Game Placebo

Homeopathy, for those who don’t know, is a form of alternative “medicine”.  It involves diluting an active medicinal ingredient into a solution so many times that there ends up being a mathematically near-zero chance of the solution containing even a single molecule of the active ingredient. Yet the United States spent 3.1 billion on homeopathic products in 2007, which can seem pretty strange to someone who understands what it really is.

Now, clearly homeopathy doesn’t “work” in the sense of having actual medicinal effects.  Homeopaths have tried to claim that it does by conjuring all sorts of bizarre theories, one of the most common being that “water has a memory” – it can “remember” the molecules that used to be in it, and somehow that memory has an effect. Whatever – pretty obviously nonsense but that’s not what concerns me here.

Delicious. It’s impossible for anyone to tell the difference between pure sugar pills and homeopathic drugs. That’s because they are the exact same thing.

What does interest me is the fact that homeopathy does work in the sense that people think it works. The placebo effect is very powerful (for some kinds of ailments, anyway), and having a “school of medicine” with practitioners all telling you that this sugar pill will stop your headache may alleviate the pain, since pain is understood to be a highly subjective sense that can be affected by the state of mind of a person.  The simple concept that “you’re being taken care of now, everything will be OK” may bring comfort to that person.  Here’s a really great Derren Brown video showing the great power of placebo.

More often, though, is probably the following scenario:  the person gets a headache, takes a homeopathic remedy, and then the headache goes away. This is erroneously marked a “hit” in that person’s mind – the remedy worked!  Nevermind that it’s incredibly likely that the headache simply happened to go away.

We tend to remember the hits and forget the misses in this way.  Psychics and astrologers have been using this against us for ages, and techniques such as cold reading are a very real method that can be used to make (help?) humans believe that which is almost certainly very fake.


The Desire to Believe

I should perhaps use the term “help” rather than “make”, because there’s another important aspect to this problem, which is that people generally want to believe the sorts of things that work because of placebo.  The very idea of alternative medicine is a pleasing narrative that appeals to both our modern first-world guilt (“I’m hip to these ancient/third-world remedies, I’m not really one of the insensitive conquering Westerners!”) and our justifiable distaste for some of the actually evil or irrational policies taken up by mainstream medicine in the past or the present.

We also have a desire to simply “have been right”.  After all, we went against the grain and chose an alternative.  We invested in this alternative, and it would definitely cost us a little more to end up having to admit that it was a big waste of time and money.

Finally, we may be evolutionarily tuned to “see agency/causality where there is none“.  With regard to survival, it’s in our interests to imagine agency behind a rustling bush and run.  Those individuals who did otherwise were taken out of the gene pool over time after enough rustling bushes turned out to be lions rather than raccoons.

The famous Jesus toast. We’re bread to see patterns.

So, before I move on, it’s good to review these two specific reasons that we sometimes believe things without reasonable evidence:  the desire to believe, and our biology making the job easy on us.



So how does any of this relate to games?  I’ve been struggling over some concepts with regards to randomness in games for a few years now, and I think I’ve come to understand some things that might be helpful to others.

Before I go on, a quick disclaimer:  I’m talking about games in a specific context here.  I’m talking specifically about interactive systems that involve winning and losing and decision making – systems wherein you express your level of understanding through those decisions and meet a win/loss outcome.  This Casino App List explains what I mean better – or there’s a bit more on my definition for games here.

Games with high levels of output randomess – i.e., card draws, dice rolls, or other random things that get between a player and his agency – tend to have more random outcomes. This isn’t really contested. A professional Poker player will win a smaller % of games against a new player than a professional Chess player will against a new player.  There’s more randomness in the system of Poker, so the win percentage naturally moves towards 50%.

This guy is really bad at the game

What’s much harder to pin down is an individual game’s outcome. If you won that game of Dominion or Settlers of Catan, how much of it was because of your actions, and how much of it was because of sheer chaos? It’s extremely hard – practically impossible – to pin this down, because of layers upon layers of many decisions that were made.  Maybe one of the decisions you made was somewhat poor, but it was only because you were put in a bad position due to randomness. And maybe that’s not so bad, except for the fact that maybe your opponent wasn’t put in any kind of compromised situation at all – his randomness dealt him a nice steady flow of problems that he was used to dealing with.

I don’t care about this question for some alpha-male “who’s better” reason. I care because if we can’t determine the cause of the outcome, then feedback for your agency is diminished. And the agency-feedback loop is what games and all kinds of interactive systems are all about.

Of course, the defense is that over a large enough number of games, it’s not really a problem. The better player wins most of the time. However, that doesn’t detract from the phenomenon I want to explore – that of imagined agency.


Imagined Agency

I’ve talked about this a bit before back on the Dinofarm blog. The idea is that after and during a specific highly random game (i.e. almost any card game or game involving dice, virtual or otherwise), players tend to imagine that they have a different level of agency than they actually do.

For instance, if a player is doing well, he’s quite likely to attribute that success to himself.  Of course, he won’t do this if the random swing is so strong that it’s patently clear.  For example, if a player lucks out and finds an extremely powerful weapon early on in a roguelike, they’re likely to point that out whenever they explain how their game went.  “I did really well – I mean, I got this crazy sword early on, so most of the game was a total pushover…” they might say. But, barring freakish outliers like that, they’re likely to take  credit for a successful game.

It makes sense. If we watch children playing Candyland, Trouble, War, or any other 100% random children’s game, we can see that despite the fact that the game is completely random, the children are assigning agency to their actions. They take credit for the 6 they rolled. We can clearly see that this was not a product of their skill, yet they can’t see this.  It’s similar to how Tic-Tac-Toe is interesting for children because they can’t see the solution, whereas to us adults it’s boring because we can.

In Candyland, everyone loses


So then doesn’t it also make sense that a game with a large degree of randomness – say Poker or Dominion – would have a similar effect? Some players will see through the illusion and point out that there is no direct line between player input and game feedback. These players will see – as clearly as every adult sees for Candyland – that the outcome of the game is not determined by the players, but rather by randomness.

In short, highly random games tend to be built on a similar illusion – a placebo effect, if you will – of agency.


The Game Placebo

What determines whether a person is able to see through a highly random game or not?  Well, it’s similar to the factors that cause a person to see through a fake drug’s illusion.  Firstly, a base level of intelligence does factor in. If the person just doesn’t even have the faculties to start asking the question “is this real”, as most children do not, then the question won’t get asked.

But the second aspect, which interests me more, is the inclination to ask that question in the first place.  Most gamers I know do not see the indirect relationship because they do not ask the question in the first place.  They aren’t thinking along the lines of “tracing the steps between my actions and the outcome of the game”.

This is understandable, of course, as most people play games to simply have a good time, and sitting there and analyzing exactly what’s going on probably isn’t in their interest for doing so. But as the film critic Plinkett often says, even if these players don’t consciously notice this disconnect, their brain does. What I mean by that is that even if they are consciously ignorant of the mechanistic processes that make up the play of the game, the effect of those processes is real and undeniable, and your brain feels them fully.  If your brain is able to detect that lack of agency-feedback, it will diminish its interest in the game. In short, your brain gets tired of the randomness, even if you don’t, and many gamers will find themselves simply wanting to move on to other games without knowing exactly why.

To paraphrase something Richard Garfield said at PRACTICE this year, something along the lines of “you don’t have to be an engineer to feel the effects of a faulty bridge”.


The Problem

So what’s the matter?  Why can’t I just let people enjoy their games for a short time and then move onto something else when they get bored?  There’s two problems with this whole setup.

Firstly, game designers should basically never, ever be in this camp.  Game designers should be fighting at all times to be fully aware of the actual stuff that’s really happening when people are playing games.  Game designers not being able to discern between the illusion and the reality is similar to a doctor who really can’t tell the difference between medicine and sugar pills.

The second issue is that this kind of creates a loop of consumers buying games, getting excited about them, and throwing them out after a year or so, and then moving onto the next one.  This is great for companies who want to sell games, but it’s a treadmill that wears people out.  After enough cycles like this, players will eventually feel as though it’s not even worth looking – that they already know how it’s going to go.  They’re going to get a game, get super excited about it, grok it, and then throw it away.

There is certainly value to this placebo effect.  It can attract people in the short term, and act as a “shortcut” to interesting or otherwise “fun” game designs.  However, it’s just worth being conscious of the downside as well.  If we’re talking about how to achieve great game designs, we should probably rule out shortcuts.

I think that the more we rely on this illusion to get people interested in our games, the less healthy our gaming ecosystem becomes.  We should be fostering games that can stand up to a lifetime of play, or more.  We should be striving to make games whose depth can be felt.  We should want players to know that if they invest time in learning this game, it will continually give back to them.  We should make games that strive to become a facet of a person’s life.  This is how we can establish trust between designer and player, and through that trusting relationship, everyone really wins.



  • rs

    I’m seeing a block of spam text on the top of the page when opening your site with JavaScript disabled or when doing view source on any page. What’s that about?

  • keithburgun

    Wha??? This is the first I’ve heard of this… I just did view source and it looks fine to me?

  • rs

    Here’s the source I get when I load your page:

    Try opening the site with a browser in incognito mode, whatever is going on may be reacting to your login cookies.

  • ooli

    Yeah I saw the same spam in the source code. Pretty scary. Not on the site hopefully.

    btw Keith, what about Pirate (with Civ it must be the game I played the most); It’s sort of special with a lot of randomness (type and number of ship on your way, prices, land fight) but you can negate most of it (flee, dont rely on trade, dont attack cities).

    And a Civ game can be ruined with a unlucky beginning location (too small island, too close to aggressive AI, lack of river, etc)

  • keithburgun

    I haven’t played Pirate, can you link me to what you’re talking about specifically? And as for unlucky Civ starting locations, I’ve played 1, 4 and 5 the most and that has never really happened to me. But if it does happen, that just means that the rules for generation/starting placement need improvement.

  • ooli

    Pirates! By Sid Meier :

    here’s some gameplay footage from the original (it was remade as of 2004):

    From the very beginning of the adventure: The number of crew men available, the events you learn about in the tavern, even city country ownership are partially randomized, etc, etc. In the end it felt far less luck dependent than Civ.

  • Dasick

    Well, most of the time, map generation in Civ is pretty good at creating fair, but interesting and varied. There will be situations that are unfair, either detrimental to you or advantageous. I honestly think that with input randomness you *will* get those fringe cases no matter how good your engine is, it’s just a matter of frequency. Bad level generation creates those fringes 1 in 10 times you play (looking at you Spelunky) and good level generation will make you see them maybe once in a million. Example, it’s possible to play a game of Tetris where you *never* get a line piece, or where every terinome you get forces you to make gaps in your structure, until the well fills up. Never happened to me or anyone I know, but still possible.

    The player would identify the unfair fringe case and either lose quickly, or restart. The most important thing though, is that the player shouldn’t play for several hours only to find out that the odds were stacked against her the entire time, and suffer the most frustrating defeat because of a few bad dicerolls.

  • Dasick

    A couple of opinions on randomness, either input or output.

    Randomness works as a framework, not as a mechanism. I think this is a better distinction than input/output randomness.

    I think the most important point about randomness is that having too much or too little of it is bad (in single-player games, and some multiplayer games as well). It’s a balancing act is what I’m saying, and you’re probably going to fall short on on side of it anyways, so it’s a good idea to prioritize.

    There probably exists a sub-category of single-player games where the measurement of your decision making is your ability to surf the noise and chaos. I’d say games have to do that to some degree in order to encourage exploration of the mechanics, but this type of game (that doesn’t exist in it’s purified form yet, as far I know) is focused on surfing total chaos. This can work in a multiplayer game if the chaos is the same for both players (using the same seed for random operands, using the same dice rolls and card draws etc. In auction based Monopoly, what property is up for sale is random, but everyone get’s an equal chance to nab it)

    Output randomness can be worked into solid game design, so long as the player can expend resources to increase probability to be 100%.

    Output randomness of little arcs (tactical decisions) can work as input randomness of bigger arcs (strategical decisions). In a Civ-like, the outcome of a battle/war can and should be random/abstracted in order to create an interesting decision field for the meat and potatoes – diplomatic and trade relations, tech research and city management. You just need to work in a couple of fail-safes to prevent fringe cases from popping up everywhere, or…

    Put a barrier between the measurement(score) and randomness, so that it affects the score as little as possible. In our Civ-like the outcome of the war/battle shouldn’t give points. Instead, the meat and potato sphere should be the one giving out points (city size/happiness, vassals/allies, researching a new way to get point as opposed to a new ability). Even if your civilization was conquered and is now being oppressed, the player still needs an opportunity to gain points.

    I think Settlers of Kattarn is an ok base for a game, it’s just that the tech cards create fringes too often (you can draw a card that just *gives* you a victory point… wtf? what if you can apply such a ‘civics’ card to a settlement or city, it gives an extra resource to settlements, and for a rolled-up price of 3 resources, you can activate it on a town for an extra VP) and it gives no control (can’t choose to invest into military, point-giving-civics or economy. Maybe they can be divided into piles), and the dicerolls determine the resource yield (should be a chip bag, or just a hat with ballots, you draw a chip with a number on it, boom, that number gives resources. You keep drawing until all chips are out of the bag, then you reshuffle. 1 season has passed. You only put in the robber and the numbers that have settlements/cities on them.). The thief rules also need to be modified in accordance to the ‘equal chaos’ doctrine, so when a thief chip comes up, he picks a hex automatically and is placed on a hex has the most points in terms of cities and settlements, or (in case of a draw) has more players drawing from it. In case of further draws, dicerolls or coin flips are OK. His ballot goes back to the hat, though, so there’s always a risk of getting a thief, since it’s a hand-size keep-in-checker. The investment in the military tech would still allow for strategic cock-blockery, since the robber would only move once per season (forced or by being drawn), unless moved by an army. His deck-discarding shenanigans are still in, and he only steals resources when it’s the army moving it.

    Well, that was me putting my observations and opinions in practice.. If you want to, play Settlers with the suggested rules and tell me if it’s better (please 🙂 )

  • Amvee

    You’d reasoned in your Game Design Theory podcast that games like Poker are “less efficient” at determining which player has higher skill, but here claim about Dominion that “It’s extremely hard – practically impossible – to pin this down, because of layers upon layers of many decisions that were made.”

    The extensive statistics and playtesting that Dominion undergoes undermines this argument. The numbers on — taking the page as an example — collate the effects of acquiring or avoiding certain cards over some 19 million games, and how those choices correspond with success or failure.

  • major_shiznick

    I’d totally be down to read a follow-up that inspects this very issue with a finer scope. What (modern?) video games do you think are illustrative examples of randomness placebo being used judiciously and/or excessively? Is there such a thing as “judicious” use of it?

  • Your suggested changes for Catan are pretty interesting and I think I may try them at some point. I like the idea of fixing the odds and then ensuring the adds are correct in the medium-term via using the pick-out-of-hat-without-replacement method. I’m not sure about how the replacement of the robber will work, but I’m sure a rule can be evolved through play that can work robber determination into this framework without breaking the game.

    Perhaps your suggested changes are just minimizing the effect of randomness while maintaining some of the variability it introduces and making it so randomness can only screw players in the short-term. I’m not sure how appreciable this effect will be in practice, but I’d like to see more games adopt it. I feel like your idea is just a half-measure along the road to attempting to annihilate randomness from the game all together, but perhaps that’s just the purity-hound in me coming out.

  • keithburgun

    Oh, that game. Yeah, I know that one. And the remake. It is less luck dependent, but instead more “silly un-related execution mini-game” related. But a cute game anyway.

  • keithburgun

    I don’t think it is possible to get a game of Tetris where you never get a line piece, because it’s not unfiltered random like that. Right now they have the horrible 7-bag, but they always had some kind of mitigated randomness to the system, if I recall correctly. But yeah – agreed with the thrust of your post.

  • keithburgun

    >>Output randomness can be worked into solid game design, so long as the player can expend resources to increase probability to be 100%.

    I’d actually say, output randomness can be worked into solid game design, so long as the game is *about* judging and understanding odds. If it’s a tactics game and suddenly there’s a diceroll, I’m sorry but that’s just bad game design in that you’re allowing something totally unrelated to what the game is about sometimes decide the game!

    Settlers changes sound good. I think almost any changes to Settlers would improve it 🙂

  • keithburgun

    I’m sorry, I don’t see where any of this counters anything I’ve said. Highly random games are less efficient at determining skill, and Dominion fits into that category. How do these statistics prove me wrong?

  • keithburgun

    Well, right now I think Puzzle Strike uses randomness judiciously, but that may just be because I haven’t gotten good enough at it yet to see how that’s not the case (I’ve been playing maybe 6 months).

    For videogames, I actually think the random spread on gunfire in *team based* FPSes might be a good tool to diminish the effects of player skill and thereby enhance the effects of teamwork, which is what those games are supposed to do. Then again, there’s probably a better way to do that.

    I don’t really think randomness – output randomness – has any place in games. There’s *always* a better way to determine something than by introducing sheer chaos.

  • keithburgun

    If you ever have a game you’re designing, and you think “Well, I’ll just resolve this issue with a dice roll”, there’s *always* a better way to do it. Base the outcome on something. It’s harder – more design work – but it’s usually worth it, and certainly worth it for anything central.

  • major_shiznick

    So you don’t think that the ability to weigh probabilistic risks and rewards is a worthwhile skill to test in a game?

    To you, is it reasonable to claim that the use of subtle output randomness (bullet spreads, unit damage/accuracy in large-scale strategy games, …) can enhance a gaming experience? This is part of what I mean by “judicious” use of randomness. In other words, if randomness is effectively used in a placebo-like fashion in a game in such a way that it doesn’t make the player feel consciously disenfranchised, does this necessarily worsen the game? I mean, in ways other than it being less philosophically appealing to you.

  • keithburgun

    Yes – probabilistic risk management is a worthwhile skill to test. But only if that’s what the game is actually about. Again, if you have a situation where it’s clearly a tactics/positioning/some other mechanism-based game, and then there’s just some dice roll to resolve shit, that’s not efficient game design.

    >To you, is it reasonable to claim that the use of subtle output randomness (bullet spreads, unit damage/accuracy in large-scale strategy games, …) can enhance a gaming experience?

    Maybe, but there’s always a better way to do it than by adding randomness.

  • Dasick

    >>I’d actually say, output randomness can be worked into solid game design, so long as the game is *about* judging and understanding odds. If it’s a tactics game and suddenly there’s a diceroll, I’m sorry but that’s just bad game design in that you’re allowing something totally unrelated to what the game is about sometimes decide the game!

    What I meant is that you can spend some type of resource to increase your chance up to 100%, and the meat-and-potatoes is trying to bring the number up to 100, but also judging when you *should* take the gamble

    Bullet spread in shooters is a good example of what I’m referring to. You can crouch, stay immobile and shoot in short bursts to get pin-point accuracy. But that diminishes your range, mobility/dodging, and fire rate respectively. So sure shot is something you can ‘purchase’ by making trade-offs.

    What about a tactical game like X-Com or Jagged Alliance, and your dude gets 100% chance to hit frequently, due to spending action points on aiming, gun and character stats, and terrain. I’ve noticed that in Fallout 1+2, JA, Xcom and Age of Wonders, and other games with to-hit dicerolls, you can never go above like 95% chance to hit, which I think undermined the systems really badly.

  • Dasick

    Well, the core mechanic of Settlers is trading and long-term planning. Th type of randomness I’m leaving in place provides framework for the core mechanics, and at the same time the randomness is removed from the core mechanic. That’s the goal anyways.

    The robber is almost always used to hurt the dominating player (or the runner up if you’re winning 😉 ), so the placement is just an algorithm to determine which hex being block is going to hurt the most. To be honest, I’m not sure there is a way, other than army cards, to make the robber movement be fair. Maybe there would be an auction to place the robber, and if no one bids he just goes after the fattest wallet?

  • keithburgun

    I think I disagree. How does one judge when they “should” take the gamble? What does 70% to hit *really* mean? The problem I have with this is that this chance the player takes could be a game-deciding thing, and the outcome is based on…. drumroll please… NOTHING. Random information. It doesn’t matter to me that the player understood that this was going to be the case, the game is basically ruined in this situation. In other words: just because you and I both know that if we’re flipping a coin, we both have a 50% chance to win, doesn’t make that a good gameplay mechanism.

    I know what you mean about bringing the % to 100 but if it’s anything but 100, it’s kind of necessarily bullshit.

  • Kdansky

    For anyone who wants some amusing Placebo/Homoepathy-related content:

    James Randi imbibing a full bottle of “sleeping pills” in stage, commenting: “I’m doing this for years now.”

    Also, Tim Minchin’s “Storm” has a great quote: “What do we call alternative medicine that’s proven to work? ‘Medicine’!”

    As for the games: There seems to be a huge gap between people. For example, many TF2 players despise the random crits, and there are a large number of servers which deactivate that randomness through mods. On the other hand, one of the biggest fighting games in the US right now is Ultimate Marvel vs Capcom 3. While there are very few actual dice rolls, a lot of gameplay devolves into jumping far away and spamming projectiles, hoping to randomly hit and capitalize on that with a deadly combo, even at the highest levels of tournament play. The players feel like they have agency, but in reality, they don’t, because very tiny differences make a huge difference. The winner is the person who got lucky AND was able to capitalize, which means superior reaction and the ability to keep your cool during a high-stress moment. Those are skills, sure, but why is that more interesting to the masses than actually reading the opponent in SF4? And then you have DOTA as a major phenomenon, and that’s a game where you have very little agency. You spend 45 minutes trying to “last-hit” creeps, with a pre-planned build, to culminate in one or two deciding moments. It really doesn’t make sense to me.

  • keithburgun

    Big fan of both Storm and James Randi!

    Something that I didn’t really talk about in this *or* my newer article is the fact that you can have a coherent game, and then some luck element, and while the luck element IS a breakdown at some level of what makes games good, the “gambling value” of “hoping good stuff happens” sort of takes over. It’s still a totally non-ideal situation, but it explains why so many people like random games. Plus, this arrangement is required for the placebo to work anyway… just as adults need someone in a lab coat to tell them that sugar pills cure their headache for it to work.