Riot is thinking about input/output randomness in League of Legends

For the past five years or so, I’ve been talking about the input/output randomness concepts, and why they’re so important for game design. While I wasn’t the one who coined the terms—that honor goes to the great fellows at the Ludology Podcast—some Googling around shows that no one has talked about the concepts, or developed them, nearly as much as I have.

In short, talking about randomness, especially in these terms, is kind of my thing.

Here’s a quick list of some of the more notable times that I spoke in depth about this subject:

Beyond that, you can (and should) read my books, listen to my podcast and watch the complete 3 Minute Game Design on YouTube.

Between that, and the fact that I recently praised League of Legends and its developer, Riot, for some of the radical things they’ve been doing to their game, you can imagine how pleased I was to see that a Riot designer, Greg Street, made a video talking about input and output randomness in League.

I want to do three things with this article. First, I want to signal to the Riot devs that I would be more than happy to help them out with this project of making the game more input-randomnessy and less output-randomnessy. Let’s chat!

Second, I will talk about the current state of randomness in the game. And finally, I will talk about a few of my recommendations going forward.

The Gist

You should check the links above for more detail, but the very short and rough explanation of input/output randomness is that they are describing the distance between new random information and the user’s ability to react.

Input randomness is stuff like the Tetris “Next” box, or a randomized map—stuff where the user has time to see what the random information is and make decisions based on it. Fog of war is also often used as a mechanism for input randomness: you see stuff usually a turn or two before it can threaten you.

Output randomness is stuff like “roll to hit” in RPGs or games like X-Com. You decide to attack a thing, and then there is random information which comes into the game and determines whether that hit connects or not. In this case, you have zero time to “respond” to the new random information. The course of the game has now been altered by pure randomness, unfiltered by your decision-making.

A quick note is that actually, input and output randomness exist on a spectrum. Randomness which is really close to the user but still technically is input randomness(imagine 1-tile-radius fog of war) may actually be so close as to function similar to output randomness. For more on this, read my article on the Information Horizon.

One more note: as Mr. Street said in the video, output randomness is bad for strategy games. (Actually, he wasn’t that strong with the language, but that is definitely my view.) You also don’t want zero randomness in games. What you want is a carefully placed information horizon.

Randomness in League

There are a number of sources of randomness in League already, but what’s exciting is what they’ve been adding, so let’s talk about that first.

One of the big ones, that they’re talking about in the video, is the new Dragons (or I guess they’re calling them Drakes? I’m going to call them dragons) system. Every game, there is a random elemental dragon. Killing each dragon gives your team a different kind of buff—something like, Air dragons make you move faster, fire dragons make you deal more damage, Earth make you destroy towers faster, and so on. After one of these dragons is killed, a new random one will spawn.

But here’s the cool thing, and when I saw it, I was impressed, because it’s exactly what you should do. The dragon takes awhile to respawn. But on the dragon’s death, the next dragon is selected and a big bright symbol is painted on the dragon’s lair wall for everyone to see. This is a great example of input randomness. Both teams know exactly what random dragon is going to spawn there, and decisions can be made around that.

This system is a really great start. There should be a lot more of this on the League of Legends map (which I guess they call “Summoner’s Rift”? I’m going to call it “the League of Legends map”).

My Suggestions

Category #1: Increasing input randomness. I believe that by making the League of Legends map more dynamic and more different each game, Riot can worry less about perpetually adding content to the system and the sort of “patching just for the sake of keeping things fresh” idea that they arguably do sometimes.

  1. Randomized geometry. Why is the map geometry—the pattern of the walls and everything—the same exact every match? Is this geometry sacred? Remove this brush, put a pillar here, change the shape of this wall. You can definitely design the parameters in such a way that it’s always fair for every character, yet slightly different every game. If you’re really worried about it causing balance problems, maybe have it change randomly at 10 minutes and at 20 minutes, or something.
  2. Randomize all jungle monsters. Having a fixed jungle means having a fixed jungle route. Instead why not have there be a few more jungle monsters, but what they are and where they are is randomized somewhat. That way you can have a more dynamic and less “memorized Starcraft build-order” automatic pathing to the jungling. Make it mirrored, so it’s fair.
  3. Random plants. I love the new “plants” system, and maybe it’s just the beginning. But why not randomize the positions and the types for these plants? Make it mirrored, so it’s fair.
  4. Push the dragons further. Right now the buffs are cool, but maybe instead of just the four elements, it’s four different kinds of monsters as well. So you could have an Earth Wizard who does lots of magic damage, or a Fire Ogre who has shitloads of health. Maybe Ogres, when killed, change the terrain somewhere else or spawn plants somewhere.
  5. Random (mirrored) cannon minions, and more? I like the idea of there being some more variance in the minion-stream itself. What if randomly, every 5-7 waves or so, a super minion spawned on one of each team’s lanes (probably, it can’t be the same lane). There could be a little map alert telling both teams about this. I spent like two minutes thinking of this idea, I don’t know—but the point is, I do think there’s a lot more that could be done with minions than is being done here.

Category #2: Removals of existing output randomness. These aren’t quite as important in my view, and they’re also less likely to be taken up by Riot for a number of reasons. But ideally, I’d love these to be changed.

  1. Remove Random Critical Hits. This is totally unnecessary for this game. It’s totally a vestigial D&D thing and all it does is unnecessarily loosen up a system which already is very loose. I’m actually surprised that high level players don’t complain about this, because a lucky crit at the wrong time can completely change the course of a match.
  2. Reduce execution across the board. This one is hard, and I don’t expect Riot to act on it for that reason. But, in an ideal world, we should be making the game be less about crazy reaction speed in team fights, and more about strategy. Removing skillshots and having them be target-based is one suggestion here. Getting rid of “burst” (massive amounts of damage in a tiny window of time) as a concept would be a much wider scope change that would really benefit the game. That would mean questioning the roles of burst-mages and assassins, which is a huge job, but personally I think it would be worth it. Because, as I’ve written about before, execution is a form of randomness, and this becomes more and more the case the faster players are asked to execute.

(On a slightly related topic, here are some more League patch notes that I’d love to see.)

Anyway, those ideas are mostly off-the-cuff (but not getting rid of random crits. That seriously needs to go), but the point is just to demonstrate some ways that Riot’s designers—and designers in general—can use the input/output randomness theory in practice to create a better experience. Using input randomness in this way, you can get the “variety” that is so sought after using extremely costly (both in terms of production costs and accessibility costs) asymmetry content, while also providing players with a fair, balanced competitive experience.

I’m really glad to see someone as high profile using these terms. If you know of anyone else talking about randomness in this way, please let me know in the comments!


If you enjoyed this article, please consider becoming my patron on

  • Samuel

    I think your whole ‘execution is randomness’ schtick is still very misguided and that the recommendations to remove entire playstyles and roles in the game is short-sighted and narrow-minded.

    Let’s start with execution as randomness. I remember you saying you don’t believe in ‘mind games’ so I guess you’ll find the rest of my argument invalid but bear with me.

    Sure, adding humans and execution to any strategy game adds some sense of the unknown, you’ll never be able to clearly say “he’s definitely going to do that!”. But you can say “he’s very likely to do that” and surely that’s enough. Let’s take the example of skill shots. (I’m rusty with League, since I’m a Dota guy, so apologies if it doesn’t make complete sense)

    – I’m in a team fight, supporting my bruiser friend.
    – We are against another bruiser and Morgana.
    – Both pairings want to take out the other one’s support, because logic.
    – I know that Morgana has a skill shot so I position myself behind creeps and move around unpredictably to avoid the potential skill shot.
    – Morgana has to make a tactical decision about whether to use the skillshot on the easier-to-hit target (the bruiser) or risk missing it in order to hit the more important target (me).

    What decision she makes is somewhat of a coin flip, but it doesn’t matter at that point. The value of the mechanics in terms of their strategic worth are already there. We’ve already been making tactical decisions that affect our behaviour because of the scenario’s context and our inherent understanding of “this is effective in League, that is not”.

    Even in this situation, it’s a 50/50 chance that she’ll do either of those options… but that’s 2 very likely possibilities that have been determined from the millions of situations the game offers as a whole. I feel you say “This is random” to mean “this outcome is completely and utterly outside our possible perception and can therefore not account for it”. This situation is CLEARLY not outside possible perception and can be strategized and played around.

    Random humans are black boxes.

    Random League players at a certain ELO playing a certain hero in a certain role in a certain environment with certain skills on cooldown and certain allies nearby are NOT black boxes. This is all information that has a pretty big horizon (from 3secs to 2mins depending on the specific info) and therefore can be evaluated and decided upon.

    For removal of burst damage, I think you’ve got a similar problem. Burst damage does not come from nowhere. It comes from a player being within range of their target with certain abilities available. And this is in a game where you have a mini-map and wards. In Dota there is a character called Lion who can basically one-shot almost anyone in the early game with his ultimate. You know that if Lion is missing from the mini map, he’s going to be looking for a kill. You can already react to this on a strategic level with ward placement, good positioning, item purchases and communication. If I am playing Dazzle, a hero who can make anyone immune to death for 5 seconds, I will be on the lookout constantly for Lion and if I suspect he will be ganking the mid lane I will have a Teleportation scroll in order to get by his side and save his life.

    Burst damage is an important tactical consideration and plays a crucial role in making shit happen. If carries didn’t have to fear assassins, then who the hell is going to stop them getting fat and winning the game?

    I like your big bold statements sometimes, Keith, but something like this I’d expect to have a lot more consideration from a design perspective.

    “Getting rid of “burst” (massive amounts of damage in a tiny window of time) as a concept would be a much wider scope change that would really benefit the game.”

    How? Will it benefit players who like that playstyle and want to be in the killer role? Would it even benefit the typical targets of that burst damage?

    I feel you often look at problems solely from the perspective of “this shit beat me, therefore I dislike it”, without taking into account the greater game experience and playstyles that do not match your own. It’s a MOBA. People spectate it. People want dramatic shifts in the battlefield, they want glorious battle and tests of skill. As a strategic player (I only really play supports) I WANT to have to consider burst damage assassins when I play these games… it’s like my main role as a support in particular phases of the game.

  • >Will it benefit players who like that playstyle and want to be in the killer role?

    This kind of “it was there originally, so now it has to say” thinking is a real problem I think. It’s arbitrary that such a thing exists at all. Suggesting we remove *anything* from the game can be met with the same “but what about players who LIKE that thing” argument. The difference between a game designer and a player is that the game designer should be able to look past a player’s personal preferences and say something like, “yeah, I know you like this playstyle, but, the overall game experience will be healthier without it”. Treating something like burst damage (or snipers, which I’m equally skeptical of for FPSes) as this kind of gospel is a mistake.

    It’s quite possible that there could have been some other mechanism in League, like, I don’t know, flying heroes. And I might be here saying “flying heroes really suck for the game”. And then someone like yourself would say “but what about the people who LIKE flying heroes?!” Hopefully this demonstrates why this argument is futile.

    With that said, if you don’t agree with my opinion on execution being a source of randomness, then I can see why you wouldn’t agree about something like burst damage. The reason burst damage is bad is because it happens in such a small window, meaning that there is more reliance on execution, and therefore more randomness.

    I assure you that none of my feedback comes from “that shit beat me”.

  • Samuel

    I get what you mean about time windows, I mentioned a good example from Dota.

    Firstly, the time window is NEVER that small in a competitive strategy game (i.e. a MOBA). Everything is consequential, it just requires an understanding of the game. A contest of understanding, if you will. The closest thing you have to a ‘too random’ encounter is if there’s a hero that can instantly teleport next to you and burst you down in 3 seconds. But that doesn’t exist, because that would be poor design for the reasons you state burst damage is bad. Which leads into my second point…

    The line where I draw ‘it’s just execution and guessing’ is probably a lot closer to real-time than you’re happy with. I suppose we appreciate the importance of quick-thinking differently in terms of strategy game design. Interestingly enough, since you discuss subjectivity in your latest podcast, there IS a case for “but this is just taste and everything is ok” for this situation. You’re trying to say “this is the lowest a time window can go that doesn’t breach the randomness boundary”. But that boundary is clearly a subjective opinion and I doubt there’s any non-personal justification for why that time window is exactly where you say it should be… coz people aren’t all the same. My example time window of 3 seconds from teleport to death is probably too long for some people who can think quicker than me. I don’t like it but it’s not bad design per se. It’s not wrong.

    You got a point about “these people like it, what about them?” but I feel you haven’t addressed the majority of my point. What IS good for the game, Keith? I think Burst Damage is a great design tool for heroes because it adds something to fear, something to defend against, something to be aware of, and (most importantly) something to use as power projection to prevent farmers from just playing Farmville with swords. I did not JUST say “what about call of duty kidz?!”, although in reality the idea of completely alienating a large section of the user base is also a valid concern. What would fill the void of burst damage’s removal? Your suggestion is too vague, so how am I to take it seriously? It’s just as bad as “what about the people who LIKE flying heroes?!”

  • If you check my most recent article from today, you’ll see that one of my criteria is that players shouldn’t be making “input errors”. They shouldn’t be making inputs that they didn’t mean to make. So having the kind of fast real time combat that exists in League is problematic for me in general, and burst damage is an extreme form of that. (The fact that you paint the “teleport, burst down move” picture shows that you DO understand where I’m coming from, but we just disagree on where it becomes a problem.)

    I don’t understand your last paragraph. Burst damage is great because it “adds something to fear”? I mean, no matter what we added (call it “X”) to the game, it would “add something to [insert word that describes response to X]” to the game.

  • Samuel

    Yes everything we add creates something in the game.

    I’m looking for your design thoughts on how you’d replace the interesting dynamics that burst damage characters create on both a micro and macro level.

  • I don’t see any need to replace the concept of a burst damage character. I think the game will be overall better without it. You can still have characters that are “high threat”, like if they catch you alone, they’re probably going to kill you, or they’re very strong brawlers, or whatever. But I think it would be a purely positive step in the direction of a better game to just remove the burst down characters. And if the only counter-point to that is “but what about people who like that?” then I would say that’s unconvincing, and I would go further to say that that kind of thinking would really, seriously hamper a game designer’s ability to do his job.

  • Samuel

    It’s strange that after explaining various strategic effects and considerations of burst damage characters and what they bring to the game, you still counter with “your only point is that people like it”. I’m yet to see an interesting counter-point to my actual point. Just saying “it would be a purely positive step in the direction of a better game just to remove the burst down characters” isn’t actually saying what the positive step is. It sounds a lot to me like “I like it” which you keep mentioning is my only point.

    The one thing that has been posed by you with substance is that people are increasingly likely to make input errors when there’s a shorter window of reaction time. Fair enough, some people are slower than others. I would counter with the fact that there’s no such thing as burst damage in a game with a massive map, real-time communication, surveillance tools, and a very clearly established metagame. Sure your HP goes from 3000 to 0 in 2 seconds, but you were dead the second that player left the lane and exploited your poor positioning. That isn’t a 2 second duration, you have at least half a minute to display your understanding of the game systems via map awareness, communication, and good positioning. That’s strategy.

    In your podcast you mention that some games need a timer to prevent someone overthinking every turn to make the game playable. Is this not just a different side of the same coin? Sure, the game is playable if everyone in League has 10 seconds to escape from any combat situation… but is it good? Is it an interesting test of skill? Is someone’s level of understanding truly being tested when you’ve got that much time to make a decision? At what point do you decide “this player has defeated you”?

    You say “10 mins is too long to take a turn in a Euro game”, I say “10 secs is too long to react to a gank in a MOBA”.

    The funny thing is, bursting a guy to death isn’t even winning… you respawn and have 40 more minutes to prove your understanding of the game better than your opponent. A MOBA is not a series of 1v1 encounters, the genre didn’t capture the hearts and minds of the world’s population by being just a tactical arena battler. Once again, I feel the way you’re focusing on “burst characters don’t give me time to react smartly” is complete tunnel vision. You’re dismissing the game as a whole when a MOBA must be considered as a complete package.

    And yes it is unconvincing for a professional game designer to not explain his point more than “I like it”.

  • >Sure your HP goes from 3000 to 0 in 2 seconds, but you were dead the
    second that player left the lane and exploited your poor positioning.
    That isn’t a 2 second duration, you have at least half a minute to
    display your understanding of the game systems via map awareness,
    communication, and good positioning. That’s strategy.

    Actually, if there was a 100% chance that this character will kill you, I’d be more okay with it. Like if that player just had to click on you once within range, and you died, that would be MUCH better than the current situation. Because the current situation is that based on some fast combos/reactions on the part of the two players, the attacking player can make an input error and screw up, or the other player could do some weird execution move and escape the burst. This is what makes it really random, and what makes it a problem.

    > but is it good? Is it an interesting test of skill? Is someone’s level
    of understanding truly being tested when you’ve got that much time to
    make a decision?

    I feel like there’s an implicit “no, it’s not good” that you’re suggesting, but I think that that would absolutely be a good thing. And yes, I think that once you are getting the inputs that everyone intends to make, that’s exactly when you are testing someone’s level of understanding (and not their muscle memory or reaction time).