作者：Jens Peter Jensen

11.处理复杂性

“严格：系统要求模拟少量步骤去预测结果。在垂直卷轴射击游戏中，你将看到子弹朝自己射过来。你并不需要费心思考是否站在原地就会被击中。”（请点击此处阅读本文上篇内容）

“宽松：系统要求模拟多个步骤去预测结果。另一方面，在《Triple Town》中，优秀的玩家需要想着向前进。让玩家基于各种可能影响你想看到的结果的计划进行思考将会带给他们一定的认知负荷。玩家计算中的一个小小错误便有可能生成不可预期的结果。”

相应指标：每个目标的机制和行动

手机游戏的一般趋势是限制游戏和复杂性处理过程。这么做是为了适应游戏的休闲属性以及较短的游戏时间。也有一些策略游戏在长期的玩家互动中呈现出复杂性，但是大多数手机游戏开发者都是围绕着快速理解和快速游戏进行设计，从而完全忽视了行动结果。所以如果你需要使用游戏指标去衡量手机游戏的复杂性，那么游戏可能对于大多数手机用户来说都太过复杂了。既然如此，设计师将考虑是否要完全改变游戏，或者至少选择其它市场。当然，你也可以创造一些全新的内容，但往往说起来容易做起来难。

一种可能方法便是去追踪游戏中的不同机制，并明确玩家使用每种机制的频率。根据游戏是基于关卡，目标还是宽松结构，我们可以选择基本比例或者目标基础做到这点。即使游戏并未使用关卡，设计师也应该单独测量。像《部落战争》和《辛普森一家：枯竭》等游戏便拥有较为宽松的结构：玩家可以在特定区域内按照自己想法安置建筑，但是他们仍需要经历一些关卡。在这些例子中，游戏的目标便是完成关卡。还有其它拥有宽松结构的游戏则是逐渐扩展玩家的基地/城堡/城市。这些游戏的目标便是扩展行动。如果游戏同时拥有关卡和扩展机制去划分目标的话，你便能够比较不同类型目标的复杂性。

Springfield（from gameanalytics）

扩展Springfield城镇是游戏目标。

clash of clans（from gameanalytics）

在《部落战争》中计算重要的建筑也是一个目标。

然后指标将变成玩家会使用怎样的行动和机制去实现一个目标并走向下一个目标，以及他们需要花费多长时间去执行这些行动。最终数据将呈现出玩家在不同游戏机制中实现每个目标所执行的行动数和花费的时间。这一信息可以用于明确游戏中不同关卡的复杂性。基于此，设计师便可以决定是该提高还是降低不同关卡的复杂性，从而实现预期的难度标准。

12.选择复杂性

“严格：考虑更少的选择。在我最近创造的升级系统中，我给予玩家3种升级选择。我本来想着提供给他们一个包含60种升级的菜单，但是考虑到这可能会让他们感到压力。所以后来决定只呈现一些重要的选择，并留给他们一定的心理空间去思考每个选择，然后挑选最重要的那个。”

“宽松：必须考虑许多选择。围棋中存在许多潜在的移动以及上百种次要移动。这些选择复杂性也是推动数千年来无数玩家反复挑战围棋的主要原因之一。”

相应指标：机制和行动以及目标和行动

选择复杂性似乎包含了游戏玩法选择和外观定制选择。虽然这是两种不同的内容，但却都增加了游戏内容的数量。这两种内容的区别在于，内容定制并不会对游戏玩法产生影响，但是游戏玩法选择却会在某种程度上影响游戏。本篇文章将着重讨论游戏选择复杂性（而非外观定制内容）。

使用游戏升级/选择机制主要有两种不同的方法。一种便是公开且无向导的方法，即玩家可以基于不同选择进行探索或升级。这一方法是无限制的，留给玩家大量的空间。这经常用于RPG游戏中，如《天际》，某些MMORPG也使用了这种方法，如《魔兽世界》和《我的世界》。

另外一种方法便是有限且严格的引导。玩家将面对非常有限的选择/升级范围，并且游戏前进道路也是线性的。如此设计是为了让玩家按照特定数序执行行动并前进。设计师可以清楚地呈现这种方法，或者让玩家产生选择的错觉。像《使命召唤》，《战神》，《辛普森一家：枯竭》和《大都市》等经济类游戏便使用了这种方法。

call of duty（from gameanalytics）

在《使命召唤：黑色行动2》中有许多武器和升级，但玩家只能在严格的前进道路上打开它们。

megapolis（from gameanalytics）

《大都市》中拥有许多建筑，但是只有在玩家到达某一特定关卡时它们才会成形。

这两种方法并不是“非此即彼”的关系，只有少数游戏会极端地走向其中一个方向。

处理复杂性中所出现的指标也可以用于校验选择复杂性。简单地使用同样的数据并寻找升级游戏机制以及能让玩家使用不同行动去获取同样效果的环节。然后你将能够感受到属于选择复杂性的指标。再一次地，设计师可以使用数据去调整选择复杂性而延长或缩短游戏进程。

13.社交复杂性

“严格：另外一个人具有明显的意图，能力和心理状态。在MMO中，玩家将扮演较高级别的治疗者并待在某一特殊群组将会在此会面的特殊地点。如果你让他们共同展开冒险的话，你将能够了解到他们会做些什么。或者在一个管理贸易窗口中，当对方拿出一瓶药剂去对换你的宝剑时，你便能够清楚他想做什么。虽然这种信息传达有点含糊。”

“宽松：另外一个人伪装了自己，不表露出自己的意图，能力和心理状态。

相应数据挖掘：聊天和电子邮件

Cook似乎只专注能够直接影响游戏玩法的社交元素，而忽视了社交互动的主要元素，也就是聊天。如果手机和社交游戏是围绕着简单机制进行设计，那么它们便不需要复杂的玩家合作，并且大多数玩家vs.玩家游戏的例子都只是关于玩家vs.静态数据（保存在服务器上）。当然也存在间接合作型玩家互动，如在《FarmVille》中浇灌其他玩家的植物之类。一个玩家贴出了某些内容而其他玩家去浇灌他的植物，这并不属于直接的玩家间社交互动，而是玩家与服务器状态间的互动。

在任何能让玩家进行互动的游戏中都设置了内部聊天功能或电子邮件系统，而游戏开发者必须时刻关注着玩家们的聊天内容。玩家间的交流将提供给开发者他们对于游戏的看法以及如何游戏等宝贵信息。所以对于数据挖掘来说聊天和电子邮件系统是非常显著的目标。不过开发者却很难从聊天中提取信息，因为这需要大量的人为解释。

最简单的方法便是设置指标去搜索开发者所感兴趣的单词或短语。即我们可以搜索“游戏，优秀，出色”或者“游戏，糟糕，无聊”等关键词。但是我们最好更具有针对性地进行搜索，特别是当结果是面向人类读者时。如果搜出了20多万行聊天记录的话，我们便不可能得出真正有意义的结果。我们可以无需基于人为解释去使用挖掘数据，但这却会让结果显得较为含糊。一般说来，搜索更具有针对性，最终得出的结果便更有效，人们也能够以更加有意义的方式去使用这些数据。

chat（from gameanalytics）

《亚瑟王国》中的玩家聊天机制，其目的是挖掘真正有用的内容。

这种方法能够用于收集许多不同类型的信息。其可能性只受到想象力和指标计算资源的限制。不过这种方法涉及到了隐私侵犯问题，所以你必须在最终用户许可协议中添加监视并检查游戏内部聊天记录和邮件的相关条款。

14.时间压力

“严格：要求基于玩家理想中的速度而模拟模式。这与处理复杂性和选择复杂性有关，因为玩家只能基于特定速度去执行模式。如果降低时间压力的话玩家更有可能创造出因果联系。举个例子来说吧，《NetHack》带有复杂的互动系统，要求真正的侦探去解谜。为了提高玩家间联系的可能性，我们可以将游戏设置为回合制游戏，并让玩家根据自己的想法去决定每个回合所需要的时间。你将会发现情境变得更加复杂，甚至连有经验的玩家也会放慢速度去理解所有分枝内容。”

“宽松：要求快速模拟模式。在《瓦利奥制造》，每个谜题中并不存在真正的复杂性。但是我们却可以通过设置较短的时间去提升玩家的认知负荷以及结果的不确定性。”

相应指标：竞争时间vs.时间限制

这里所说的时间压力并不只是关于游戏中的时间限制挑战，同时还包含了游戏速度。提高游戏速度也是提高游戏难度的一种方法，就像经典的《吃豆人》那样。

Cook所写的宽松方式是在玩家所习惯的速度上加强时间压力。但是这些游戏经常会逐步使用更多时间压力去提升难度，即最终会让玩家感到不安。而设计师所面临的挑战将变成基于能让玩家感到压力并充满兴趣的方式去平衡难度比例。

如果玩家购买了游戏，他们便具有额外的利益动机而继续游戏，所以设计师便能够创造出更具挑战性的游戏，并期待玩家能够继续游戏。在免费模式中，难度平衡会发生改变。大多数免费游戏使用的是“提升游戏速度”或“降低游戏难度”的盈利策略。这便意味着平衡难度的挑战并不能让游戏变得更有趣，反而会让开发者很难去判断玩家是否会在游戏中消费。所以免费模式应该在保证不会赶走玩家的前提下，推动他们的极限，并超越他们所习惯的范围。

我们可以创造一个反应基于时间关卡的游戏（例如像《Candy Crush》，《宝石迷阵闪电战》以及《Bubble Mania》这样的三消游戏）难度的指标。除了三消游戏，还可以是其它带有时间限制且基于目标结构的游戏。设置指标去收集每个玩家完成每个关卡所花费的时间，收集玩家死去或失败的时间，收集他们在每个回合所投入的时间。然后你便可以获得玩家完成每个关卡的平均时间，以及每个关卡中有多少玩家遭遇失败，并发现他们是在何时选择放弃。

bejeweled blitz（from gameanalytics）

在《宝石迷阵闪电战》中，Cyborg Cat一直在注视着玩家游戏，这便是游戏所使用的时间压力。

基于这些数据，设计师便能够判断游戏难度是否符合预期设想。如果所有玩家在第一个关卡便遭遇失败，而设计师最初的设想是玩家可以存活到第十个关卡，那么游戏便太复杂了。很多设计师都在一开始赋予游戏过高的难度。而基于上述指标所收集到的数据，设计师便能够有效地调整难度。

如果游戏遵循的是免费模式，那么我们便可以使用同样的数据去调整难度而刺激游戏内部购买。如果玩家的前进速度过快，他们也许会更加喜欢游戏，但却会因此错失了花钱购买额外时间和能力的机会。在这种情况下，设计师可以使用上述指标去选择一个让玩家遭遇失败的适当时刻。

例如设计师可以瞄准10分钟无缝游戏去吸引玩家的注意。然后引进一些机制让玩家遭遇失败，并为此呈现出付费游戏体验的优势。

在手机设备上，严格更加重要

当着眼于这14条内容时，我们更加清楚“严格”游戏更容易尝试与理解，并且更加适合于手机市场。所以在面向一般手机游戏用户设计游戏时，“严格”方法总是更加适用。如果设计师想要在手机市场中尝试一些新内容，他们便可以选择“宽松”方法。如果设计师的目标是面向主流手机用户设计游戏，他就需要仔细浏览这14条内容并确保游戏是否足够严格。而如果设计师只是想要进行实验的话，他们便可以着眼于这14条内容并选择适合的“宽松”方法去激发游戏机制的新理念。

（本文为游戏邦/gamerboom.com编译，拒绝任何不保留版权的转载，如需转载请联系：游戏邦）

Building tight games with game metrics (Part 3)

By Jens Peter Jensen

11. Processing Complexity

“Tighter: System requires simulating few steps to predict an outcome.  In a vertically scrolling shooter, you see the bullet coming towards you.  It doesn’t take a lot of thought to figure out that if you stay in that location you are going to be hit.”

“Looser: System requires simulating multiple steps to predict an outcome.  On the other hand, in Triple Town, good players need to think dozens of moves ahead.  Thinking through all the various machinations necessary to get the result you want adds a serious cognitive load to the player.  A single mistake in the player’s calculations yields unexpected results.”

Corresponding Metric: Mechanics and actions per goal

In mobile games the general trend is to limit both game and processing complexity. This is in order to accommodate the casual nature and short play times that characterize the mobile platform. There are some strategy games that exhibit complexity in long-term player interaction, but most mobile game developers design for quick understanding and quick play, thus ignoring action consequences completely. With that in mind, if you need a game metric to measure the complexity of a mobile game, the game is probably already too complex for the average mobile consumer. In that case, the designer should consider whether the game should be fundamentally changed or at least directed at another market. Of course, you might also get away with making something new and revolutionary that breaks all conformities, but that is easier said than done, right?

One possibility is to track how many different game mechanics there are in the game and how often the player uses each one. This can be done on a general scale or on a goal by goal basis, depending on whether the game is level-based, goal-based or has a loose structure. Even if the game does not use levels, the designer should somehow separate the measurements. Games like Clash of Clans and Simpsons Tapped Out have a more loose structure: the player can place buildings where they like within a difined area, but they still have to go through levels. In those cases, the goal of the game could be achieving levels. Other games with a loose structure gradually expand the area of the player’s base/castle/city. In those cases the goal could be the act of expanding. If the game has both level and expansion mechanics that delimit the goals, you can compare the complexity of the different types of goals to each other.

Expanding the town of Springfield can be a goal to track for.

Or count the important buildings in Clash of Clans as a goal.

The metric could then check for what actions and mechanics the player uses to get from one goal to the next and how long it took to execute those actions. The resulting data would show how many player actions and how many times the different game mechanics were used to achieve each goal. This information can be use to determine the complexity of the different stages of the game. Based on this, the designer can choose to increase or decrease the complexity of the different stages to reach the desired rate of progression and difficulty.

12. Option Complexity

“Tighter: Fewer options are available to consider. In a recent upgrade system I was building I give players 3 choices for their upgrades.  I could have given them a menu of 60 upgrades, but that would be rather overwhelming.  By focusing the user on a few important choices, I give them the mental space to think about each and pick the one with the biggest impact.”

“Looser: A large number of options must be considered.  In a game of Go there are often dozens of potential moves and hundreds of secondary moves.  This options complexity is a large part of why the game has been played for thousands of years.”

Corresponding Metric: Mechanics and actions pr. goal and options

Option Complexity seems to cover both gameplay options and appearance customization options. These are two very different things, but both add to the amount of game content. The main difference between the two is that appearance customization does not have to influence gameplay in any way, while gameplay options should influence the game in some way. In this article the focus will be on the game option complexity and not the appearance customization.

There are two different approaches to game upgrade/options mechanics. One is an open and unguided approach, where the player has many different options to choose from and aspects to explore or upgrade. This approach is non-restrictive and leaves everything up to the player. This is often used in RPG games like Skyrim,to a lesser extend in MMORPGs like World of Warcraft and Minecraft.

The other approach is narrow and strict guidance. The player has a very limited range of options/upgrade possibilities and game progression is linear. The game is designed so that the player must do things in a specific order so as to advance. The designer can choose to do this in an obvious way or he can give the player the illusion of choice. This is used in games like Call of Duty, God of War, Simpsons Tapped Out and Megapolis, and basically all tap based economy games.

There are a lot of weapons and upgrades in Call of Duty Black Ops II but they unlock trough a strict line of progression.

Megapolis has a multitude of buildings, but they are not made available until the player reach a certain level.

The two approaches are not to be thought at as “one or the other” as very few games go to the extreme in either direction. It is more a case of leaning towards one approve with more or less intensity.

The same metrics as described in the Processing Complexity section can be used to examine Option Complexity. Simply take the same data and look for upgrade game mechanics and for sections where the player can use different actions to achieve the same effect. Then you will get a sense of the Options Complexity metric. Again, the designer can use the data to tweak option complexity in order to lengthen or shorten game progression.

13. Social Complexity

“Tighter: Another human broadly signals intent, capabilities and internal mental state.  In an MMO, a player dresses as a high level healer and stands in a spot where adhoc groups meet up. There’s a good chance you know what they’ll do if you ask them to go adventuring together.  Or in a managed trade window, you know exactly what you are getting when he puts up a potion for your sword.  There is little ambiguity.”

“Looser: Another human disguises, distorts or mutes intent, capabilities and their mental state.”

Corresponding data mining: Chat and mail mining

Cook seems to focus only on the social aspects that directly influence gameplay, ignoring a key element of social interaction – namely the chat. As mobile and social games are designed to be simple, there is no need for complex player cooperation, and most examples of PVP are really just player vs. static data saved on a server. There are, of course, indirect collaborative player interactions like watering each others’ plants in Farmville and the like. One player posts something and someone else waters the plants, so this is not direct player on player social interaction, but actually player on server state interaction.

In any game where players can interact an ingame chat or mail system is usually implemented, and it is always very interesting for the game developer to keep an eye on what the players are talking about. The player on player communication is a treasure trove of information of what the players think of the game and how they play it. That makes the chat and/or mail system an obvious target for data mining. It is however difficult to properly extract information from the chat, because a lot of manual (human) interpretation is needed.

The simplest way is to set up a metric that searches for words or phrases that the developer is interested on. This could be done for general concerns (for example, if the player likes the game) by searching for words and phrases like “game, good, awesome” or “game, bad, sucks”. But it is better to be more specific, especially if the results have to read by humans. If the search query comes up with 200.000 lines of chat, it is not possible for a human to get something meaningful out of it. It is possible to use the mined data without human interpretation, but that makes the results very ambiguous. Generally, the more specific the search the more useful the results, and the more easily humans can engage the information in a meaningful way.

The player chat, here from Kingdoms of Camelot, about everything, the task is to mine only the good stuff.

This approach can be used to gather many different types of information. Possibilities are only limited by imagination and the resources necessary for metric computation . There is an element of privacy invasion in this method, so be sure to always include a clause for monitoring and examining ingame chat and mail in the EULA.

14. Time Pressure

“Tighter: Requires simulating the model at the player’s preferred pace.  This is related to processing and option complexity since players can only execute their models at a given pace.  Players are more likely to make causal connections if the time pressure is greatly reduced.   For example, the game NetHack has complexly interwoven systems that require real detective work to decipher.  In order to increase the likelihood that players will make the connection, the game is set up as a turn-based game where players may take as much time as they want between turns.  You’ll see that as the situation becomes more complex, even good players will slow down their play substantially so they can understand all the ramifications.”

“Looser: Requires simulating the model quickly.  In a game of WarioWare, there isn’t really much complexity involved in each individual puzzle.  However, we can dramatically ramp up the cognitive load and increase outcome uncertainty by setting a very short timer.”

Corresponding Metric: Completion time vs time limits

Time pressure here does not only refer to time limit challenges in games, but also to the speed at which the games actually play. Increasing the speed of the game is a guaranteed method of increasing difficulty, dating back to classics like Pac Man.

Cook writes that the loose way is to enforce time pressure at the rate that the player is most comfortable with. But such games usually increase in difficulty by gradually applying more time pressure, which means that they will eventually become uncomfortable anyway. The challenge of the designer then becomes to balance the difficulty rate in such a way that the player feels challenged and interested.

If the player bought the game, then there is an extra financial motivation to keep playing, so the designer can make the game very challenging and still expect the player to play along. In the “free to play” or freemium model the balance of difficulty is changed. Now there is the added aspect of trying to motivate the player to spend money, while not pushing the player away. Most freemium game use a “make the game faster” or “make the game easier” monetization strategy. That means that the challenge of balancing the difficulty is not necessarily to make the game more interesting, but to make it difficult enough to determine the player to make in-game purchases. So freemium games should push the player’s limits beyond the comfort level, while at the same time trying to not drive them away.

It is possible to make a metric that reflects the difficulty of games which use time-based levels (for example “match three” games like Candy Crush, Bejeweled and Bubble Mania). It doesn’t have to be match three games, it can be any game that has a goal-based structure with a time restriction. Setup the metric to collect how long each player takes to complete each level, and to collect when the player dies or fails, and how long each play session is. Then you can get an average time to complete each level, get a count on how many players fail each level and find out when they give up.

The Cyborg Cat is watching the player in Bejeweled Blitz, that uses time pressure as a game mechanic.

With this data, the designer can then find out if the game is more or less difficult than expected. If all the players failed the first level and the designer expected the average player to get to level 10 without dying, the game is too hard. It is a common design mistake to make the game too hard at first. But with the data collected through the metrics above, the designer can fine tune the difficulty.

If the game follows the freemium model, the same data can be used to tweak the game difficulty towards motivating in-game purchases. If the players are progressing too fast they might enjoy the game more, but they would also be less likely to buy extra time or powerups. In that case, the designer could use the above metrics to choose a suitable moment when the player should fail a level.

For example, the designer could aim for 10 minutes of seamless playing, only to get the player interested. Then he could introduce some mechanics which make the players fail a lot in order to introduce the advantages of a paid game experience.

Tighter is more important on mobile devices

When looking at the 14 different parameters, it becomes clear that “tight” games are easier to play, understand and therefore most suited for the mobile market. So the “tight” approach is clearly the better way when designing a mobile game for the general mobile game consumer. If a designer wants to try something new on the mobile market, then taking the “loose” approach would surely achieve that. If the designer is trying to make a game for the mainstream mobile consumer though, it would be a good idea for him to go through the 14 parameters and make sure the game is as tight as possible. If on the other hand the designer is into experimentation, looking at the 14 items and choosing where to be “loose” might inspire new ideas for gameplay mechanics.

These blog posts have given some examples of how to make use of game metrics in game development. Game metrics have to be very specific in order to be effective, so it they do not apply to your particular game maybe they can at least inspire you to come up with your own relevant metrics.

In any case, if you are designing a game, no matter the platform, do yourself the favor of reading Daniel Cooks blog post.(source:gameanalytics)