WEBVTT 00:00:04.190 --> 00:00:08.430 If you’ve spent any amount of time in a multiplayer lobby, you’ve probably heard 00:00:08.430 --> 00:00:12.030 words like overpowered, cheap, and unfair. 00:00:12.030 --> 00:00:16.610 What these players are arguing about is the game’s balance. 00:00:16.610 --> 00:00:21.500 Balance is the art of making sure that all options in a multiplayer game are fair: so 00:00:21.500 --> 00:00:26.710 none are underpowered, and thus pointless to use. And none are overpowered, and thus 00:00:26.710 --> 00:00:28.550 dominate everything else. 00:00:28.550 --> 00:00:33.660 Here’s the thing though: most video games aren’t just striving for balance. But balance 00:00:33.660 --> 00:00:36.809 among a wide range of distinctly different options. 00:00:36.809 --> 00:00:41.870 You don’t have to work that hard to balance a symmetric game - which is one where all 00:00:41.870 --> 00:00:47.350 players have the exact same starting conditions. But most games are asymmetric - which means 00:00:47.350 --> 00:00:51.390 players are facing off against each other with completely different stuff. 00:00:51.390 --> 00:00:56.159 And in a game where players can pick from 74 different fighters or 140 00:00:56.159 --> 00:01:01.550 unique champions, the developers are counting on them all being equally viable among players 00:01:01.550 --> 00:01:03.659 of roughly the same skill level. 00:01:03.659 --> 00:01:04.790 So how do they do it? 00:01:04.790 --> 00:01:10.860 Now, I should say, balance is an incredibly difficult pursuit. It can be an entire department 00:01:10.860 --> 00:01:16.540 at certain companies, and Riot’s League of Legends has had more than 200 balance patches 00:01:16.540 --> 00:01:22.330 in the last decade. Plus it’s not just about numbers, but player psychology, with Overwatch’s 00:01:22.330 --> 00:01:26.800 Jeff Kaplan saying “the perception of balance is more powerful than balance itself”. 00:01:26.800 --> 00:01:33.240 So this is not going to be an intensive tutorial. Instead, Game Maker’s Toolkit presents 00:01:33.240 --> 00:01:40.060 a whirlwind tour through the ways games are balanced - and rebalanced - and rebalanced 00:01:40.060 --> 00:01:42.310 - and rebalanced. 00:01:42.310 --> 00:01:47.270 So how do developers go about balancing a game in the first place? Well, the first consideration 00:01:47.270 --> 00:01:53.300 is trade-offs. This is when you essentially cancel out a character’s competitive advantages, 00:01:53.300 --> 00:01:54.680 with drawbacks. 00:01:54.680 --> 00:02:00.310 Think of Mario Kart characters, where heavy racers like Donkey Kong have a high top speed, 00:02:00.310 --> 00:02:05.140 but low acceleration, while featherweight racers like Toad are the opposite. On the 00:02:05.140 --> 00:02:08.630 right track design, they’ll be almost evenly balanced. 00:02:08.630 --> 00:02:13.550 You can think of characters as having a “power budget” - at least that’s why Riot calls 00:02:13.550 --> 00:02:19.870 it. Advantages are a cost, but disadvantages are a discount. If all characters are just 00:02:19.870 --> 00:02:24.500 about hitting the limit of the same power budget, they’ll be closer to being balanced. 00:02:24.500 --> 00:02:29.530 It’s rarely that easy, of course. I mean, okay, sometimes you’ll get a card that does 00:02:29.530 --> 00:02:36.010 1 damage to all minions and another that does 4 damage to all minions. That’s an easy one: just make 00:02:36.010 --> 00:02:39.010 the second card cost a bit more energy. 00:02:39.010 --> 00:02:43.680 But how do you calculate the power budget for completely incomparable options like heroes 00:02:43.680 --> 00:02:48.950 in Overwatch? Or options with dozens of stats to tweak? Like, when Bungie was reigning in 00:02:48.950 --> 00:02:54.150 the initially overpowered sniper rifle in Halo 3, it had loads of stats it could tweak 00:02:54.150 --> 00:02:59.050 such as clip size, time to full zoom, reload time, and max ammo. 00:02:59.050 --> 00:03:03.480 (It ultimately decided the best knob to tweak was the time between shots, which it bumped 00:03:03.480 --> 00:03:06.370 from 0.5 to 0.7 seconds). 00:03:06.370 --> 00:03:11.500 What’s important, though, is to celebrate the big differences between choices. The sniper 00:03:11.500 --> 00:03:16.380 rifle and the shotgun offer a more exciting choice to players than two types of assault 00:03:16.380 --> 00:03:19.900 rifle - even though the latter is much easier to balance. 00:03:19.900 --> 00:03:25.420 So I agree with ex-Blizzard designer Rob Pardo when he warns designers against using the 00:03:25.420 --> 00:03:28.716 maths to balance games into mediocrity, saying 00:03:28.716 --> 00:03:31.510 ROB PARDO: "you’re gonna end up with a game where 00:03:31.510 --> 00:03:35.489 everything kinda feels the same. And you can high five each other and say it’s balanced, but 00:03:35.489 --> 00:03:38.240 is it fun? Probably not”. 00:03:38.240 --> 00:03:43.550 Another consideration is counters. This is when we give characters the ability to negate 00:03:43.550 --> 00:03:49.069 each other’s moves and strategies. For example: a quick Zerg rush in Starcraft is all well 00:03:49.069 --> 00:03:54.520 and good - unless your opponent is one step ahead and has already built defensive bunkers. 00:03:54.520 --> 00:03:59.900 And what we ideally want is for everything to have a counter. So a defensive Starcraft 00:03:59.900 --> 00:04:06.160 player can, in turn, be countered by a more economical strategy, where you save up resources 00:04:06.160 --> 00:04:10.480 to build units that can eventually crush those bunkers into dust. 00:04:10.480 --> 00:04:15.450 And we could make a counter to that counter, and so on - but then we’d be here for forever. 00:04:15.450 --> 00:04:19.780 There’s a more elegant solution though, because how do you deal with someone who’s 00:04:19.780 --> 00:04:23.360 sitting around saving up money? Well… a rush. 00:04:23.360 --> 00:04:29.520 And, wouldn’t you know it… it’s rock, paper, scissors. This goofy game you play 00:04:29.520 --> 00:04:34.440 to see who has to do the washing up might be incredibly simple and lacking any strategic 00:04:34.440 --> 00:04:40.680 depth, but it is perfectly balanced - because everything has a counter, and everything is a counter. 00:04:40.680 --> 00:04:43.880 EDDIE: Damn man, killed those scissors. 00:04:43.880 --> 00:04:48.940 And that’s why it forms the backbone of a lot of multiplayer games. Pretty much every 00:04:48.940 --> 00:04:54.160 fighting game has a system like this, such as Dead or Alive which boasts about its triangle 00:04:54.160 --> 00:04:59.160 system, where strikes beat throws, throws beat holds, and holds beat strikes. 00:04:59.160 --> 00:05:04.490 In strategy games, it’s not just the strategies that work like this, but the individual units. 00:05:04.490 --> 00:05:08.480 And the different Pokemon types all sit in a massive web of interlocking counters - but 00:05:08.480 --> 00:05:12.250 starting, of course, with fire, water, and grass. 00:05:12.250 --> 00:05:17.690 Rock, paper, scissors is a great balancing framework to start from, because you can ensure 00:05:17.690 --> 00:05:22.650 that no element is overpowered - it’s countered by something. And no choice is irrelevant 00:05:22.650 --> 00:05:26.310 - it at least works as a counter to something else. 00:05:26.310 --> 00:05:32.460 And also, in strategy games at least, it encourages mixed strategies, it makes you into a multi-disciplinary 00:05:32.460 --> 00:05:37.470 player, and it forces you to switch tactics on the fly in a really dynamic way. 00:05:37.470 --> 00:05:42.520 And in class-based games, it’s a great way of automatically making mixed teams. 00:05:42.520 --> 00:05:48.660 Take Team Fortress 2, where seven of its nine classes fit into a complex web of interlocking 00:05:48.660 --> 00:05:54.190 and interchangeable triangles of rock, paper, scissors. Here, teams must pick complementary 00:05:54.190 --> 00:05:59.870 classes to protect each other from weaknesses. If you’re an Engineer and Spies keep sapping 00:05:59.870 --> 00:06:03.490 your sentries, then you’re going to need to get one of your team mates to switch to 00:06:03.490 --> 00:06:05.000 Pyro. 00:06:05.000 --> 00:06:09.410 These counters are often described as hard counters if they completely shut something 00:06:09.410 --> 00:06:14.680 down - like a punch is a hard counter to a throw in ARMS because it will nullify the 00:06:14.680 --> 00:06:20.690 effect every single time. But soft counters just mean one choice will have an advantage 00:06:20.690 --> 00:06:28.190 over the other. McCree will outperform Tracer, but his chance of winning is far from 100%. 00:06:28.190 --> 00:06:33.100 When it comes to counters, it’s really important to figure out what are the hands and what 00:06:33.100 --> 00:06:34.310 are the throws. 00:06:34.310 --> 00:06:39.270 The hands are the things that get locked in before the match even starts. You know, the 00:06:39.270 --> 00:06:45.740 characters and the races. The throws are the things you pick during the match. The moves, 00:06:45.740 --> 00:06:51.320 the units, and the strategies. And in a team-based game, like Overwatch, the entire team is the 00:06:51.320 --> 00:06:55.190 hand, while the individual players are the throws. 00:06:55.190 --> 00:07:00.100 The throws are specifically designed to be unbalanced against each other, to create that 00:07:00.100 --> 00:07:05.669 back-and-forth counter-play and teamwork. But the hands are supposed to be balanced, 00:07:05.669 --> 00:07:11.240 and so they should have access to all of the throws. If Zangief simply couldn’t block, 00:07:11.240 --> 00:07:14.410 for example, he would be unusable. 00:07:17.320 --> 00:07:22.040 So you’ve got a bunch of characters, with trade-offs and counters, and you think you’ve 00:07:22.050 --> 00:07:26.520 made them balanced. But how do you actually make sure that’s true? Well this is when 00:07:26.520 --> 00:07:31.560 we start collecting data - either from internal play-testers, or the millions of people playing 00:07:31.560 --> 00:07:32.720 your game online. 00:07:32.720 --> 00:07:36.940 Now you might think that all you need to do is track how often each character results 00:07:36.940 --> 00:07:42.440 in success - i.e it’s “win rate”. And if a character has a 50 percent win rate, 00:07:42.440 --> 00:07:43.500 it’s balanced. 00:07:43.500 --> 00:07:49.120 But, like all stats, this can be misleading. Imagine a fighting game with three characters 00:07:49.120 --> 00:07:54.660 - and if Ryu won every match against Chun-Li and lost every match against Cammy, his win-rate 00:07:54.660 --> 00:07:58.979 would be 50 percent. Perfectly balanced, though? I think not. 00:07:58.979 --> 00:08:02.979 That’s why match-up charts, where you where you can see the win rate of a character, when 00:08:02.979 --> 00:08:06.500 played against all other characters, are so important. 00:08:06.500 --> 00:08:10.729 But even that’s not going to tell you everything. Riot had a problem with the League of Legends 00:08:10.729 --> 00:08:15.939 character Akali. The numbers said she was pretty balanced, with a 44 percent win-rate 00:08:15.939 --> 00:08:21.789 - perhaps a tad underpowered. So how come she secured a 72 percent win rate at the 2018 00:08:21.789 --> 00:08:25.870 World Championship, and was banned more times than any other champion? 00:08:25.870 --> 00:08:31.361 It’s because while she was really powerful, she was difficult to play effectively. She 00:08:31.361 --> 00:08:36.080 had a super high skill floor, in other words. So while top-tier players could use her to 00:08:36.080 --> 00:08:41.340 wipe the floor with the competition, the low-ranking players using Akali were getting killed left, 00:08:41.340 --> 00:08:45.270 right, and center. Therefore, her win-rate was being dragged down. 00:08:45.270 --> 00:08:49.930 That’s why it’s important to look at a character’s win-rate and match-ups across 00:08:49.930 --> 00:08:51.370 all skill levels. 00:08:51.370 --> 00:08:56.360 And finally, win-rate doesn’t really tell you what’s actually going on in the game. 00:08:56.360 --> 00:09:00.540 We need to know what characters people are actually picking. People might be avoiding 00:09:00.540 --> 00:09:05.300 a character who is otherwise well balanced because that character is not much fun to 00:09:05.300 --> 00:09:08.670 play, or is only useful in certain situations. 00:09:08.670 --> 00:09:13.871 Blizzard found that Overwatch hero Symmetra was a largely balanced character, but she 00:09:13.871 --> 00:09:19.020 wasn’t being picked as much because her use was highly situational. So in her first 00:09:19.020 --> 00:09:23.370 complete redesign, they tried to make her more popular by giving her two ultimates to 00:09:23.370 --> 00:09:26.230 pick from: a teleporter or a shield generator. 00:09:26.230 --> 00:09:31.250 That’s why player feedback is so important - as well as pick-rate, which tells you how 00:09:31.250 --> 00:09:37.300 often a character is actually getting used. For Rainbow Six Siege, Ubisoft uses a matrix 00:09:37.300 --> 00:09:41.720 to cross reference both win rate and pick rate -with different considerations needed 00:09:41.720 --> 00:09:45.450 for operators who fall into these four buckets. 00:09:45.450 --> 00:09:50.480 And the pick-rates help tell you the state of the meta - which is essentially just the 00:09:50.480 --> 00:09:54.910 characters, cards, strategies, and so on that the community at large have found the most 00:09:54.910 --> 00:09:57.470 effective and are currently using. 00:09:57.470 --> 00:10:03.910 This is often shared through forum posts, fan-made tier lists, YouTube videos, and eSport 00:10:03.910 --> 00:10:08.740 victories. When a kid called Jason won the Clash Royale tournament in Helsinki, his chosen 00:10:08.740 --> 00:10:12.160 cards suddenly became massively popular. 00:10:12.160 --> 00:10:16.600 The meta can actually act as a self-balancing force. Let’s say everyone discovered that 00:10:16.600 --> 00:10:21.810 a certain character was overpowered, and everyone started using it. It’s now in everyone’s 00:10:21.810 --> 00:10:26.920 best interest to try and discover strategies that can counter or out perform that favourite. 00:10:26.920 --> 00:10:29.960 And if players find it, the meta might change. 00:10:29.960 --> 00:10:34.480 This rolling meta keeps the game fresh, and gives the players who found the counter a 00:10:34.480 --> 00:10:39.910 real sense of satisfaction. Overwatch’s Jeff Kaplan says “regarding the meta changing 00:10:39.910 --> 00:10:44.960 because players have innovated a new strategy – well – this is the best-case scenario. 00:10:44.960 --> 00:10:48.220 We’ve seen this happen time and time again.” 00:10:48.220 --> 00:10:52.910 Of course, that’s not always going to work. Sometimes the designers will have to go in 00:10:52.910 --> 00:10:58.380 and change things. If a strategy is overpowered, if a character is never getting played, or 00:10:58.380 --> 00:11:02.519 if a play-style is proving annoying then it’s time to swing the hammer. 00:11:02.519 --> 00:11:07.370 First, the devs need to figure out the exact reason why that character, or strategy, or 00:11:07.370 --> 00:11:11.820 whatever is unbalanced. It’s easy to see that a character is dominating the match-up 00:11:11.820 --> 00:11:14.850 charts, but can be harder to pin-point why. 00:11:14.850 --> 00:11:19.250 So for a character like Meta Knight in Super Smash Bros Brawl, it was mostly because of 00:11:19.250 --> 00:11:24.380 his extremely fast attack speed, and an ability to cancel his momentum in mid-air and avoid 00:11:24.380 --> 00:11:28.910 being KO’d. He had lots of advantages, and not enough trade-offs - and other characters 00:11:28.910 --> 00:11:31.580 don’t have the tools to counter him. 00:11:31.580 --> 00:11:36.269 Once the source has been found, you’ve got to figure out what to nerf and what to buff. 00:11:36.269 --> 00:11:40.680 Nerfing means making something less powerful, like reducing their speed, limiting their 00:11:40.680 --> 00:11:45.470 range, or cutting down their strength. Buffing is the opposite: making it more powerful. 00:11:45.470 --> 00:11:50.730 You don’t necessarily have to buff the weak characters and nerf the strong ones, though. You could 00:11:50.730 --> 00:11:55.000 leave an overpowered character alone, but buff the characters who counter them, and 00:11:55.000 --> 00:11:59.970 still solve the same problem. Make sure you watch this Core-A Gaming video on why buffs 00:11:59.970 --> 00:12:02.579 are, generally, better than nerfs. 00:12:02.579 --> 00:12:06.649 Balance changes can be anything from a tiny tweak to a character’s movement speed, to 00:12:06.649 --> 00:12:11.149 a complete overhaul of how a character works. It might be a fundamental change to the rules 00:12:11.149 --> 00:12:15.730 of the game - Rainbow Six Siege made attacking and defending more balanced by changing the 00:12:15.730 --> 00:12:19.839 match time to three minutes. And sometimes you’ve just to pull things from the game 00:12:19.839 --> 00:12:25.730 entirely, like when Epic scrapped the overpowered infinity blade in Fortnite. 00:12:25.730 --> 00:12:31.210 Any change is going to affect players - especially those who are very used to the way a specific 00:12:31.210 --> 00:12:35.279 character, or its counters, work. So when the game gets patched, it’s important to 00:12:35.279 --> 00:12:38.820 communicate the changes through patch notes, videos, and so on. 00:12:38.820 --> 00:12:43.740 In fact, patch notes are so important that Riot once put out of a note saying a champion 00:12:43.740 --> 00:12:49.060 was nerfed, but forgot to actually implement the nerf in the code. Even so, the character’s 00:12:49.060 --> 00:12:54.140 pick rate plummeted, and even his win rate decreased a bit. Didn’t I say that player 00:12:54.140 --> 00:12:57.260 psychology was an important factor? 00:12:57.260 --> 00:13:01.980 Now, at the beginning of this video, i said that balance was about trying to make characters 00:13:01.980 --> 00:13:07.120 equally viable among players of roughly the same skill level. But what happens when players 00:13:07.120 --> 00:13:09.160 aren’t at the same skill level? 00:13:09.160 --> 00:13:12.480 Well, a lot of highly competitive games use matchmaking 00:13:12.480 --> 00:13:15.340 systems to pair up similarly skilled players. 00:13:15.340 --> 00:13:20.080 But for more accessible, party-style games, we may want to build in negative feedback 00:13:20.090 --> 00:13:25.529 loops, or catch-up mechanics, where players who are doing poorly get a helping hand. Examples 00:13:25.529 --> 00:13:29.910 are the deathstreak mechanic in Modern Warfare 2 where you get a special bonus for dying 00:13:29.910 --> 00:13:35.000 a whole bunch. And the item system in Mario Kart where powerful items - including that 00:13:35.000 --> 00:13:39.950 pesky blue shell - are only given to players at the back of the pack. These are pretty 00:13:39.950 --> 00:13:42.980 contentious, and must be used sparingly. 00:13:42.980 --> 00:13:48.710 We can also reduce the value of skill by adding in more luck. We see this in most family board 00:13:48.710 --> 00:13:53.110 games like Snakes and Ladders and Monopoly which are heavily based on the luck of the 00:13:53.110 --> 00:13:58.320 die roll. But in video games, you see this in games like Apex Legends, where your chances 00:13:58.320 --> 00:14:03.209 of winning are shifted, based on what goodies you find when you drop into the map. 00:14:03.209 --> 00:14:09.230 Game can also offer handicapping modes. And in team-based games, we can give players alternate 00:14:09.230 --> 00:14:14.110 play styles that allow them to contribute to the team without needing to do highly-skilled, 00:14:14.110 --> 00:14:18.990 front-line action, like being a medic or an engineer. 00:14:18.990 --> 00:14:24.230 So balancing a game is a really challenging job. The more you make characters distinct, 00:14:24.230 --> 00:14:28.350 the harder it is to put them on an even playing field. And that’s not taking into account 00:14:28.350 --> 00:14:30.670 players of unequal skill level. 00:14:30.670 --> 00:14:35.170 We can try to design in trade-offs, to ensure characters don’t have too many advantages. 00:14:35.170 --> 00:14:39.089 And give characters counters, so they can keep each other in check. But even the best 00:14:39.089 --> 00:14:43.160 designs won’t stand up to scrutiny when put in front of millions of players. 00:14:43.160 --> 00:14:48.329 So we need to constantly determine the balance, by watching win-rates, match-ups, pick-rates, 00:14:48.329 --> 00:14:52.889 and player feedback. And while hopefully the meta will naturally shift in response to imbalance 00:14:52.889 --> 00:14:57.050 - sometimes devs have to go in and make the hard changes. 00:14:57.050 --> 00:15:01.550 And then you introduce a whole new character and everything breaks again. Sigh. I said 00:15:01.550 --> 00:15:06.230 this wasn’t an easy job. So let me know: what do you think is the most balanced game 00:15:06.230 --> 00:15:11.050 around, and have you ever played a game where the devs just got it oh so wrong? Let me know 00:15:11.050 --> 00:15:15.420 your experience with balance in the comments below. 00:15:15.420 --> 00:15:19.560 Thanks for watching! I had a lot of help on this one, from people who know multiplayer 00:15:19.560 --> 00:15:24.720 games really well to developers who have worked on games like League of Legends, Dirty Bomb, 00:15:24.720 --> 00:15:28.589 and Rainbow Six Siege. There’s definitely more to talk about - like balancing multiplayer 00:15:28.589 --> 00:15:34.100 maps in shooters. But we can get to that in the future.