Game of War

Game of War is a mobile game currently making $1.2M/day and having 2.2M daily active users. Its developer, Machine Zone, is valued at $3 billion (while Zynga is valued at $2.7 billion and Supercell at $5.5 billion.).

The game is quite high up the charts of the App Store, only surpassed by Clash of Clans' $1.7M/day and 4M daily active users and probably Candy Crush as well. CoC and GoW are similar in that they are both free-to-play empire-builder war games for mobile. Whereas I find CoC to be a polished game with elegant mechanics, Game of War is said to be everything wrong with mobile monetization. Let's have a look!

Copycats

In an interview in 2012, Gabriel Leydon, Machine Zone's CEO, said they are making games that are very, very special and unique in the market. But in the same interview, he also said If you want to scale fast and you have the ability to do it cheaply, you just clone. With GoW, they did not go the special/unique route: they cloned 99.9% of their gameplay and UI from Kabam's The Hobbit (which launched for mobile in October 2012). Shameless cloning is a common practice among mobile games, and The Hobbit itself had cloned a lot of its gameplay and UI from Evony (a Flash game from 2009), which itself had cloned most of its gameplay from Civilization.

Going back to GoW, the trailer features a battle in real time scene that actually never happens in the game. In the game, the player can see their armies move in real time, but when battles happen, the player only sees a report of the troops expended and resources gained. No visual battling actually happens. So the game launched with a lot of hype and exaggeration. Since it's a clone of The Hobbit, the marketing team had to exaggerate to compete.

Translation vs innovation

When GoW launched in July 2013, Venturebeat wrote they were using advanced technology, including a real-time translation engine, a sophisticated communications platform where players can send threaded emails, text chat, make comments, and share their feats on social forums. There really is nothing new or sophisticated with in-game emails. The translator, however, is GoW's unique attribute, what the game is known for. How good and useful is that famous translator?

When typing on the miniature keyboard of their iPhone, people use slang and make spelling mistakes. So the translator has to crowd-source a lot of sentences to players. For example, let's say a French player says "lu, a va bi1?". To translate this for American players, I think the sentence is first automatically-translated into English, leaving non-translatable words untouched: "read, have go bi1?". This gibberish is then given without context to 4 players who are told they'll receive in-game currency if they can fix the sentence. The first player fixes the sentence into "I read you, I have to go", the second suggests "Read (what I wrote above), I have to go and buy one", and so on. A fifth player receives the four tweaked sentences and is awarded some in-game currency for picking the sentence he thinks makes most sense. The player who wrote the chosen sentence receives coins, whereas the other three receive a thank you message with no reward.

This sounds like a good idea, but there's a reason why after all these years Google Translate still sucks, and why translator is still a job. "lu, a va bi1?" in proper French is "Salut, ça va bien?", which Google can translate easily to "Hi, how are you doing?". But most French players can write "hello" by themselves. In Kings of the Realm, a game very similar to GoW, I've seen many American players write to Russian players in Russian. In Hearthstone, the game has 6 emotes (for "hello", "well played", and so on), and needs no translation system whatsoever. Did GoW focus on and boast its (mediocre) chat translator at launch because it was the only thing that differentiated it from The Hobbit?

Cash grab

MachineZone was actually called Addmired when it started. They were developing dating apps. Maybe the name made sense at the time, but even for a dating app, it's a terrible name: are users going to get mired in ads? That no founder flagged the name as inappropriate tells a lot about the company's business model.

GoW reminds players that they can spend money every time they login. The first screen showed when logging in is the one below. The font is inelegant, reminiscent of Asian mobile games like Puzzle and Dragons. The fireworks at the bottom obstruct one of the products offered in the bundle. The bundle name is "Summer MEGA GOLD Sale!!!". And a timer of 30 minutes pretends that the sale will end within 30 minutes. In fact, the player only needs to log back in to get spammed by the sale again...

In Clash of Clans, no upgrade takes longer than 14 days. The game provides 5 builders, so the most intense players have something to upgrade roughly twice a week. In GoW, the longest upgrade times is 23 years. Nobody would wait 23 years for an upgrade to finish! The designers do not expect players to wait 23 years. Everybody understands that year-long waiting times are here only to make players skip them with money. No surprise that some players report having spent $9,000 in the game. A kid in Belgium even spent $46,000.

Sexy advertising

GoW spent $40M to run an ad where model Kate Upton shows her cleavage during the Superbowl. Some people were outraged, but this advertising strategy was already used heavily in 2009 by Evony. Using sex in ads is not new, but that does not excuse it. It's interesting that a company valued at $3 billion has only one product, and the content of and advertising for that product were both 99.9%-copied from another company's product. In contrast, Clash of Clans' core gameplay may not be original but the ad they ran during the Superbowl was funny, silly, and just more elegant.

In fact, GoW's ads could be worse. For example, Evony's advertising strategy was much more questionable than GoW's. Through iEvony, they rewarded players with game coins when they referred a friend or when a friend they referred purchased game coins with real money. In order to bootstrap the pyramid scheme, iEvony was invite-only. Evony also used any search-engine optimization trick they could, stole ad photos and game art, and showed soft-porn images that have nothing to do with the game. So GoW still has some way to go!

Why do GoW (and CoC) use ads? After all, they have so many players already that word-of-mouth should bring them more. In practice, only a small percentage of new players will keep playing the game past the one-week mark. An even smaller number will spend money. Thus these games aim at attracting as many players as possible, so as to increase the number of spenders. But there will come a time when there will be no more new players to draw to the game. But since 2013, the cost for attracting new players outweighs the average amount spent by players ($2.73 vs $1.96). Since players get tired of games eventually, these greedy business models are unsustainable.

Conclusion

GoW is an over-the-top cash grab using sex [ads] to fuel a game where sex really isn’t a factor. The game has a terrible UI plagiarized from another game, and a horrible gameplay also plagiarized from others. This is the kind of game that gives a bad name to free-to-play mobile gaming. I'm disappointed that millions of players follow the ads and don't see that it is so crappy.

Mobile crap - daily login

This series of articles is a rant about mobile game design. It may extend to console and PC gaming, but I only talk about what I experienced through most games featured on the iPad app store in the past year or so.

Designers could invent new mechanics to improve the core gameplay, and thereby improve retention metrics in the long run. But they cheat: they invent new mechanics to immediately improve retention metrics, thereby ignoring or forgetting to improve the core gameplay.

Mobile games try to maximize various metrics: their number of daily and monthly active users (DAU and MAU), the average session length, and so on. To do so, they introduce game mechanics that encourage players to come back every day or play longer. In games like Clash of Clans or Farmville, the come-back mechanic is the core game loop, so it is not a problem. Waiting and coming back is what these games are about. But in traditional fighting, racing, or match-three games, the only reason why players would come back or play longer is because they like the fighting, racing, or match-three mechanics, not because they like waiting.

More and more games have been adding come-back mechanics on top of their core gameplay. In 2004, World of Warcraft started granting a rest XP bonus to players who do not login for a long time. Blizzard designers originally implemented the rest mechanic to prevent hardcore players to level up too quickly. Beta-testers hated it, so Blizzard made it a reward for casual players instead of a punishment for hardcore players. Nothing changed mathematically, but players prefer mechanics if they are framed as a reward.

More recently in mobile games, Plundernauts has been giving out a welcome-back reward for logging in. They also give out bounties which must be completed within 24 hours - an incentive to play at least a few battles during the day, most likely in one session. 24-hour bounties are great for metrics such as the average session length, but they do not make the game more fun in the long run.

Puzzle and Dragons gives two login bonuses: 1) a consecutive login bonus, with rewards increasing every consecutive day, and reset the day the player does not login, and 2) a cumulative login bonus, increasing every day the player logs in, and never reset. Also, there are 5 exclamation marks for 7 lines of text in the login bonus message. By abusing the exclamation mark they over-emphasize my login, and it looks very amateur.

And finally, mobile games such as Birzzle or Mother of Myths show a daily login calendar when the player logs in. Some calendars focus on the week (4x7 cells), others on the month (5x6 cells). The calendar resembles an Advent Calendar, except 1) nothing happens at the end, and 2) the calendar mechanic is not properly tied to the core gameplay (receive more and more gold, or increasingly better gifts). The only point of the calendar is to keep people logging in for the rewards, not to have more fun playing the game.

A lot of mobile games have mediocre and dull core gameplay. These games are often sugarcoated in happy-shiny graphics and habit-forming mechanics. None of those improve their core gameplay. These games remain shallow.

Using achievements for analytics, retention, or skill

Achievements can have many purposes in a game.

For analytics

At GDC 2010, Bruce Philips from Microsoft showed how he used Xbox Live achievements to evaluate progression and campaign completion in a dozen FPS. These achievements already existed, and were determined by the game designers, yet provided very useful for analytics.

Traditionally, achievements are binary: you have unlocked it, or you haven't. You can treat achievements as milestones that the player reached. Milestone achievements include: FPS campaign milestones or Forza 4's "import data from Forza 3".

Achievements can also take several levels. For example, Clash of Clans' Release the Beasts achievement takes 4 values: 0 when you start, 1 when you unlock Archer, 2 when you unlock Wall breaker, and 3 when you unlock Dragon. This achievement, by itself, suffices to segment the player base into new, intermediate, advanced, and expert players. It's a double-win: players have more achievements to go for, and you get a better picture than with a binary achievement.

An achievement can keep counting even after it's been unlocked; for example, Clash of Clans' Gold Grab has 3 steps (4 values) but keeps track of the gold you looted. That way, achievements can also be numerical, not just binary or Likert-like.

Achievements can be even more complex and hold metadata such as when you unlocked it, how many days it took you to unlock it, after how many tries, in how much time, and so on. With this metadata, you can label player behavior more precisely. For example, a Clash of Clans player may unlock Gold grab 3 after 3 months where another would only take one month. The later certainly plays much more (number of sessions, session length) than the former.

For retention

Achievements provide long-term goals to the player. Moreover, they can be made so the player always has a near-complete achievement in sight. For example, you reach 490/500 mortars destroyed in Clash of Clans' Mortar Mauler achievement. You play another hour to destroy 10 mortars and get the 10 gems, but by then, another achievement has reached near-completion too, and you could play another hour.

Retention achievements don't always involve grinding. They can encourage the player to explore the design space.

For skill (or luck)

For example, in FTL Repair back to full health with only 1 HP remaining. Skilled players can risk it, for the thrill. But other players can also wait for it to happen by itself.

Other FTL examples include completing a mission in less than X minutes, or not using any weapon until sector 5. Or even taking 5 turns in a row in the card game Ascension.

The social strategy game genre

Here is a quick list of games that resemble Clash of Clans in some ways. Some wargame mechanics may date back from RPGs of the 90s, such as Age of Empires, or even from tabletop wargames of old. These old games are off-limits.

Released	Name	Theme	Description
Sep. 2004	Travian	Antiquity	The player can pick one of three tribes: the aggressive Teutons, defensive Gauls, and average Romans. Teuton players usually farm resources from Gaul players. Players can trade resources, message each other, and join alliances. Troops take time to produce, and can be unlocked and upgraded. A hero can equip gear, complete quests, and gain XP in battles. The game ends when a player upgrades a Wonder building to level 100. Premium accounts finance the game and have game advantages over free accounts.
June 2009	Farmville	Farm	Facebook game by Zynga. The player plants crops that grow with time, but wither if not harvested on time. Players can receive crops or animals from friends. In-game promotional campaigns for real-life brands like McDonalds or 7-Eleven. Reached 80M MAU and 30M DAU in Feb 2010.
March 2010	Backyard Monsters	Gory monsters	Facebook game by Kixeye. The design goal was to play an RTS game in short sessions. The game targets a hardcore player segment in several ways. First, the art is unusually gory, bloody, and industrial for a Facebook game, yet it was praised as the prettiest game on Facebook when released. Second, the monetization relies on the players who hate losing and are ready to spend money to gain an advantage. As a result, 85% of the revenues come from selling speed ups, retention is 5x longer than other Facebook games, 97% of the player base are males between 25 and 45, and the average user plays 3-4 session per day for 30 minutes per session on average. They peaked at 4.5M MAU and 1M DAU in Summer 2011.
May 2011	Battle Pirates	Warships	By Kixeye. Peaked at 1.6M MAU and 240k DAU in Q4 2012. The player controls a military island, and can be attacked synchronously, in real-time, by the boats of other players. Watch some gameplay.
Aug. 2011	Edgeworld	Aliens	Facebook game by Kabam. 610k MAU in Dec 2011. The developers did not expect players to be attacking each other for resources.
Sep. 2011	War Commander	Post-apocalyptic	Facebook game by Kixeye. Peaked at 5.2M MAU and 660k DAU in Sep 2012.
Aug. 2012	Clash of Clans	Medieval Norse	One month of soft launch: first released in beta on the Canadian iOS appstore in July 2012, then on all iOS appstores in August 2012. Released on Android in October 2013. Supercell had two goals: adapting the genre for the tablet, and bringing new segments to the genre. The name of the strongest unit in CoC, P.E.K.K.A., reminds of the name of the strongest monster in Backyard Monsters: D.A.V.E.. Some say that CoC is Backyard Monsters with polish. Troops are consumed if they are deployed in battle: even if it stays alive until the battle times out, the player loses the troop for ever. A lot of community management and communication goes on between SuperCell and the players through the forums or Facebook. Depending on where you look, CoC had 8.5M DAU in April 2013, and 4.3M MAU and 2.7M DAU in January 2014.
July 2013	Ninja Kingdom	Medieval Japan	Also called Dojo Mojo. Facebook game by Zynga. The Jade Mine can only be staffed with captured enemy troops, not with the player's workers. Each captured enemy troop generate 1 jade per 24h. If destroyed by an enemy raid, all the jade in the mine is lost. 4.6M MAU and 600k DAU in January 2014.
July 2013	Battle Beach	Modern warfare	The troops are a complete ripoff from CoC. 21k MAU and 9k DAU in January 2014.
July 2013	Jungle Heat	Modern warfare in a tropical jungle	Android game by Mail.ru Games. The troops are a complete ripoff from CoC. 260k MAU and 110k DAU in January 2014.
July 2013	Castle Clash	Medieval Fantasy	Around 20 heroes with different skills. Twelve troops: 3 tiers of power for each of the 4 attack types: heavy (strong against ranged), ranged (beats magic), magic (beats heavy), and heavy anti-building.
Sep. 2013	Amazing Clan War	Medieval Norse	Blatant CoC ripoff by a Chinese company. Even the loading screens have the same layout.
Sep. 2013	Kingdom Clash	Medieval fantasy	Soft launch in Australia in May 2013. 5k MAU and 2k DAU in January 2014.
Sep. 2013	Total Conquest	Roman Empire	iOS and Android game by Gameloft. Soft launch in Canada and New Zealand app stores a month earlier. The troops are very similar to CoC, except they don't seem to fly. 100k MAU and 30k DAU in January 2014.
Oct. 2013	Lord of the Guardians	Forest animals	Three solo campaign modes (normal, heroic, legendary), each with 150 mission. 5k MAU and 2k DAU.
Oct. 2013	Samurai Siege	Medieval Japan	iOS and Android (implemented using Unity). 1.2M installs and 300k DAU when it launched, 100k MAU and 44k DAU in January 2014. Same core units as CoC. Troops and buildings are unlocked through the solo campaign. The campaign has a (mediocre) storyline. Loot items are stolen from other players. When all 6 of them are gathered, they provide an extra army camp (more troops) or a cannon (more defense). Alliance War pits groups of players against each other for 12 hours. This was probably based on the periodic trophy push organized on the CoC forums to break the farming routine.
Nov. 2013	Call to Arms	WWII	Developed by GREE. Soft launch in October. Lets you simulate an attack on your own base.
Nov. 2013	Galaxy Factions	Space	A month of soft launch. 10 heroes and 12 units.
Dec. 2013	Boom Beach	Pearl Harbor	Soft launch in November. Developed by SuperCell. Troops that stay alive at the end of a battle are available in the next battle. The player can direct the units to attack a particular building through flares. Explore the ocean (at a cost) to find islands controlled by NPCs and other players. Attacking a neighbor player also has a cost. Nearby captured islands generate resources for you. The units are inspired from Team Fortress.
Dec. 2013	Dark District	Dystopian futuristic city	Developed by Kabam. 6 months of soft launch, probably because of all the crashes and bugs. Assign your troops a building to attack.
Dec. 2013	Sensei Wars	Medieval China and Japan	Soft launch in New Zealand and Australia in October. Developed by 2k Play. 3D graphics. One hero available from the start. Heroes gain levels and learn skills. There is a skill tree, and skills can be reset by paying the hard in-game currency.

Improving the genre

In-game base building: During the first few weeks, shuffling the dozen-or-so buildings around and trying a new base design is quick and easy. Three months later, the village holds around 80 buildings and there is little room to move buildings around. Since players can't design their base in-game anymore, they turn to unofficial websites where they can share their designs and receive feedback. This is bad because 1) it's a feature needed by the players, and it's missing from the game, and 2) the game developers lose touch with their players.

Community: CoC clones do not have a large community. In fact, very few of them have a community at all. The CoC player community is large, and provides valuable feedback and ideas to SC for free! The recipe is simple: the large amount of player suggestions and feedback is filtered by the community managers who then forward the key parts to the developers. For example, a player sketched out a very convenient in-game base builder in the SC forums on September 11. 20 days later SC released a patch with a simpler version of that player's idea. In my opinion, the community is crucial for innovation and polish, and none of the clones have it.

Innovating the core mechanics: The clones must innovate if they want to have people playing their game. A player wrote: if I had invested a ton of money in Clash of Clans, I would find it unlikely that I would do that again in a game that is so similar. I think there is a spectrum of innovation. On one side, Amazing Clan War is a blatant ripoff of CoC (see the CoC vs Amazing Clan War loading screens below). When your power in the game is all about how much time you have been playing for, why would you switch? Most other clones are heavily inspired by CoC, but add, tweak, or twist a couple elements. Yet the core mechanics remain unchanged since Backyard Monsters! For example, all games have the slow meat-shield troops luring defense buildings and the fast and weak troops targeting resources. On the other hand, the hero skill tree in Sensei Wars, the jade mine of Ninja Kingdom, or the 12-hour inter-alliance war in Samurai Siege are very promising PvP concepts, but they're too shy and peripheral to really matter.

Soft launches: CoC clones follow CoC's formula to launch the game in Canada, Australia, or New Zealand one or two months before worldwide release. But since CoC clones are all about time, soft launches give a head start of a couple months to some privileged players. That is probably why SC had no trouble giving new players 500 gems to catch up when they released the game worldwide. Of course the Canadian players who had started playing during the soft launch were not happy - they missed out on 500 gems! How could a game soft-launch and make everyone happy?

Clash of Clans - match-making and retention

Match-making

A player can only be attacked when she is offline. Hardcore trophy hunters are online most of the time, because trophies are mostly gained in offense, not in defense (some gemmers even ignore upgrading their buildings). Thus it can take a while for high trophy players to be given a match when they are attacking. Players complained about the player base being too small to support trophy hunters, when the game came out in August 2012. Trophy hunters kept complaining, and the most SC did was force players to logout at least every 8 hours.

The fact that players can only be attacked when offline has another consequence. When a player logs in while under attack, she does not see what is going on. Instead, she is given a loading screen for up to three minutes. This is inelegant, and it could be improved.

As others have noted, in Clash of Clans, the clans don't really clash. Players of the same clan can not gang up against a rival clan. The only thing they can do is chat together and donate troops to each other. Some even say that the game only provides the illusion of a multiplayer game. Players can not even choose who they will attack. The system gives them villages from other players one at a time, and the player has to next sometimes for 10 minutes to find an appropriate target.

Match-making is also SC's way to remove churners from the game. Churners are called inactives in the community. Inactive bases have full mines, so they provide a lot of loot for little effort. To keep the loot in check, SC removes inactives from the match-making queue every few weeks. Players notice it very quickly: one day, they loot 300k gold per raid on average after 30 seconds of nexting. The next day, it takes 10 minutes of nexting for a 150k-gold raid. I think a lot of players stop playing when they suddenly can't find loot.

Retention and monetization

Achievements such as unlocking the Dragon, reaching a certain amount of trophies, or looting 100M gold, all reward the player with gems. These gems are the only way to acquire builders to speed up the upgrading of buildings. Thus for players who do not spend real money to buy their builders, these achievements are long-term goals. For many players, the game is about the journey, not the end. Achievements also work in the beginning as quests guiding the player through the basic game mechanics.

WoW subscribers and Blizzard's strategy

I found this graph of WoW subscribers from launch until end of 2013, but could not track any of its sources. So here's my version of the graph, with the sources below.

The most interesting part of the graph is that Mists of Pandaria (MoP) is the first reported expansion to suffer from churn a few months before release. Blizzard's CEO says that have become much better and much faster at consuming content, and so they would leave when they're done. Have players changed, or has the game changed? Probably both.

The game has fundamentally changed, and this is a result of Blizzard's design approach of depth first and accessibility later. But as the game becomes more accessible, new content is consumed faster. Veteran players agree.

Have players changed? Probably. When WoW launched in 2004, mobile and social games did not exist. Blizzard has always targeted PC gamers. That segment was growing in the early 2000s, but has probably started shrinking in the last 3-4 years. Free-to-play PC games have also bloomed since the late 2000s, and lots of PC gamers would rather play for free than for $15 per month. WoW remains subscription-based because, according to Blizzard's CEO, it was not designed as a free to play game and would not transition to F2P. Keeping WoW subscription-based is a very conservative and obstinate stance, but Blizzard probably lost faith in F2P after the Diablo 3 real-money auction house failure.

Blizzard is conservatively milking its existing games. But I don't see the pay-to-play PC-gamer segment increasing. So Blizzard will have to adapt their game design or die.

date	millions of subscribers
23-Nov-04	0 (launch)
17-Mar-05	1.5
14-Jun-05	2
21-Jul-05	3.5
9-Nov-05	4.5
19-Dec-05	5
1-Mar-06	6
11-Jan-07	8
24-Jul-07	9
22-Jan-08	10
31-Jul-08	10.9
28-Oct-08	11
23-Dec-08	11.5
11-Feb-09	11.5
7-Jul-09	6
10-Feb-10	11.5
5-Aug-10	11.5
1-Oct-10	12
4-Nov-10	12
31-Mar-11	11.4
1-Sep-11	11.1
8-Nov-11	10.3
9-Feb-12	10.2
31-Mar-12	10.2
4-Oct-12	10
31-Dec-12	9.6
31-Dec-12	9.6
31-Mar-13	8.3
30-Jun-13	7.7
30-Sep-13	7.6

Player Experience Panel, Phillips 2010

Player Experience Panel, Phillips at GDC 2010.

(slides, mostly about FPS games)

Each dot represents the average number of days taken to complete a particular achievement. DLC achievements are farther because DLC were released later (not because they are harder or take more time to complete).

Box gameplay peaks when new DLC is released: new content does increase the number of people playing. However, later DLC packs do not peak as much as the first DLC pack.

Influence of gameplay on skill in Halo: Reach

Huang et al. Influence of Gameplay on Skill in Halo Reach, 2013

Data: 3m player data (entire population) for the first 7 months of Halo Reach, and 70 players in a survey. Mixed methods: use survey data to help explain the big dataset.

Most played mode is Team Slayer: from 3v3 to 5v5, one point per kill, first team to reach 50 kills. Match ends after 15 min.

Player skill metric: TrueSkill's mean μ. More frequent players have bigger μ drop at week 1, but their μ increases faster over the weeks.
The longer the break between two games, the bigger drop in skill. It takes 10 games (3h of gameplay) to regain the skill lost after a break of 1 month.
Most of the top 100 players use the DMR (same range as sniper rifle) and sniper rifle.

Rewards, pacing, and dopamine

Rewards and pacing

There is something mysterious when people read through 1000 pages of Cryptonomicon, a 1999 novel from Stephenson, or through the thousands of pages of Martin's A song of ice and fire. What keeps people reading? Here are some rewards that both authors use throughout their books:

Reward	Pacing	Description
Character progression	Randomly (but fits the story)	The protagonist learns new skills, grows up, meets love, learns a lesson of life, or gets rich. We care about this character's well-being because we empathize.
Action and passion	Every 3-10 pages	When stuck in difficult situations, the reader knows that the hero can not die, so she asks herself: how is the hero going to get out of this alive? (Except sometimes, main characters actually die, leaving us in shock). The same applies for romantic scenes, where we wonder not if, but how it's going to work out between two characters.
Story progression	Every 10-30 pages	Right from the start, the authors put under the reader's nose a bunch of questions: what is happening, why is this guy so mysterious, and so on. The reader wants to know the answers, so she keeps reading until answers are provided, along with new mysteries to figure out.

Looking at blockbuster action movies such as The Bourne Ultimatum, the same kind of pacing emerges: action scenes are followed by the protagonist learning about his past. Then the plot moves on, the character meets new people, finds himself new tasks to do, and back to action scenes.

These rewards and pacing also echo those found in games and mentioned in a class on game balance from Ian Schreiber: gains in character power, discovery of new areas, and progression of the story are rewards to be used at irregular intervals to keep the player engaged. So reading books, watching movies, and playing games seem to provide similar rewards.

Dopamine

Let's dig down to the physiological level of rewards. Most of the articles dealing with the physiological response of gaming revolve around addiction. For instance, a 2012 NY Times article reports that Zynga helps addict millions of people to dopamine, a neurochemical that has been shown to be released by pleasurable activities, including video game playing, but also is understood to play a major role in the cycle of addiction.. To that, the Zynga co-founder replies: Given that we're human, we already want dopamine. And that does not calm things down. So let's look at a less controversial topic: the physiological response of reading. This is not a survey of the field, but rather some picks from a few Google searches - nothing very serious.

First, according to Farland, a current writer, the dramatic structure of stories (exposition - action - climax - denouement) matches the bio-feedback of hormones such as dopamine, adrenaline, and cortisol. He says: As a person "hunts" for clues, or for a way out of a problem, the brain rewards the person by releasing dopamine as a reward. [...] When you reach the climax of the novel, [...] you reach the climax of your emotional exercise. When the story ends [...] your stress is released. The adrenaline and cortisol stop flowing.

As notes from a psychology class on stress can tell us, cortisol is the key hormone of stress. Adrenaline is the hormone that tells our body to be alert. And finally, dopamine is in charge of rewarding our brain.

So what happens when we start reading? Some have guessed that we are having pleasure because reading is a tough task, and our brain rewards us for completing such a hard task. This seems confirmed by a 2001 study who showed that transitioning from rest to reading produces the same increase in dopamine concentration as transitioning from rest to memory-intensive tasks (something cognitively demanding). Although, a 2000 study seems to reject this hypothesis. Something that might be worth investing is whether reading a microwave-oven manual generates as much dopamine as reading an exciting short story where the action starts right from the start.

It may actually be more complicated: the dopamine concentration could actually not indicate our pleasure, but our expectation of pleasure: dopamine motivates us, increasing our energy and drive and compelling us to engage in the pleasurable activity. If everything is as nice as the brain predicted, dopamine levels remain elevated. If things turn out even better than the brain hoped, dopamine levels are increased; we engage in the pleasurable activity even more vigorously. If, on the other hand, the activity is less pleasurable than we thought it would be, dopamine levels plummet.

Back to games

So, what can we conclude about games? First, much like movies and books, reward us by generating dopamine when we succeed at a difficult cognitive tasks such as a head-shot in an FPS, or a successful Chess trap. That is pretty close to fiero, and in fact, Bateman already suggested in 2008 that fiero is a cocktail of epinephrine and dopamine. So, nothing brand new here.

Perhaps more interesting, long-term enjoyment seems to require the dopamine to yo-yo - which is bad. Let us assume that our brain produces dopamine by expecting a nice event. Then when our brain is done with those nice events, the dopamine level will decrease. That would be a horrible yo-yo if there were only one kind of event, but books, movies, and games, have at least three: action, story, and character events. So alternating events and interweaving them in such a way that our brain is always on the lookout may keep dopamine, and pleasure, at high levels instead of making it yo-yo.

There might be a few gotchas, though. First, the magnitude and the lifespan of the dopamine burst depends on a lot factors. Grinding monsters for a drop is only pleasurable so long as the brain is expecting that item to drop. After a couple days, the dopamine is all gone, and it's just boring. Some studies should investigate the average magnitude of an "achievement unlocked!", or the lifespan of a boss kill or a level up.

Second, dopamine bursts may not stack. The brain may be too busy expecting a story progression event that it may ignore, or even worse, be displeased, by a character progression event: it was not expecting it! This limitation does not seem cognitive, but rather emotional, so maybe people with higher EI could stack expectations and dopamine bursts more easily than indifferent people who say "it's just a game"?

Third, and to finish as we started, with books: Stephenson and Martin alternate their character viewpoints chapter after chapter, possibly to keep the reader's attention. By the way, they rarely handle more than 7 characters at the same time, since 7 is a magic cognitive number. These changes in views often cut short the action, so the reader may get frustrated, or even vexed for being tricked to continue reading by the exact same three mechanisms every chapter. It is interesting to know why people keep reading those long books, but that last reason is exactly why I stopped.

Gold buying patterns

Picks from a paper I wrote about gold buying patterns for FDG 2012. The data comes from an online questionnaire completed from March to May 2010 by 2800+ WoW players from around the world. Unless mentioned, all results are significant with a p-value below 0.01.

• Overall, 14% of people have ever bought gold.
• Men are twice more likely to buy gold than women (17% vs 8%), but there is no difference between Asians and Westerners.
• Achievement increases the likelihood to buy gold, while immersion decreases it. The effect of achievement is stronger on men.

• 12% of people who only play with people they know IRL have bought gold. This ratio increases to 15% for people who play with both RL relatives and friends made IG, and to 21% for people who play only with friends they made IG.
• Overall, people who have taken longer breaks from the game are more likely to buy gold.

• But really, it's a big mess to know which variables influence gold buying: in the correlation graph below, vertices represent variables, and edges bearing positive/negative values indicate positive/negative correlations between two variables. Values closer to 1 in absolute value indicate higher correlations.

• That's where GLM come in handy: they account for interactions between variables in regressions from multiple variables.
• Controlling for all other variables, the odds of buying gold increase when playing on a private server, being a man, having frozen one's subscription, having made friends IG, playing for achievement, and having played the game for a long time. On the other hand, the odds
• Controlling for all other variables, the odds of buying gold decrease when having had a college education, playing for immersion or socializing, and playing with cousins, siblings, or spouse.

MapReduce for MMOs

MapReduce is a powerful tool to parallelize batches of computations. MMOs may sometimes have to run batches, but from what local game companies tell me, nobody in the game industry is currently using MapReduce. I guess, this is mostly due to studios not knowing what to do with it. Here are some examples.

Business intelligence

Basic metrics such as weekly play time or stop rate can give a rough perspective of the retention of an MMO. These metrics can be estimated with a couple SQL queries on dumps of the production database(s). It starts taking more time and effort to distinguish accross server shard, faction, race, or class. Still, a SQL script running for a few hours can do the job. Fancier analyses such as machine learning or social network graphs explorations take even more time and effort. MapReduce can be used to tune machine learning algorithms through Mahout, and even to process graphs (Google's Pregel also seems interesting for parallel processing of graphs: the Pregel version of PageRank takes 15 lines of code).

Detecting bots, hacks, or gold farmers is not as straightforward, but I think it is doable. First, the typical deviant behaviors have to be determined and made explicit by humans. For instance, speed-hackers send too many messages per second to the server, while gold farmers interact with less players, but more intensely, than normal players. Then, detecting deviant behaviors can be a machine learning classification or a graph parsing problem. In both cases, MapReduce can help.

Game-specific

Matchmaking and ladder: Some pre-calculations or updates to parameters of the ladder and match-making algorithms could be done offline by a small MapReduce cluster. A player's skill is unlikely to change much in 12 hours, so a cron task could run the job twice a day. According to Josh Menke from Blizzard, matchmaking involves gradient descent or Gaussian Density Filtering. Not sure whether Mahout supports GDF, but gradient descent is supported.

Tuning and balancing can take days for system designers. MapReduce could do that automatically: each mapper job is given a particular set of system parameters: player 1 has skill A (cost x SP and inflicts y damage) and skill B (cost z SP and heals w HP), player 2 has skill C (...) and skill D (...). Mappers run a few hundred Monte-Carlo simulations of a player 1 versus player 2 match with a fixed set of parameters (player1:A,B; skillA:x,y; skillB:z,w; ...). When done, mappers pass average statistics (win/loss ratio, average amount of gold at the end of the match, ...) of the 100 matches to reducers who sort them. The interesting configurations for balance are those with a win/loss ratio close to 50%. Naturally, this brute-force way of balancing assumes a proficient AI, and designers will still have to tweak the configurations returned by MapReduce so that they feel fun.

Practical concerns

Engineering detail: MMOs have hundreds of shards, but really only one MapReduce cluster should be needed. Each shard could send its jobs to the MapReduce cluster when it needs them done, and wait asynchronously for the MapReduce answer on a particular port. If the MapReduce job uses data from the production database, producing a daily dump may induce a temporary extra load on the shard's database machines, but this should be fine during empty hours.

MapReduce can be a double-edged sword if overused. Exploring the parameter space of learning algorithms too aggressively may lead to less accurate models.

Edit: Some people have been using MapReduce for analytics: mogade's platform and keighl have been using it through mongodb, but it's more of an engineering constraint (scatter-gather queries in a nosql DB to build a ladder board) than an analytics or machine-learning endeavor.

Impact of updates on retention

It does not matter if WoW has around 12 million subscribers. What matters is how many users connect monthly to the game. As shown in my paper, half of the 40% of players who stop playing the game for more than 6 months never freeze their subscription (ie they keep paying but never login). Since they pay, some would consider them active.

Who are those inactive people? It is hard to guess. Therefore, let's flip the question: what are the patterns of (in)activity for various player demographics? By player demographics, I do not mean achievers, explorers, etc. from Bartle or any psychologically-based categorization of players. I would rather focus on the extent to which people play in relation to the game updates, in WoW raiding more particularly.

Category	Ratio of total player pop. (estimate)	Behavior
Pro gamers	<0.05%	Intensive play on PTR right before update deliveries. After four to six weeks, all heroic-mode bosses have been downed by their guild (with eventually some world firsts). Then they stop until the next update.
Dedicated raiders	<10%	This category could be called hardcore, but hardcore is an umbrella term that does not mean much. Those players eventually try out new bosses on PTR. Server-firsts are their goal. They raid three to seven times a week, and spend a considerable time looking for strategies or theorycraft data. Once active guild members have received most of the top-ilevel gear, the raiding activity decreases, and people log in less often. They take a break after 3 to 6 months depending on how dedicated and skilled their guild members are.
Amateur Raiders	>90%	Continuous raiding. They take their time to raid (once to three times a week), and their raids look more like spontaneous pick-up groups than organized expert guilds. Every body complains about the few who really do not pay enough attention and supposedly cause the wipes. Only patient officers spend time reading strategy guides and coach their guild members. These players have content at least until the next update.

Since amateur raiders make up the bulk of the WoW player population, this slow release cycle does not apparently harm the WoW player base so much. However, it seems to me that a diverse range of players (dedicated + amateur, for instance) gives more stability to the player base. Therefore, it might be worth it trying to retain dedicated raiders longer (say, a year's worth of content for each biyearly expansion).
Below, a totally fake graph to show how all this would look like.

Limitations

These qualitative categories may not be the most accurate, but they reflect trends observable in the game, in the few informal player interviews I conducted, and in quantitative studies.

The speed of updates may affect the retention rate of particular player demographics. Blizzard is known for taking its time to release well-polished updates every other year. SOE, on the other hand, has been releasing extensions for EverQuest three times faster: every 8 months on average. What does that mean for retention?

Each player category is affected differently by the gameplay (PvP vs PvE vs PvPvE vs sandbox vs ...). In this article I focused on PvE, but maintaining a top position in the PvP ladder on one's server takes dedication. Another example: the number of EVE Online players has been increasing steadily up to 350k subscriptions, possibly because fans of this kind of gameplay have to be very dedicated. Can these players be considered loyal, though?

[Literature] Retention in WoW

Thomas Debeauvais et al. 2011. If you build it, they might stay: Retention systems in World of Warcraft. In FDG2011.

I looked at what keeps people playing WoW and which mechanisms retain most effectively which kind of players. Here are my picks from the paper I wrote and sent to FDG2011.

Around 2800 WoW players from Europe, North-America and Asia completed an online questionnaire. Player commitment (and therefore retention) was measured by three metrics: weekly play time, ratio of respondents who have ever stopped playing, and number of years spent playing WoW. All the results mentioned in this article are significant with a p-value below 0.01.

• 96% of respondents have been playing WoW for more than a year, 70% for more than three years.
• 23% of respondents have stopped playing for more than 6 months and have never canceled their subscription - they keep paying even if they do not play!
• On average, people play 23h/week. Asians play more than Westerners. No noticeable difference between men and women.
• Achievement and social actions are motivations that increase the weekly play time. Immersion does not influence the weekly play time.

• Asians are more immersion-oriented than Westerners.
• A higher guild rank (officer or GM > basic member > non-guilded) increases retention.
• Women play with people from real-life more than men.
• People who play with their partner play less than single players, but more than players not playing with their partner. They also stop playing less often.
• 13% of players have found a real-life boyfriend/girlfriend/spouse in WoW.
• There are more players over 45 than players who play more than 40 hours per week (another sample may contain a different ratio, though)