Tuesday, November 3, 2009

Testosterone and Competitive Play

Lately I've been digging into research on testosterone. Over the past decade, scientists have been placing players in competitive situations and then measuring how their testosterone fluctuations predict future behavior. What you find from looking at the studies is that both winners and losers will leave your game if they are placed in a set of predictable situations involving dominance, luck, and friendship.

There are four points that have experimental support:
  1. How playing with friends affects the testosterone in winning and losing players
  2. How playing with strangers affects the testosterone in winning in losing players
  3. How perception of the role of luck or skill in the outcome affects the testosterone of players.
  4. How players differ by pro-social or pro-dominance inclination.

1. How playing with strangers affects the testosterone in winners and losers
When strangers play a competitive game based off skill, the results fit the common sense understanding of winning and losing.
  • Winner testosterone increases. Dominance and/or aggressive behavior increases. Dominance is defined as behaviors that are intended to "gain or maintain high status" (ref) Physical energy increases (and in some cases men become aroused.) Winning is exciting.
  • Loser testosterone decreases. The losing player attempts to avoid fighting the same opponent, even in situations challenges unrelated.
This is the classic description of winners and losers in a competitive game. The winners get a huge rush from beating the strangers and the losers are sent home with their tail between their legs, humiliated and subdued.

The upside
Beating strangers is a guaranteed source of entertainment. If you want a highly reliable, inexpensive means of making your game fun, toss some strangers together in a game of skill (it barely matters what sort). To boost the emotion even further, place the winners on a high status pedestal. Voila, instant fun, at least for the winners.

Typically designers look for 'fun' in a game and then build the game around what we find. The hard fun or fiero is easily detectable on the faces of the victors and acts as a clarion signal of fun. This overt signal has driven designers to create hundreds of competitive games between strangers. "Hark! Here be fun!" and we flock like moths to the flame. Our fun finding, hill climbing algorithm is predisposed to overemphasize competitive play due to the strength of the delight exhibited by winners.

The downside
Yet there are clear tradeoffs that occur when we go down this design path. Losers leave. First, they know that they cannot gain status by pursuing the game, especially against the winning players. Second, if there is some way for winners to communicate, losers are subjected to degrading displays of status. Losers may react in turn with defensive behaviors if they feel they cannot escape. Especially in games where only a few people can be winners, your player retention will suffer.

The result is an intriguing purification of the community. Only the elite winners stay around. This elite community creates an even more competitive environment that in turn creates and drives out more losers. New players attempting to enter into the community are inevitably of low skill compared to the hardened veterans and are immediately classified as losers. They also leave. Competitive games slowly boil their community down to an elitist core that actively resists and inhibits audience growth.

2. How perception of the role of luck or skill in the outcome affects players
Notice that the above case applied only to games where the loser felt that they were participating in a game of skill. The testosterone response changes when players feel they are playing in a game of luck.
  • Winner testosterone does not increase: Winners feel that their victory was not a true demonstration of superiority.
  • Loser testosterone does not decrease: Losers feel like they still have a chance of winning.
Luck is another name for an unexpected environmental factors outside the control of the player. If humans were to fall into a funk every time they lost due to the weather or an unexpected mishap, we would not have survived very long as a species. Luck turns a loss into a lesson about the environment or game mechanics, not a lesson about which player is superior to another player. As such, our innate social dominance systems fail to kick in and the social penalties from losing are avoided.

The upside
By introducing luck into a game, you can mitigate the ill effects of losing. Losers are often willing to give the game another shot. The fact that humans are notoriously poor at judging their probability of success plays out in the game designer's favor, since even poor players will think they still have a chance of winning.

The downside
Winners fail to feel the rush of victory. Strangers playing against one another in a game of luck will often complain that the game is 'cheap' or 'not a real test of skill'. Many highly competitive players will consciously avoid competitive games involving a high amount of luck since such systems reduce the psychological benefit of winning. What is the point of playing against strangers if you can't beat them into a pulp and demonstrate your dominance?

Pure competitive games of luck between strangers are rare beasts and for good reason. They manage to keep losers around, but the games hardly ever considered fun. Some gambling games may qualify (such as horse betting), yet it is telling that the vast majority of players lose.

An example
Mario Party is an example of a high luck competitive game. The game awards crazy bonuses that appear arbitrary and many games end up with the person in last place winning because they happened to have landed (randomly) on the correct square. Due to the high degree of luck is easy for losers to claim that the victory doesn't matter. The relative status of players barely changes over the course of the game.

3. How playing with friends affects the testosterone in winners and losers
So far, the the previous two studies of competition shouldn't be of much surprise to folks that have designed competitive games. However, the response of players is quite different if they consider one another to be friends. The following is what occurs if friends face off in a competitive game.
  • Winners testosterone decreases. In essence, dominance behavior dips sharply if you win in front of friends. Friends are generally are people you need to get along with in order to live your life. Imagine for a moment, if you were to win a game and then yelled at them to lick your boots (and you meant it). They probably wouldn't be your friends for very long. Our innate social response is to repress our instinctual dominance urge so as not to damage our friendships.
  • Loser's testosterone briefly falls and then recovers: The loser is under threat of being put in a low status position. However, once they receive signals that their trust in their friend is justified, they have no reason to fear a loss of status.
If dominance responses are missing, where is the fun? In general, you see both winners and losers focusing on bonding activities after a competitive game.
  • They discuss the great shared moments in the game. Shared experiences create a common ground between players that they can reference in the future.
  • They compliment one another. Compliments are often reciprocated, creating an opportunity to build mutual respect and indebtedness.
  • The winner claims they got lucky. This defuses the notion that the winner is in some way dominant or higher status. They frame the game as one of luck which makes the loser feel much better.
  • Typically, the winner does everything they can to avoid rubbing their victory in the face of the other player. And the loser does everything they can to not dwell on their loss of status. We even have names for friends that engage in inappropriate dominance behaviors. We call them 'poor winners' or 'poor losers'. Players that behave in a manner conducive to bonding are called 'good winners'. It is rare that you hear the term 'good loser' since the loser is the victim to be consoled.
  • Mutual smack talk is a form of bonding: This can be confusing for the untrained observer, because good friends will often act like they are engaging in dominance behavior by using smack talk. Yet this is just for show. The moment the smack talk actually infringes upon existing expectations of status, the mood of players will change abruptly. You'll often see accusations of one player 'taking it too seriously.' It is a good demonstration of trust to play at dominance, but to actually assert dominance between friends is considered out of bounds.
Bonding requires some form of communication channel. In a game played in a living room such as Mario Kart or a board game like Carcasonne, there are plenty of ambient opportunities. In PC games, text is the common channel. In console games, chat serves this purpose. Game mechanics can also be used as a form of in game communication. Tagging in Counterstrike is a good example of a game mechanic used to demonstrate status or shared affiliation.

The upside
With the increased popularity of couch gaming on the Wii and social gaming between friends on platforms such as Facebook, understanding the dynamics of competition between friends is critical to creating a successful game.

The most important realization is that typical form of 'fun' that we associate with competitive games is either reduced or turned into a negative experience. Competitive game play with friends becomes less about winning and more about shared experiences. This is a very different emotion. The ability to tell player stories, communicate, discuss and joke with one another are all features that enable the core delivery of value to the player. In some sense, the actual competition is secondary to the bonding that occurs around the activity. The 'fun' that comes from playing with friends is completely different than the 'fun' associated when playing with strangers.

The downside
Again, you can't rely on 'hard fun' to deliver the same jolt as you would in a competition between strangers. The simple switch from playing with strangers to playing with friends results in such a shift in player psychology that you now need to rethink your reward and communication mechanisms.

It is easy to be fooled. The mechanics of the game like Unreal Tournament when played with strangers or friends are apparently identical; you shoot and you move. Yet the experience ends up being radically different. It turns out that existing social relationships and ambient communication methods are as much a part of the game as is the level design and the bullet physics. All too often I see designers building a game that they play with their buddies on the dev team. The group knows one another, can yell out in victory and ends up having an immense amount of fun. Then that same game is released online and immediately strangers begin griefing one another and creating an actively offensive elitist environment. The social graph of the playtesters is not the same as that of the actual players. As a result, the playtest sample is massively flawed.

Here's a little chart to keep it all straight:

An example
Let's return to Mario Party. Why would anyone play a luck based competitive game that provides poor rewards for winning? One clue is that Mario Party is always played with people sitting together on a couch. It is a social game about improving your friendship, not about beating the snot out of someone. Due to the game being played in person, there is immense communication between players and almost all communication is focused on bonding over a shared experience. The key gameplay yields is social fun, not hard fun.

It is perhaps not surprising that Nintendo multiplayer franchises have been slow to move into the online world. Most Nintendo games are designed to be played with friends. Due to low concurrency, synchronous play models and a lack of scheduling, most console gaming services are populated by strangers playing with strangers. Changing the dominant type of fun that forms the core of your game changes your value proposition to the player. This is a major brand mismatch that likely needs an entirely new franchise (such as Halo), not a minor design tweak.

4. How players differ by pro-social or pro-dominance inclination
To complicate matters, there are in fact two distinct populations of players in all these studies. The first are pro-dominance players who are predisposed to react to situations in a dominant fashion. They tend to have a higher base level of testosterone in their system and their level rise or fall more strongly in situations where they win or lose.

The second group are pro-social players who are predisposed to react to competitive situations with a focus on relationship building. In general, they have a lower base level of testosterone. Intriguingly, they do not experience the same misery of failure. In some sense, they aren't playing to win so they don't mind losing. In fact, some studies suggest they even experience increased stress and reduced performance on complex cognitive tasks when they are thrust into a high status position. Winning is a punishment.

Age may also be a factor. Testosterone peaks in the late twenties and drop steadily after age 30. By age 40, 19 to 47% of males fall into the low testosterone category, depending on the accepted cut off.

The downside
From a game design perspective, this split in your population has some interesting implications. When you create a game that rewards players by winning alone, there are two groups that you fail to address. The first is of course, the losers. The second however, are pro-social players that are motivated more by forming relationships than by demonstrating status. You can give them opportunities to 'be the winner', but these rewards will fall flat.

These patterns of competition give designers some useful tools.

Note 1: Your design should explicitly differentiate between friends and strangers
You need to differentiate up front between friends and strangers in your design. If you fail to separate these two populations, you'll end up creating system that inevitably alienate multiple segments of your player base. Many of the problems stem from how communication channels are used by each group.

If you create a game for friends:
  • Winning strangers will use the communication channels intended for building reporte to instead act out their dominance and aggression urges. Teabagging is an example of a humiliation behavior that tends to encourage losers to leave.
  • Losing strangers will use the communication channels to denigrate the winners or claim luck or environmental issues were at work. This makes the winners more likely to leave since this is not the 'good job!' pat on the back they were hoping for. Instead of bowing and fame, they are greeted with yells of 'cheaters' and 'lucky'.
If you create a game targeted exclusively at strangers
  • Due to lack of communication channels, winning friends will have no way to reduce the bite of their victory. There is the risk of permanently damaging your relationship with the loser.
By separating friends from strangers, you can offer each population rewards and game mechanics appropriate to their desires. Winning strangers can be complimented in isolation. Losing strangers can be given feedback that emphasizes the luck of the situation and their increased future chance of victory. Friends can be given communication tools that allow them to bond.

Note 2: Games that focus on playing with friends result in stronger retention across a broader audience.
  • Friends encourage other friends to join since they want to share the experience with them in order to increase their bond.
  • Friends tend to encourage existing players to play more since they want to deeper their bond.
  • There is only one class of player that is alienated by bonding oriented play: pro-dominance players that are not able or willing to play amicably with friends. This is arguably a big group (upwards of ~50% of males age 14 to 39) Yet this is distinct minority in comparison to the broader population.
This insight gives some indication why asynchronous social network games grow so rapidly. People typically play with friends and are predisposed to communicate their game experiences and feel social pressure to repeat them. In contrast, competitive activities between strangers tend to result in a steady decline in player populations.

Note 3: Test with strangers and friends separately
As tempting as it is to test your multiplayer game with the readily available team playing within shouting distance, understand that you are fatally polluting your data. Larger scale online tests that allow strangers to interact and figure out how to dominate and insult one another will yield a much more realistic understanding of the culture that will evolve out of many competitive multi player game systems.

Note 4: If you must include communication channels in your online game, create a design that turns strangers into friends.
If you include rich communication channels in a competitive game, strangers will use them to exert their dominance. The way around this is to explicitly create groups where people act as friends. This leads to bubbles of cooperation even within a competitive game.
  • Assign players to a common affiliation. Counterstrike does this by having sides that you join from the start of each mission.
  • Create a common goal: Horde mode in Gears of War does this by giving players the goal of surviving the onslaught together.
  • Create a common experience of suffering or joy: In Eve Online, players partake in vast highly destructive battles. Even after vicious losses, companies still stick together since the suffering gives them a visceral common experience that strengthens their bonds.
  • Offer opportunities for reciprocation: In Left 4 Dead, players can help one another if they are in trouble.
  • Provide channels of communication: In Farmville, players can send messages to one another in game and via Facebook notifications. This helps players negotiate group norms and bond over shared experiences.
  • Allow individual choice: In WoW guilds, players actively choose to participate in a particular group. Players that allowed to choose freely will have a greater affiliation than players that are forced to rely on other players. I find designs were performance is improved with other players works better than ones where players are punished if they do not cooperate.
When we design a game, we are constantly on the lookout for 'fun'. However our ability to identify and augment fun is only as good as our mental model of what fun looks like. Our commonsense models of competition overvalues the delight expressed by winners and undervalues the reactions of other player populations. By adopting a more sophisticated model of how winners and losers react in various situations, a designer has a much better chance of knowing why their design fails and how they might fix it.

The data I've covered is not complete for all populations. For example, there are fewer studies looking at how testosterone changes in women. Though we commonly think of it as a 'male hormone', testosterone is actively produced by both sexes and appears to serve similar purposes in regards to dominance. However, not all behaviors found in men have been reliably produced in studies involving women. Nor have all the studies been validated on older populations, different cultures or children. Scientists have a tendency to use male college students because they are readily available and it is much easier to measure their testosterone. This can skew the results. The solution is to use these guidelines as a starting point and then continually test your hypothesis about competitive play. Put your game designs in front of a diverse group of players and see if they react as you expect. By looking through the lens of a richer mental model, your informed experiments will guide your game in the right direction.

My personal take on these studies is that there is vast potential for new pro-social competitive games. The market took an odd turn for a short while:
  • Early consoles involved 2 to 4 players gathered around the TV. Play was primarily social.
  • We lost the focus on playing with friends with the advent of online play and low concurrency platforms. Since we were playing with strangers, the primary class of fun switched to games of dominance.
  • The advent of social networks again allows us to target online multiplayer games at audiences guaranteed to be friends.
Now we have a fresh opportunity to design friendly competitive games that build relationships instead of breaking them down.

take care