Interesting Times – Modelling Time in Multi-Player …...analysing RPGs across different formats....

Vol 16 Issue 2 – 3

of 12 Leonardo Electronic Almanachttp://leoalmanac.org/ Vol 16 Issue 2 – 3

Last updated on 28 September 2008

1

1}

Interesting Times – Modelling Time in Multi-Playerand

Massively Multi-player Role Playing GamesAnders Tychsen & Michael Hitchens

Interactive Systems and Virtual Reality Research Group, Macquarie University,Sydney, Australia {atychsen, [email protected]}

AbstractTime is a key concept in the design, playing and studyof games and can be viewed from multipleperspectives, e.g. the player and the game world. Herea comprehensive, iteratively developed model of gametime, based on empirical games research as well asrecent theory, is presented. The model is tested inpractice and its applicability across tabletop, digitaland other forms and formats is demonstrated. Specialemphasis is placed on multi-player and massivelymulti-player games, as well as role playing games, asthese feature complex game time behavior that cannotbe explored in existing models of game time. Themodel includes seven viewpoints of game time, andallows for the mapping of time as an interactivelycreated and non-linear feature of games and gameplay.

1.0 INTRODUCTIONComputer games are inherently linked with theconcept of time. It takes time to play a game, time hasa definite place in our understanding of the gamingactivity and time is an important factor in the design ofcomputer- as well as tabletop games. From calculatingreload times of weapons in FPS games to achievingperfect balance, level and story design, pacing of play,timing of events and triggers, time is a major elementin the development and design of games [e.g. 2,4,17].Time is furthermore a key consideration to address ingames research, as it is one means of understandingwhat happens in a computer game, how the game andthe player(s) interact, as well as how the game isexperienced by the player(s).

While an important aspect of play and design, therehave been few contributions towards developing acomprehensive theory of time in computer games[10,11]. What efforts have been made draw from avariety of sources, including semiotics [13], physics[15], literary narrative models [8], and user-gameinteraction [9,10,11]. In contrast, time is a subject wellstudied in other academic areas of study, such as film,literature and theater. Given the importance of time incritical understanding in these areas and its obviousimportance in relation to games the need for a clearer

understanding of time for the study of game should beobvious.

In creating a theory of game time, it is important torealise that there are differences between time ingames involving one player, and those involving morethan one player. In a single-player game, only twoentities are involved, the player and the game. In amulti-player game, the actions of one player can affectthe game playing experience and actions of the others,or players can be engaged in different activities whilethe game operates. This creates a complexity that amodel of time from a single player perspective will notencompass [9]. Most of the contributions mentionedabove pay scant, if any, attention to the inclusion ofmultiple players in their models.

One of the most popular categories of multiplayergame is Role Playing Games (RPGs). RPGs span ahuge range of formats, from tabletop, virtual reality,mobile-based and online. Despite the commondenominator – role-playing – there is an incrediblylarge variety of formats employed, and it is thereforechallenging to describe and analyse these games in thecontext of a coherent framework.

As a shared – if highly variable – feature of all gameforms, game time offers a venue for studying andanalysing RPGs across different formats. A coherenttheory and model of game time would for example beuseful in modelling player interactions in RPGs, andthe collaborative storytelling process of Pen and PaperRPGs (PnPs).

RPGs often demonstrate very intricate temporalbehaviours. Many forms of RPG present players witha high degree of freedom, allowing them to operateindependently, and affect the game story. Furthermore,RPGs are based on detailing the activities of playercharacters (avatar-characters in computer games) –activities that do not need to follow the chronologicalflow of time in the fictional worlds Player charactersmay exist in different instances of the game worldchronology, or even form personal perceptions of timethat operate completely or partially in the mind of theplayer. Briefly, multiple players and their interactionscreate a comparatively larger degree of complexity




2

2}

when compared to single player games. This meansthat a model of time developed for such games, whilenot necessarily universal, should form a basis for awider understanding of the nature of time in games.

The purpose of this paper is to provide a model of howtime operates and flows in multi-player (MP) andmassively multiplayer (MMP) games, and apply it tothe four major formats of RPGs – the traditionaltabletop PnPs, Computer-based RPGs (CRPGs),Massively Multiplayer Online RPGs (MMORPGs)and the physically embodied formats collectivelyreferred to as Live Action RPGs (LARPs). Focus ishere on digital MP and MMP games, notably RPGs,however, the model is equally applicable to non-digital games and single-player games, and severalexamples from e.g. tabletop games have been includedin the model description.

The model presented expands the single-player modelpresented by [9], which is itself an extension of[10,11], into the MP and MMP range. The model isnot intended to be the final word on game time,however it benefits from being scaleable and adaptableto specific purposes. For example, the use ofsegmentation of temporal activities integrates theinteractive nature of games and allows different levelsof granularity in the analysis of time in games to beutilised. While previous models are based on theoryand/or observations of games, the model presentedhere is derived from theory, play testing and empiricalresearch. Multi-player PnP and CRPG experimentswere employed in order to test and refine the gametime model and provide a test bed for the applicationof the model to RPGs in practice.

2.0 PREVIOUS WORKDiscussion of time and its related issues has featuredin numerous computer game design books as well asgames studies publications [e.g. 1,2,4,13,14,16,17]Games have even been classified based on their use oftime, e.g. real-time strategy, turn-based games. Timeis integrated in game design terminology withincomputer games as well as other game formats, e.g.the concept of game speed, and the utilisation of directtime control (e.g. the rewind function of Prince ofPersia: The Sands of Time, or bullet time as used inMax Payne). However, there have been very fewdedicated studies of time in computer games from theresearch community.

[8] Adapted concepts from narrative theory for thedescription of temporally related phenomena in games,noting that events can be described in terms of theirduration, ordering, speed, frequency, simultaneity andthe time of action. [8] also noted that the dominanttemporal relation in computer games is between thetime of the user and the events of the game. This in

opposition to the temporal relation observed intraditional narratives between the discourse time (thetime of the telling of the story) and the story time (thetime of the events narrated).

[15] Discussed time from a physics point of view, e.g.separating relative and reversible time. [5] Consideredtime in computer games from an aesthetic perspective.

[13] Outlined a semiotic approach based in literaturestudies. [13] Created a framework for ludic semioticsbased on the recognition of different temporal semioticsystems: simulation, game and narrative. As thisapproach is focused on structuring events whendesigning a game, rather than the actual experience oftime when playing a game, it cannot be directlycompared with the approach adopted here.

[10,11] (Figure 1) presented the possibly first modelfor game time from the viewpoint of the playingexperience, rather than game design. That modelincorporates time in the form of layers: Play Time –the actual real-world time spent playing - and EventTime – the time of the fictional world of the game -and describes the ways that the game and the playercan interact via these two layers. The model allows themapping of time between the two layers, i.e. projectingthe play time and the actions of the player onto theevent time layer. This approach allows a description ofgame speed (the relation between play time and eventtime). [10,11] describes game features such as cutscenes, loading and saving as specific effects that canoccur in the play time-event time relationship.

Figure 1: Mapping player actions between two layersof time on a 1:1 basis (after [10]).

[9] Adopted the approach outlined by [10,11] (Figure2), refining the model to include non-linearity [1],expressed as a branching of certain time layers torepresent the ability of players to choose betweendifferent paths, and to abandon their current path andreturn to an earlier point in the game. Non-linearity inthe model allows the integration of the potentiallycomplex relationships between the player’s experienceof time in the real world and progress through thegame. For example, by reloading a previously savedgame, the player is effectively turning back the eventtime of [10]. Non-linearity is relevant to all gamesbecause all games feature choice. This even includes




3

3}

PnPs, as those games permit the rolling back/forwardof the game story, should the participants agree.

The model of [9] also introduces multiple modes ofengagement, by dividing the temporal layers up intosegments and breaks (Figure 2). Segments represent asection of time of the relevant temporal layer, withinwhich the activity is homogenous - for example,waiting while a module is being loaded in the CRPGNeverwinter Nights. Breaks represent changes in theactivity, and have no temporal dimension, being

instantaneous. For example, starting to play (interactwith the game) after the loading of the NeverwinterNights module finishes. Breaks should not be confusedwith periods of inactivity, i.e. a period where theplayer is not playing the game. These have a non-zerodistribution in the relevant temporal layer.

The model of [9] has four time layers:

1) Playing Time: The objective real-worldchronological time experienced by a player during andbetween game play sessions.2) Engine Time: The objective chronological realworld time in which a game engine executes.3) Progress Time: An abstract (non-chronological)measure of time tracking movement towards gamecompletion and allowing events to be related in termsof happens-before and happens-after. This layer allowsnon-linearity to be described and modeled in terms of

the player’s achievement of thegame-play aims.4) World Time: Chronological orabstract and logical (dependingon the game) time within thegame world. While this iscomparable with the event timelayer of [10,11], the latter doesnot include the non-linearityincluded in the view of the gameworld time.Like the model of [10,11],activities within one layer can bemapped to another layer. The firsttwo layers are linear, while thesecond two represent views oftime that can be non-linear(Figure 2).

Figure 2: The four layered modelof [9]. Segments are separatedwith a vertical line. The dottedlines represent mapping ofsegments between layers.Branching occurs in the twolower layers when a saved gameis re-loaded (dotted arrows) andplay continues along a differentpath (stippled arrows). Note thatwhile time passes in Playing Timeand Engine Time (marked as asegment), the reloading isinstantaneous in Game ProgressTime and Game World Time(marked as a break).

The model presented hererepresents a refinement of the

[9,10,11] models to include multiple players,incorporating further theory and empirically deriveddata, as well as extending the model to include thecomplex temporal relationships of multi-player RPGs.

3.0 METHODA core aim in developing the model for game timepresented here was to ensure that it was directlyapplicable to the study of games, rather than simply an




4

4}

abstract theory. An empirical approach providessupport for the theory behind the model anddemonstrates its applicability by evaluating its abilityto function in a real world context, as well asproviding the hard data needed to analyse how gametime operates in RPGs.

Earlier work on game time – both within gamesresearch and design – has shown that game time is avaried feature. Prior work on RPGs [7,17,18,19,20,21]indicates that time in these games can form complexconstructs, with e.g. player characters operating atdifferent intervals of the chronological time of thefictional game worlds that these games are set in. Inorder to develop a model that is not only applicablebut also broad enough to encompass the differentfeatures and viewpoints of game time, theoreticalmodels were combined with a thorough, case-basedstudy in an empirical framework (experimentation andobservation).

The empirical part of this study was focused on PnPsand CRPGs. PnP and CRPGs include complexmanipulations of game time, due to the high degree ofplayer freedom as well as the unrestrictedcommunication between the participants [7,21]. Thiswas combined with extensive MMORPG play testingto supplement the available research and designliterature, notably with World of Warcraft; in order toassess the applicability of the model to MMORPGs,and to expand the description of time in these games.The practical difficulties in observing the largenumber of participants over a large geographical areaprecluded LARPS from being included in theexperiments, however, all available theory on LARPswas included in the study [e.g. 3].

The model was developed and refined over the courseof two rounds of empirical experiments. A basicassumption is that the experimental situation isrepresentative for PnP and CRPG games. This ispotentially problematic, because the variation in PnPand CRPGs [7,20,21], makes establishingexperimental conditions that are standardized difficultat best. This assumption is alleviated by combining theempirical experiments with literature studies, whichserves to locate any game format variations that wouldimpact on the game time model. Furthermore, byusing time layers as the foundational building block ofthe model, it becomes resistant to error introduced byminor game design variations (for example, thenumber of players or distribution of directorialcontrol). For example, whether the PnP in questionutilises a GM or not does not impact on the modelframework, only on the application of the model to thespecific game situation.

The CRPG Neverwinter Nights used in theexperiments is a fairly typical example of the CRPGgenre, and examination/playtesting of a range of otherCRPGs such as the Might and Magic series, Baldur’sGate, Sacred, Diablo and Summoner revealed noissues in relation to the model that needed to beintegrated. Undoubtedly, with the huge variation ofgames, there are aspects that have not been captured inthe current model, and it must be stressed that themodel presented here is not meant to be the finalworld on game time.

The PnP and CRPG experiment setups were designedto provide as similar conditions as possible. This is inorder to limit bias caused by differences in story, rules,players etc. between the two game formats. During thepilot test, one multi-player PnP and one multi-playerCRPG game session was run. The PnP sessionconsisted of five players and one GM, while theCRPG session consisted of the same playersinteracting via and with the game engine. The samegroup of players played both game sessions. Themodules being played were based on the same widelyused rules system; the D20 system originally designedfor the Dungeons & Dragons PnP, version 3.5 ofwhich forms the rules basis of the CRPG used,Neverwinter Nights. Each player had his/her owncomputer monitor.

In order to limit the number of variables involved, thetwo modules both featured story themes of reversaland revenge. In both game sessions, the players weresituated around a table with full verbal and visualcommunication access. The game sessions wererecorded on video. For the CRPG, a client tool bundledwith Neverwinter Nights was utilised to capture screen-based actions. For both game sessions, selectedsegments of the player communication was transcribedand coded, and for the CRPG a game log extractedshowing all text-based chat as well as the use of emotes.The players were asked a series of questions beforeand after the game sessions. This was both to locateany tensions in the groups that could bias results, toobtain information about how players relate to theirfictional characters in the different layers of timeproposed in the models of [9,10,11].

The 2nd round of experiments consisted of a series of10 PnP and 10 CRPG sessions, each involving 5players (plus 1 GM in the PnP sessions); runningbetween 3-7 hours. All game participants were adults(18-54 years of age, 28.8 average, 27 median). Thegroups carried over between the two game setups.

The recordings, questionnaires and game logs formedthe basis for iteratively revising the game time model,and if the model encompassed the full variety ofplayer actions. A substantial amount of revision was




5

5}

necessary following the experiments, for example, theneed to separate game progress time into two distinctlayers (game progress and -story time), was realizedafter the first round of experiments. The experimentsalso gave rise to the boundary between the EngineTime and Server Time layers being redefined (seebelow); and gave rise to the notion of perceived time.Finally, examples of the different time layers inoperation were located to serve as documentation.

4.0 KEY CONCEPTSBefore describing the current game time model, a fewkey concepts need to be introduced and defined. Theseconcepts are used to describe how game time operatesand how players interact with game time.

Player state: The player state describes the situation ofthe player for a given segment of time along any of theseven time layers. Three general states are defined:

Active: The player is actively playing the game,interacting with it and/or the other players (Figure 3).Passive: The player is playing the game but iscurrently in an observatory mode, i.e. not activelyinvolved in the game play. For example, a goal keeperin a soccer match when the ball is in the other half ofthe field, a player in a PnP who is watching the GMand two other players acting out a scene; or auto-crafting in MMORPGs.Inactive: The player is not playing the game, forexample, during the real-world time between a saveand exit; or during the start and load of a computergame. During inactive periods, neither the player norany avatars are active in the game.

The identification of player states is not strictlynecessary in order to model time flow in MP andMMP games, however, it is of relevance whenconsidering how multiple player can affect the gametime of each other, and how the game utilises time.This is of special interest to PnPs, where the playersconstantly change between active and passive statesdepending on whether or not their characters areactively involved in a given interaction.

Figure 3: An example of the Playing Time of a three-player game, where the players change states. Note

the period of inactivity of Player 1. The game howevercontinues because the other two players are active.

Segment granularity: By dividing the layers of gametime into segments of a defined duration, separated bybreaks, a useful tool for mapping different types ofplayer activities is gained. However, it is obviouslypossible to use different levels of granularity whendefining the segments of a game time layer (Figure 4).For example, while actively playing a game, a timelinesegment can be labelled simply ‘interaction’.However, in a more detailed study of the playeractivity it might be of relevance to apply a finer grain.Irrespective, the level of detail must be tailored to thespecific game analysis, with sections labelledaccording to the specific task the player is performingwithin the segment (e.g. actions such as loading andsaving can occur in different ways and have varyingimpact, e.g. conditional saving in Animal Crossing orserver save/load ‘roll-backs’ in MMOGs).When applying the model presented here to therecorded game sessions, it was observed that ingeneral, the finer the grain of the segmentationapplied, the more variation there was in the activity ofthe different players within each segment. Note that inthe models of [9,10,11], individual time lines canpotentially represent any number of entities. Whenconsidering interactions in MP and MMP games, thevarious time lines can be used as representations ofspecific players or groups of players (Figure 5). In thiscase, one time line represents the specific group orguild of players.

Figure 4: Example of how the activities of players canvary within the same Playing Time segments inmultiplayer games. Note the two levels of segmentgranularity in the centre of the two time lines (the‘interaction’ segment is broken down into segments ofa smaller granularity).

Game Speed: [10,11] defined the ratio of timebetween his play time and event time layers as thegame speed. While this is a useful definition whenapplying his dual layer model to game study, thecurrent model utilizes additional layers, meaning thatdefining game speed based on only two layersbecomes impractical. The term relative game speed is




6

6}

therefore utilized here in order to denote the ratiobetween the time flow of two specific time layers. Forexample, the relative speed (or ratio of time flow – theprogression of time within a given framework -difference) between Playing Time and World Time ina CRPG like Neverwinter Nights is 1:12 – NeverwinterNights operate on a two-hour day cycle determined bythe active Playing Time.

5.0 THE SEVEN LAYER MODELThe model of time presented here is an extension ofthat presented in [9]. That model is extended here toaccount for the presence of multiple players and theeffect that has on the nature and perception of time inmulti-player games. The presence of multiple playersmeans that the individual player not only interactswith the game, however usually also with otherplayers. This impact might be limited to specific layersof time or effect all of them and the magnitude of thiseffect can vary tremendously between different MPand MMP games.

The model includes seven layers or viewpoints oftime. Four of these are adapted from [9]: playing time,engine time, progress time and world time. Asubstantial amount of new detail is added toaccommodate the multiplayer situation, as described inthe following sections. The three new layers (servertime, story time and perceived time) are introduced toaccommodate the presence of multiple players. Thesenew layers are necessary in order to describe gametime operation and the way multiple players caninteract with each other and the games they play. Notethat not all seven layers are applicable to all forms ofgames, e.g. PnPs and other table top games do nothave Server Time and Engine Time layers, as these arebound to the hardware and software of computergaming. The model is intended as a toolkit for theunderstanding of games and it is expected that not alllayers would be necessary in all applications

5.1 Playing TimePlaying Time (Figure 4) is a conceptually straightforward measure of game time from the viewpoint ofthe player. It is defined as the objective real worldchronological time experienced by a player during andbetween game play sessions. Playing Time has aspecified beginning and end, between which gameplaying occurs. The playing time includes all the timethe player interacts with or otherwise participates inthe gaming activity within the magic circle of [17].The core of the time spent playing MPGs and MMPGsis taken up by performing game related actions andinteracting with the other players, via various inputdevices – verbal, game pieces, human-computerinteraction devices such as mouse and keyboard, etc.Via these actions the players affect the game state,

progressing through the game. Each player (or group)would have their own separate representation ofplaying time when modelling a game.

The presence of multiple players means that multipledifferent actions can be taken in the game world at thesame time by the different players. For example, aplayer can choose to attack a MOB while anotherplayer opens a treasure chest. This means that within agiven time segment, players can perform differentactivities. Some of these, for example bartering ornegotiating, can temporarily align the activity of theplayers, as seen in Figure 4.

5.2 Engine TimeEngine Time considers time from the perspective ofthe actual game engine or software running the gameapplication, and can be used to map e.g. loading,saving, briefing and interaction segments. It forms incomputer games one of the two aspects that the playercan interact with – the second being the other players.

This time layer is defined by the objective, linear andchronological real-world time in which the clientsoftware part of the game engine executes, and can befairly similar to Playing Time, however, there aresubtle differences (Figure 2). For example, in EngineTime there is no measure of time flow when the gameengine is not operating, as the software is notfunctioning during that time (Figure 2, the break in thePlaying Time does not register in the Engine Timelayer). In multi-player games engine time refers totime as experienced by the client software for aparticular player. The client software for each playerin a multi-player game will experience time and eventsin a potentially different fashion, due to the nature ofdistributed computing systems.

Engine time is also used for the software in multi-player (“hot-seated” games that run on a singlecomputer). The concept can also be applied to boardgames such as Monopoly, Axis & Allies and Chess, orPnP games. In these cases Engine Time represents theviewpoint of the game itself, and is mapped to PlayingTime in a similar fashion as for computer games.

5.3 Server TimeFor single player games the game software is normallylocated on a single machine and time from its point ofview can be represented via engine time. Multi-playergames are often implemented using the client/serverparadigm. Each player has a copy of the clientsoftware on their machine, with a single server(possibly on the same machine as one of the clients)providing central control. The server has its ownviewpoint on time, which is normally different (atleast in detail) from that of any of the clients. This isespecially true in MMORPGs, where the server can




7

7}

remain running for extended periods, even if no clientsare operating.

Server Time (Figure 5) is here defined as thechronological, real-world time in which a game serverhosting the server-side software executes. Software onseparate machines does not share the same view oftime [12] and this requires a separate layer of time inmulti-player games. In modelling multi-player gamesnot implemented via client/server software (e.g., aspeer-to-peer software) this layer is not required,although such architectures are very rare in gamessoftware and the remainder of this section refers togames implemented using the client-serverarchitecture. Server Time is useful when mappingdifferences in player activities in the Playing Timelayer to the game itself in a MP or MMP situation.Because each player has their own Engine Time layer,the Server Time layer allows the activities of theplayers to be reconciled. In the case of MP games suchas Counterstrike, where the computer of one playercan acts as both the server for the game as well ashosting the game engine for that player, the gameengine time of that player is mapped directly to theserver time. In general, precise synchronizationbetween computers is nigh impossible due to networklag and processing time, which causes variations inwhen each instance of a digital game receivesinformation about events in the others. Thisimprecision can be important a game analysis,depending on the segment granularity employed.

In persistent-world games, the Server Time layer isalmost continually active, save for occasional downtimes for maintenance or similar issues. Players arefree to log on and off as they choose, leading to acomplex mapping between player and server time(Figure 5). While in a single-player computer gamethe Engine Time maps to part of the Playing Time(Figure 2), the situation in a MMORPG is reversed.Even when players are inactive, the server continues,even if the player avatar is effectively removed fromthe game (some MMORPGs allow certain repetitiveactivities such as crafting or transportation without theplayer needing to actively interact with the game). Thegap in Playing Time is therefore mapped to a segmentin Server Time, as opposed to a break in Engine Time.

In summary, for each MP or MMP there will be oneServer Time layer, mapping to multiple Playing Timeand Engine Time layers. Note that e.g. MMORPGscan have multiple shards, or copies of the gamerunning simultaneous, however these can be viewed asseparate games or systems.

5.4 Progress TimeWhile Playing, Engine and Server times are all linearand chronological, games can provide a non-linear

Figure 5: Multiple players of a computer game with apersistent server. The top two timelines are PlayingTimes for Player 1 and 2, while the bottom timelinerepresents time as viewed from the perspective of thepersistent server, i.e. the Server Time.

experience [1,9,17]. Non-linear can have multiplemeanings in a games context, one being that there ismore than one possible path from the beginning to theend of the game, referring to the topology of the nodeswhich make up the game structure. A game which hasmultiple paths to completion is nonlinear in this senseand the choices of the player when navigating thegame impacts on the experience of the game and gameworld. For example, in a game of Chess, a player willusually have more than one option in making the nextmove.

Considering game time from the point of view of theplayer, rather than the game (world), non-linear canalso refer to the ability of players in most digitalgames to choose between paths, and later abandon orrevisit paths. For example, when reloading a savedcomputer game and replaying a specific section, thusexperiencing the same content. The replayed sectionwill generally be different from the originally played(Figure 2), as the player has knowledge of the pathbeing replayed. This therefore leads to a branching inthe game time (Figure 2, 6), as a similar amount ofgame World Time is experienced, but at a differentpoint in Playing Time and normally with differentplayer choices and outcomes, although as noted by [9],mistakes can be repeated. Branching can also occurwhen progress can be lost without reload, for examplein strategy board games such as Twilight Emperiumand Settlers of Catarn, as well as many card games,the goal of players is to reach a certain amount ofpoints. These are scored by reaching specificconditions, such as controlling a specific planet inTwilight Emperium. In such games, players canhowever loose points already gained. In terms of gameprogress, this type of setback reduces the player to aprevious checkpoint (one with fewer points), causing abranch to occur.




8

8}

Progress time is a logical measure of a player’sprogress through the game. It can be measuredobjectively by mapping the specific game events fromthe beginning to the end of the game, and organizingthem into logical relations of happens-before andhappens-after, for example, by using a series of gameobjectives or checkpoints in the game story. Progresstime can be measured based on different types ofcriteria, depending on the game in question. Twoprimary forms of progress observed in the empiricalexperiments for this study mechanic progress andtask progress. Progress time allows the mapping ofbranches caused by e.g. a save-reload cycle (Figure 2).While different branches in progress time canrepresent similar intervals of world time, they have noeffect on the future of game except for that the playerhas learned about the replayed section of the game.Also note that in a multi-player game there will be agame progress time measure for each player.Mechanic progress changes the game state in terms ofthe rules, e.g. acquiring a character level in aMMORPG such as Ultima Online or the accumulationof money in Monopoly. Progress in these terms can belost via a reload, and non-linearity introduced. Intabletop games this is possible where the basis of themeasurement can be lost and regained. Task progressis related to the requirement of the players having tocomplete certain tasks (objectives, quests, etc.) toadvance in a game. Examples of this include the levelsof games such as the classical Pac Man and AtomicBomberman. Note that some games (e.g. many PnPs)have no defined completion conditions; however theprogress of players from one point in the game toanother can still be readily observed. This approach isuseful in games that do not enable players to revisitpreviously explored paths, i.e. most non-digital games,in which Progress Time would otherwise be linear andpossibly chronological.

5.5 Story TimeStory Time is time of the actual dramatic story of thegame, if present. Story Time is chronological if thegame World Time is chronological, and otherwiselogical. However, most games without chronologicalworld time have limited story, so Story Time is mostuseful as a chronological measure. Following the samegeneral principles as Progress Time, Story Time canbe non-linear (Figure 7). Story Time tracks themovement towards game completion according to thegame story, allowing events to be organized in ordersother than the strict chronology of the game world(eg., via foreshadowing or flashbacks). The term“game story” is here used in a loose sense, as theconcept of game narratives is a contested issue[11,13]. However, irrespectively of whether or notgame-based stories or plotlines fulfil the formal

criteria for being narratives in the traditional literarysense [11], stories form an important component inmany genres of games, notably computer game genressuch as adventure games (e.g. Beyond Good and Evil,Gabriel Knight, Indiana Jones and the InfernalMachine), FPS games (e.g. Deus Ex, Chronicles ofRiddick, Half-Life), and especially in all forms ofRPGs.

In Story Time is included most of the narrative toolsused in storytelling such as flashbacks, (direct)foreshadowing and visions, contracted time (e.g.Civilization and similar turn-based games) or extendedtime (e.g. the bullet time of Max Payne), and the useof these time-based narrative tools impact on therelationship between time layers (Figure 6). Storytime and progress time are closely related, but are notsimply the logical and chronological sides of the samecoin. Story time can measure elements (such asoptional quests) which may not be represented inprogress time. Logical progress towards gamecompletion, such as obtaining character levels andchoosing the new abilities that flow from them, mayinvolve no story time (although the activities that earnthem may). Viewing game progress from the point ofview of the Story Time can be useful, e.g. in theplanning of story-based vs. mechanics-based rewards,or to measure the rate with which players progressthrough the game story, locate “dead” story elementswhich the players ignore or get stuck in, etc.

5.6 World TimeAs mentioned above, World Time is the chronologicalor abstract time within the game world, and mirrorsthe non-linearity of game progress rather thanproceeding linearly (Figure 6). World Time isundefined if no game world exists, e.g. traditional cardgames such as Poker, Blackjack and Bridge. WorldTime is possibly the most variable layer or viewpointof game time. Abstract and logical World Time is usedin e.g. the classic computer games Pac Man, AtomicBomberman and Space Invaders. The players are notinformed about how long a given action in the gametakes in the fictional world of these games. Firing ashot at the invading aliens in Space Invaders couldtake a second or a minute in terms of World Time, thisis not known. However, the World Time of thesegames can still be mapped logically. This leads to acomplex mapping between World Time and PlayingTime. In some computer game genres, notablyCRPGs, FPS’ and adventure games, the time flow ofthe game world is commonly to synchronous withreal-world time, however, the day/night cycles can besubstantially faster; frozen (e.g. the never-changinggloom of Quake III or Hexen ) or without anydiscernible order beyond acting as ambience (e.g.Serious Sam). For example, in the later installments of




9

9}

the CRPG Might & Magic-series, a game could lastyears in terms of the World Time, but a few dozenhours of Playing Time. In this example, World Timeoperates at more than one level (world time asmeasured by the in-game calendar and as apparentlyoccupied by in-game actions) and in mapping to otherlayers of time a decision must be made about which

World Time level/-s to map from. While World Timein PnPs can vary immensely from game to game, theyare generally associated with fictional worlds wheretime flow is chronological and asynchronous withreal-world time (e.g. player characters can travelbetween distant locations in second of Playing Time).

Figure 6: Mappingbetween PlayingTime and the threenon-linear layers(Progress, Worldand Story Time).Dashed lines:‘Mechanic’progress mapping,dashed/dotted line:‘Story’ progressmapping, dottedlines: Indicatesmapping betweenlayers, bolddashed arrows:Reload, thindashed arrows:Branchcontinuation.

Uniquely to PnPs, player characters canbecome distributed at different points in the gameworld chronology. This leads to the existence ofparallel and interacting instances of World Time, eachtracking one or more player characters, and with theplayer characters being asynchronously or plesio-synchronously (almost synchronized, with thevariation in rate being constrained by specific limits,e.g. the temporal limits of a combat sequence)distributed in the World Time chronology. Forexample, during the PnP sessions it was observed thatwhen the group of player characters split upgeographically within the game world, the subgroupswould often end up at different points of the gameworld chronology, because the GM would focushis/her attention on first one subgroup, advancingthem through the game, while the other were unable toprogress through the game world chronology,effectively frozen at an earlier point in World Time.Although limited in their ability to move theircharacters in World Time, these players would oftenremain in an active state, continuing to role play theircharacters and maintain e.g. Story Time progress,rather than passively observe the action involving theremaining participants. Note that the PnP formatallows situations where PnP players distribute the in-game directorial control, so that all participants areable to progress World Time, not just the GM. Howdirectorial control is distributed in PnPs is decided

jointly by the participants during the formation of theinitial premise or agreement of the game [7].

Massively Multiplayer Games: LARPs andMMOGs/MMORPGs face unique restrictions on theuse of Story and World Times due to the potentiallyhigh number of players involved in these games[3,18]. When hundreds or thousands of players arecontemporan- eously involved in a game, thequantitative problems related with manipulation ofWorld Time increase markedly. In theory LARPs andMMORPGs could utilise flashbacks and similar time-based narrative tools, however the practical problemsassociated with performing these temporalmanipulations in a large-scale game requiressubstantial organization. For example, transportingfive players in a 500 participant LARP back in time,calculating the effect of their visit to the past on thepresent and implementing the effects to the entiregame structure would require massive resources. Incomparison, handling a flashback in a PnP group offive is computationally a much simpler task.

While MMORPGs generally observe a 24-hourday/night cycle, like other RPGs, this pattern issuperimposed on a world which is almost static. Thereis often no calendar and even when there are changesof the seasons (Saga of Ryzom) this typically forms acosmetic layer over an essentially static world, or




10

10}

more precisely a rapidly reasserting game world andWorld Time. It is possible for the actions of theplayers to alter the game world, and thus create aWorld Time where actions can be organized logicallyin order of occurrence. However, these changes arewith few exceptions only maintained for a limitedamount of time, creating a situation where a persistentworld game can run for years of Server Time but not aday goes by in World Time. For example, monsterseliminated by players will quickly respawn, and beginwandering around in the same pattern as before. Thegame world never changes to any substantial degree,and the actions of the players have only a very limitedor temporary consequence on the appearance of theworld [21]. Permanent changes to a game world aswell as World Time are most often caused by contentupdates and patches from the developer, e.g. TheBurning Crusade expansion for World of Warcraft.Other MMOGs, such as Ultima Online and A Tale inthe Desert, do allow players to cause more permanentchanges to the game world, e.g. in the form of buildingpermanent structures or changing the political systemof the game world, and these can be considered ingame time from the point of a logical World Time.

The reasserting feature of real-time run MMORPGsalso give rise to complex mapping between PlayingTime and Progress Time. As MMORPGs do notinclude save and load features, game progress can beviewed as linear from the point of view of a singleplayer. At the same time, content can in some cases beexperienced multiple times; e.g. in the form ofinstances, repeatable quests or the elimination of thesame opponent multiple times. Each time the samecontent is accessed by the player, the experience ismost likely different; however, it can be debatedwhether backwards movement along the GameProgress timeline occurs. A player will in World ofWarcraft gain experience points and/or object rewardseach time the content is accessed. On the other hand,the game story – from the point of view of the playercharacter – does not change as the same content hasalready been experienced.

On a final note, the structure of the game worldsthemselves can impact on the relationship betweenPlaying Time and World Time. For example, the useof geographic compression means that travel in thegame world can occur at a faster than real-time.

5.7 Perceived TimeWhile Playing Time, Engine Time and Server Timeare chronological and objective, these three layers oftime are – to a varying degree depending on the gamein question – subject to personal interpretation by theplayers. Perceived Time (Figure 7) defines WorldTime, Progress Time and Story Time as perceived in

the mind of the individual player. Perceived Time is auseful tool in mapping e.g. the understanding of gameprogress by players. Personal interpretation isespecially important in RPGs where for example gameprogress can be measured in different ways and havedifferent meaning to players who can vary in theirunderstanding of game progress – e.g. in terms ofcharacter levels, items accumulated, NPC contacts,reputation in the game world etc. Similarly, playerscan have different understandings about the pace ofthe game story, and the relative placement of theircharacters or avatars within the story framework, aswell as the magnitude or importance of story segments(which are logically ordered by abstract in terms ofduration). Finally, players can have differentperceptions about the World Time duration of eventstaking place in the fictional world, time between cut-scenes, the time it takes in World Time for a piece tomove on a gaming board (e.g. in games such asSettlers of Catarn). Perceived Time therefore consistsof three sub-layers, a Perceived World Time,Perceived Progress Time and Perceived Story Time.

Perceived Time is important to PnPs which are basedon a shared understanding of an imagined fictionalworld [20] and the operation of the imaginations of theplayers within the framework of the game rules. Thismeans that one of the major challenges of designingand running PnPs is to ensure a sufficiently similarperception about the game events among the players inorder to avoid confusion. Perceived Time can beillustrated using an example from the PnP gamemodule utilized in the experiments: A scene featuredthe player characters being flown in a military dropship from their base of operations to the inceptionpoint of a mission. Because the actual time it took tofly from the base to the drop-off point was notimportant to the game story or mechanics, most of theGMs of the PnP groups did not specify the duration ofthe flight in terms of World Time. Only in one groupdid one of the players ask what the flight time was,and the GM made up a number on the spot. Becausethe duration of the flight was not specified, it is up tothe individual player to define the length. In the groupsessions after the game sessions, it became apparentthat players had very different perceptions about thelength of the flight, as well as other similar scenes inthe game (Figure 7).




11

11}

Figure 7: Perceived Time in a PnP. Thechronologically defined World Time segments can beperceived as having a varying duration by the players,when these are not explicitly defined. At regularintervals, an alignment is performed which brings theplayers to the same perceived point in World Time.

Because Perceived Time operates within the cognitivemind of the players, it is difficult to objectively extractand map to the other layers of time. A player can beinterviewed about how he or she perceived specificevents in terms of their World Time duration, however,there is no guarantee that the perceptions of the playerhas not been altered later in the game, or that theperception can be difficult to articulate verbally in agame form that relies on a “minds eye”. Even though itis important to the personal experience of collaborativeand interactive narrative games, the cognitivefunctionality governing the perception of World Time inPnPs remains unexplored, and it is therefore difficult todescribe in detail. From a more generalised viewpoint,game time can also be perceived differently by playersin terms of their experience of the speed of game timeflow. For example, one player can feel time flowsrapidly while playing a specific game; while anotherfeels time passes slowly (related to the concept of flowby [6]. While such features of game time are related toquestions of immersion, engagement and enjoyment,perceptive differences of game time are relevant ingame studies and analyses, and important in theunderstanding of how PnPs operate. Irrespective,perceived time is a concept that needs moreexploration. Approaches developed in studies of playand engagement in game psychology [e.g. 6] couldform a venue for advancing the current knowledge.

6.0 CONCLUSIONSGame time is a potentially complex construct ofdifferent layers of time. When events occurred withina game can have different answers depending on theviewpoint of game time. This paper has presented amodel for understanding time as it relates to gameplay, where these separate viewpoints of time arerepresented by specific layers. This allows objectiveanalysis of game time the relationships between theviewpoints of time. The approach has been adoptedfrom previous models [8,9,10,11,13,14,15,16], gametheory as well as empirical experiments. Focus is onMP and MMP games and RPGs, however, the modelis applicable to other genres as well as single-playergames. Empirical research has been utilised toadvance the current theoretical concepts by capturingcontextual data, explore and confirm the theoreticalconcepts of existing models, and map the complexityof the interacting viewpoint of game time. The modelincorporates non-linearity, these viewpoints beingrelated to linear measures of game time - that of the

time of the playing of the game, and for digital gamesthe time of the game engine/software as well as forMP games the server hosting the game in question.

While the model is not intended to be the final wordon game time, it is wider in scope than previousmodels and has been developed and tested in anempirical framework. It is intended to be a useful toolin game studies, classification and comparison ofgame design and play styles, as well as forunderstanding the game play experience. Future workwill focus on developing a taxonomy of playeractivities in PnPs and CRPGs based on the recordedgame sessions as well as existing theory, which willform the basis for a segment-based analysis of thetemporal dynamics and differing play styles of thesegame formats. An associated case study in theapplication of the model to design is being planned.

7.0 ACKNOWLEDGEMENTSThe authors would like to extend their gratitude to theICT Centre of the Macquarie University, the playersand GMs who participated in the experiments and Ms.Thea Brolund for invaluable research assistance.Anders Tychsen received financial support from theMQPGRF and ICS PGRF. The experimental work wasapproved by the Macquarie University Ethics ReviewCommittee (HE23SEP2005-D04313).

References[1] Aarseth, E. Cybertext – Perspectives on ErgodicLiterature. John Hopkins University Press, Baltimore,Maryland, 1997.[2] Bartle, R.: Designing Virtual Worlds, New RidersGames, Berkeley, California, 2003.[3] Bøckman, P., & Hutchison, R. Dissecting LARPCollected papers for Knutepunkt 2005. GrimsheiTrykkeri, Grimshei, Norway, 2005.[4] Crawford, C. Chris Crawford on InteractiveStorytelling, New Riders Games, Berkeley, CA, 2005.[5] Crogan, P. Gametime: History, Narrative andTemporality in Combat Flight Simulator 2. In TheVideo Game Theory Reader, edited by Wolf, M. andPerron, B., Routledge, New York, 2003.[6] Csikszentmihalyi, M. Flow: The Psychology ofOptimal Experience. Harper Perennial, 320 pp., 1991.[7] Edwards, R. GNS and Other Matters ofRoleplaying Theory, Adept Press. Online publication,accessed 11th January 2007. URL: http://www.indie-rpgs.com/articles/1/ (2001)[8] Eskelinen, M. Towards computer game studies,Digital Creativity, 12, 3 (Sept. 2001), 175-183.[9] Hitchens, M. Time and Computer Games. CGIEConference 2006, Perth, Australia (2006).[10] Juul, J. Introduction to Game Time. In FirstPerson: New Media as Story, Performance, and




12

12}

Game, edited by Wardrip-Fruin, N. and Harrigan, P.,MIT Press, Cambridge, Massachusetts, 2004, 131-142.[11] Juul, J., Half-Real, MIT Press, Cambridge, MA,2005.[12] Lamport, L., Time, Clocks and the Ordering ofEvents in a Distributed system, Commun. of the ACM,21 (July 1978), 558-564.[13] Lindley, C. The Semiotics of Time Structure inLudic Space as a Foundation for Analysis and Design,Game Studies, 5, 1 (Oct. 2005).[14] Murray, J., Hamlet on the Holodeck: The future ofNarrative in Cyberspace, Free Press, New York, 1997.[15] Rau, A. Reload – Yes/No. Clashing Times inGraphic Adventure Games. Presentation at ComputerGames and Digital Textualities, Copenhagen,Denmark, 2001.[16] Ryan, M., Narrative as Virtual Reality, JohnHopkins University Press, Baltimore, Maryland, 2001.[17] Salen, Katie and Eric Zimmerman. Rules of Play:Game Design fundamentals. MIT Press, Cambridge,MA, 2003.[18] Tychsen, A., Hitchens, M, Brolund, T. andKavakli, M.: Live Action Role-Playing Games: Control,Communication, Storytelling, and MMORPGSimiliarities. Games and Culture, 1, 3 (2006), 252-275.

[19] Tychsen, A., Heide-Smith J., Hitchens M. and ToscaP., S. Communication in Multi-player Role PlayingGames - The Effect of Medium. In TIDSE 2006, SpringerVerlag, Berlin, Germany, 277-288 (2006).[20] Tychsen, A. Role Playing Games - ComparativeAnalysis Across Two Media Platforms. In CGIEConference 2006, Perth, Australia (2006).[21] Tychsen, A. and Hitchens, M. Ghost Worlds -Time and Consequence in MMORPGs. In TIDSE 2006,Springer Verlag, Berling, Germany, 300-311 (2006).

Interesting Times – Modelling Time in Multi-Player …...analysing RPGs across different formats....

Documents

Transcript of Interesting Times – Modelling Time in Multi-Player …...analysing RPGs across different formats....