devtrackers.gg
Hello Heroes!
Over the past couple of months, we have been hard at work preparing our combined league structure and are excited to bring Storm League’s Preseason to the Nexus in the next major patch! We will go into more detail about that in the near future, but in the meantime, in our most recent Reddit AMA, we got some questions about the 2019 Gameplay Updates and wanted to share some insight into our thinking behind them. We’d like to discuss our goals and share some data to show what has been happening since we rolled out these changes.
There was a recent post on www.Heroeshearth.com that went through some in-depth analysis of Hotslogs data and plugged it into some machine learning algorithms. The author, Ghostdunk, did an incredible job putting together the data and graphs in an easy to read format - we highly recommend you take a look at it! While Ghostdunk was working with data from Hotslogs, we’ve got the benefit of having access to much more data and wanted to share some additional context. While end-of-game level differences is a useful datapoint, it does not tell the story of everything else that led up to that point; addressing or changing that wasn’t a focus for our 2019 Gameplay Updates. That said, we have been analyzing the data from the last couple patches compared to the past and are excited to share some of that information with you!
Our high-level goal for the 2019 Gameplay Updates was to make matches more competitive, especially in the early-to-mid game. To do this we were expecting the following outcomes:
- Reductions in the time that teams spent on differing talent tiers
- A push back of the time that ‘snowballs’ tended to begin occurring
We spent months working on different designs and number changes to accomplish that goal, and each iteration pulled us closer to the major culprit – burst experience advantages in taking early game structures. Teams that won the first objective would often focus on getting front walls of each town down before ever attempting a Town. At these early levels of the game, there was so much experience in these structures that teams that won the first major objective were often able to put the losing team into a massive disadvantage as they gained a two to three level lead heading into Heroics. At this point, the team with the level advantage could control the map and force uneven fights on future events, effectively putting the enemy team into a chokehold that too often limited choices for the lagging team and often dictated the outcome of the map.
The MOBA genre revolves around these types of power advantages, but in ours, it felt doubly bad to be down since it wasn’t a single Hero or two on a team taking over, but the whole enemy team having both statistical and talent-tier advantages. As most players know, it is extremely difficult to team fight in our game unless you are at least on the same talent-tier as your opponent. This greatly limits the options for the team that is behind. We figured we could combat this disadvantage by taking some of the raw statistical advantages away from teams that were winning early, and give them an instant strategic advantage, plus a long-term experience investment. This led us to our final changes:
- Less experience in all structures (removing the instant statistical advantage)
- Periodic Catapult rewards for killing Forts (strategic advantage)
- Increased trickle experience for killing Forts and Keeps (long-term statistical advantage)
Stas, our wonderful technical director has been crunching our data in order to see if our goals were being met or if we fell short, and here is the high-level bulleted findings and some graphs to help players interpret them.
- The average experience differential between teams is down about 13% in minutes 5 through 20. Reaching a peak difference of 19% at the 9 mark. This shows us that teams are much closer in levels throughout the entire game, but most importantly during the critical mid-game phases.
- As you would expect, if teams are closer in experience levels throughout the game, they should also spend less amounts of time at talent-tier disadvantages. The data very effectively backs that assumption up as well. Again, we see an average of around 13% less time spent with talent disparities, with a peak difference of 23% at level 13.
- Another interesting thing we learned from this is that the win-percentage of a team with any kind of level gap has slightly increased. While this may seem counter-productive to what we would actually want, it is an expected occurrence; since it is more difficult to gain a level lead now, the performance requirement for gaining that lead is higher. Meaning that teams that can start creating these larger gaps are highly out-performing their opposition. Since the focus of our changes was making sure that early game leads didn’t lead to late game ones, we’re okay with teams that have earned a lead in the later stages of the game winning more often. We think this is a good reflection that the better team is still winning most of the time.
Overall, the most important thing to take away from all of this, is that even when losing, the game should still feel close. You can still compete, fight the enemy team, and not be at too great of a statistical disadvantage. If the enemy team still beats you, it’s because they’re actually the better team, and not because early game mistakes lead to unsurmountable level advantages that don’t feel fun to play against.