Game Design in DotA 2: Icefrog’s Battle with Positive and Negative Feedback Loops
Good god I’m lazy
This blogpost is an extrapolation of the idea’s behind Game Maker’s Toolkit(GMT) on Feedback Loops. You should watch it because GMT/Mark Brown is wonderful and I apply some of the ideas within DotA 2, albeit loosely. I get kind of lazy on the particulars to be honest, but probably because no one will read it.
DotA2 is an interesting specimen to examine game design and balance, but interestingly enough we see a rather small amount of actual meaty analysis on it; likely this is because DotA 2 is such a ridiculously hard to examine game and is insanely complicated. To look too deeply into the game is to stare into the void, and at the expense of having it chrono us back, I thought it might be interesting to look at some more general concepts behind it.
DotA 2 has long been considered a game with fantastic balance, often being put into the hall of greats alongside Starcraft: Brood War. Over-glorifying this ignores Icefrog’s long term battle to balance games from the winning team getting out of control too quickly, or allowing teams that are behind from recovering too easily. Every patch, he looks to find the sweet spot between these two mechanics, allowing for a reward for bringing yourself to a winning position and for punishing for becoming overconfident.
I wanted to look at what Mark Brown (GMT) refers to as “Positive and Negative Feedback Loops”. In DotA 2, these can be filtered into several different concepts that have become popular in our dialogue over the last few years. These ideas come in names such as: Snowballing, Comeback Mechanics, Momentum, Map Control, and others. Each of these concepts look to either accelerate the team or hero that is ahead, or look to bring that advantage back the more they are ahead.
A “Positive Feedback Loops” is pretty easy to explain – a player that is ahead is given advantages to become even more ahead. GMT gives the example of Call of Duty’s Killstreaks Bonuses – getting enough kills will allow you to get bonus airstrikes, which may allow you to get access to helicopter attacks, which will get more kills to get to the next stage, and so on. Simply put, the kills start to snowball once you accrue enough.
Experienced DotA 2 players should recognize this concept immediately. DotA 2 as a whole is built with the underlying core of Positive Feedback Loops – a player that kills another hero receives gold and experience, which makes them stronger and leads to the easier ability to kill more players, where they will grow stronger and earn enough gold to buy powerful items, which allows you to be even stronger, and so on. Not only that, but kills often lead into destroying towers, limiting map control to the other player (making it harder for them to farm safely), opening the map for killing more neutral camps and creeps, and allowing a stranglehold on the enemy. Teams with map control also get easier access to Roshan and the many bounties he provides. The fundamentals of how DotA 2 inherently functions lead to snowballing and momentum – to change these ideas changes how the game is played dramatically.
Its easy to go back to Pre-The International 2 Na’Vi and remember them pushing down bases to weaker teams by 10 to 15 mins, or The International 4’s uneventful finals where games were quickly decided by the 10 minutes mark, with GG’s being called by 20 minutes. Arguably now, we are seeing games be decided very quickly, as Positive Feedback Loops, or perhaps the lack of enough Negative Feedback Loops hinder the possibility for teams to comeback easily (though, the recent Captains Draft Finals may have something to say about that).
Remember when this networth distribution meant the game was over?
(source) youtube I guess XD
Negative Feedback loops provides the counter-point to Positive Feedback Loops in DotA 2. GMT defines them as an advantage being given to a player that is losing or a disadvantage given to a player that is winning that attempts to prevent the lead from becoming out of control. DotA 2 has had this mechanic within it to different degrees. Within the game naturally different concepts and items allow players to come back. Scaling bounties for killing heroes on a killing streak has always provided greater gold returns, and killing heroes of higher level give more experience than lower levels, which helps a player behind get back in the game a little easier if they are able to kill a cocky player out of position. Defending objectives that are being pushed usually offers an advantage – particularly high ground positions (notably the main base) where the defending team will have a bonus from being hit and have more vision over the defending position.
Over time, Icefrog has looked to experiment with Negative Feedback Loops – we’ve seen him give a “comeback mechanic” where the team that is behind gets extra gold based on how far behind his team is, or how poor (or rich) the heroes that get the kills in relation to the heroes that die. This helps smooth out the curve from snowballing, though it has been done too dramatically in the past (most notably after The International 4 in version 6.82, which was slowly walked back after numerous patches). Icefrog has also changed some mechanics, such as allowing an extra Glyph of Fortification after Tier 1 towers are destroyed, in hopes of slowing down “Death Ball” pushing teams. He randomized when Rosh respawns, to make it harder to secure for teams that are ahead via map control. He even added Wells in the base to allow defending teams to heal (though, those have been removed from inside the base).
Icefrog’s battle has been to reward teams for playing well but not so much that the lead becomes insurmountable. He wants to counter-balance this by allowing teams that are behind to not fall too far behind and not give up from any small lead. He ideally wants to provide enough incentive for teams to be able to play themselves back into the game from any disadvantage, but not so much that teams that are ahead lose their advantage too easily. To this day, he still is struggling with that balance – recently removing wells from inside the base from preventing from base defense becoming too easy, allowing pushing teams to glyph their pushing creeps from taking damage, among other measures. He also is tuning the game from snowballing still, making early kills less rewarding than previously, and trying to promote late game play (so teams that are too invested into early game play are punished by teams that can hold on) by some more complicated ideas. He’s constantly looking at this balance, fine tuning heroes and mechanics within the game to get the gameplay just right.
So if you think the game is looking a bit instable right now – don’t worry, Icefrog’s looking at it too.
post-note: sorry, this stuff is probably pretty obvious, but I thought it was kind of interesting to look at some of GMT’s concepts and see how they can apply to the game we love. DotA2 is widely ignored or often discounted in Game Design circles, and I think its cool to take a look at these things occasionally. Something I didn’t mention was how Positive and Negative Feedback Loops affect player mentally and psychology and how they lead to bad manners, feeding, and flaming, but I’m sure you can figure that stuff out.