the player stops playing. Pauses allow and encourage quitting - players are constantly evaluating the very next thing they can do in a game, and if their level of interest drops below the draw for another activity, they stop. If this happens through burnout, they may not go back - the gamble with variable ratio contingencies - but if they stopped because of the pause after achieving a sufficiently large reward in a fixed schedule, they will likely come back. In this regard, one must be careful that the rewards themselves maintain at the very least a constant (and at most an escalating) level of reward.
Of course, the best case is that just playing the game is inherently enjoyable to the player - that the core play is its own reward (when the core play is a flow state for the player). Still, even when this is so, the player is likely to stop playing when 'they have seen everything'. This is when multiplayer elements can extend the play window of a game, of course, by providing new rewards - provided the player happens to be motivated by agon (like many Type 1 and Type 2 players).
If one creates a game which is inherently fun for the player, an exponentially structured fixed ratio schedule can be sufficient framework to keep playing - such as the monuments in The New Tetris (which are wholly cosmetic, and equivalent to Rank rewards). For some reason, this structure seems to work better than other schedules for a high level framework. Interval schedules lack the connection with player action, and variable schedules only work until the rewards are exhausted. But diehard players of The New Tetris routinely reset the monuments and start over again, with little loss of interest. Perhaps it is the exponential gearing which drives the appeal, pulling the player forward by gradually increasing the jump to the next reward.
If you have ever wondered why games with poor game mechanics can still entertain players, it is perhaps because many poorly designed games are at least easy to play - and if they provide rewards according to a reliable schedule, they will entertain. As long as they keep providing rewards. Good game mechanics can aid a game by eliminating rough edges and inconsistencies, and some players (those who fit the template of our H1 and H2 clusters in particular; those with access to the Rational temperament) are actively drawn to elegant game mechanics - but it is the delivery of rewards, and not the quality of the game mechanics, which maintains a player's interest.
You've finally helped me understand why i enjoyed Deus Ex 2!
The majority gameplay was horrible, yet the story made me keep playing. Not because it was a terribly good story (it wasn't - the plot may have been but the narrative was broken by the non-linear nature of the game), but rather because enjoyed the 'narrative rewards'.
I wanted to see it through in order to complete the story (being a huge fan of DX1 i had to see how it all fit together), and try out the different branches.
I have always know this about the game, but until now I have not fully understood it or thought to attempt relating it to a specific game design mechanic such as rewards - perhaps because I have never considered the notion that narrative could be a reward before.
Of course the enjoyment only lasted until I exhausted this, thus confirming what you've said!
Posted by: Ben | November 02, 2005 at 04:02 AM
I have written my own piece on this phenomenon as well. It can be found here:
My piece concentrates no what you hint at towards the end; games which use rewards as a crutch to keep the player going through what would otherwise be poor game mechanics. Tons of RPG's work on this principle and are viewed by fans as "deep" games, despite the fact that the gameplay consists of carrying out mundane menu navigation devoid of thought or strategy for hours at a time in order to build currency in the form of XP points or items.
I personally wish someone would outlaw these mechanics for some of these developers, so that they would be forced to actually create legitimately good games for a change.
Posted by: GnaM | August 06, 2007 at 04:43 AM
Well, my link got cut off, but if you go to my blog you can find it under the "top 10 old favorites" category, titled:
"(False) Achievement-Driven Gameplay"
You may find it interesting. Sorry if it seems like spam, just trying to contribute to the discussion.
Posted by: GnaM | August 06, 2007 at 04:48 AM
GnaM: thanks for sharing your perspective; it doesn't look like Spam at all - it's completely on topic. :)
Posted by: Chris | August 06, 2007 at 05:32 PM
Hello I read your article and I enjoyed it thoroughly. I just wanted to ask what do you mean by all the type of players that you talk about. I would like more information on that subject if you could help me with that it would help me out a lot. Im new to game design and just beginning my path to making games. -Danny wilson - [email protected]
Posted by: Dannysjobs1986 | November 27, 2012 at 11:03 PM
Danny: the types mentioned here are from the DGD1 model described in "21st Century Game Design". Briefly:
Type 1 = Conqueror, players who enjoy striving against tough challenges.
Type 2 = Manager, players who enjoy solving problems.
Type 3 = Wanderer, players with an experiential attitude towards play.
Type 4 = Participant, players who enjoy the social aspects of play.
However, the DGD1 is now very old! My latest model is BrainHex which you can find at www.brainhex.com
Best of luck with your game design!
Posted by: Chris | November 28, 2012 at 10:00 AM
Seven years later, and I dare to make a comment. :) I miss one reward in your list and that's feedback. You know, when you push a button and a lamp is glowing as an answer.
I would like to take this a step further and talk about feedback loops, where one thing reacts on another thing that created a reaction something that the one thing started; where everything loops back.
Feedback loops are really important in tabletop roleplaying games, where your character's action will create consequences. This is something I really miss in computer roleplaying games, where what I do will perhaps create an outcome in the end, but it will never come back and haunt me.
I want to see the world's response of my actions.
Posted by: Rickard Elimää | January 24, 2013 at 08:02 PM
Rickard: no problem with the late comment - happy to resurrect dead conversations and pick through their bones! :)
Feedback is an important general point that was not on my mind when I wrote this, but since Jane McGonigal laid out her definition in terms of it I have been much more willing to see all of the above as forms of feedback i.e. to see reward systems and feedback as intimately connected. The above list is far from complete.
What you're calling a 'feedback loop' sounds more like a 'feedback snowball' to me - since it's less about the loop and more about the compounding of consequences of action... Quite hard to do this in digital games.
Interesting to make reference to tabletop RPGs, which of course are the kings and queens of agency and consequence. No videogame will ever match them on either front, because they are not mediated solely by imagination. And even if it were possible to produce AI of sufficient imagination (and it may not be), we would only be turning digital games into tabletop RPGs with better props than lead miniatures for representation. :)
This desire to see the response of actions is what the issue of agency is all about - and it's so much harder to make this work in a digital game than around the table. But that makes it all the more satisfying when a game is able to deliver that extensive agency, or at least the illusion of it.
All the best!
Posted by: Chris | January 25, 2013 at 10:39 AM
Yes, you're totally right in that all rewards are nothing more than feedback.
Feedback loops (improv term) or reinforcement and mirroring, as it's also been called in freeform circles, shouldn't be that hard to create in digital media.
All structural story design that I've come over in digital media have been built up through a tree-like structure. I wonder how hard it would be to instead build a relationship map with reactive relations and active agendas, where a response is triggered when the relations are dabbled with. I've written an article about the fishtank model on how to build mysteries in TTRPGs. I don't think it would be hard to design a computer game in the same way.
Posted by: Rickard Elimää | January 27, 2013 at 10:31 PM
Rickard: the tree structure is actually a terrible way to build games because of the combinatorial explosion (the increasingly diverse sets of possible outcomes) so almost all trees use a narrative that is recombinant - more like a helix than a tree, really.
When I was doing ambitious narrative design I liked a threaded system that compartmentalised the story content into sequences of possible events (like a sub-plot in a TV show) but allowed the player to thread these up in whatever order made sense. I would have liked to perfect this method.
Dynamic narrative tools is an interesting research area but one that is just not an option for major game projects which are necessarily risk averse and cannot afford to be inventive in areas like this that are perceived as secondary to commercial appeal (correctly, I made add, however sad that might be).
If these were to be explored, it would be in indie games - and I would suggest the way to begin would be in very small experiments. There's a reason Facade could handle so much variety of response, and that's because it was only meant to play for 30 minutes. :)
All the best!
Posted by: Chris | January 29, 2013 at 09:49 AM