- 1:35: Starting, Introduction
- 1:38: Georg Zoeller: KOTOR was about 60 hours of content, SWTOR is an astoundingly more than this. Managing this massive volume of content is the focus on this talk.
- 1:41: I talked two years ago about DA:O, about making a massive title. I had not idea how small that was. A single section of Tatooine has more creatures in it then the entirety of DA:O. A single class will spend between 6-10 hours on Tatooine alone; 15 if you are a completionist. This is only the class content.
- 1:43: How do you create and manage content on that scope? You need lots of resources, smarter workflow, and smarter tools. The second and third will be what I focus on.
- 1:46: One testing weekend can generate 100,000 pieces of testing feedback. The different types of feedback we are dealing with are: Testing forums, screenshots, /bugs, context sensitive surveys, chat log, customer service tickets, in-game feedback. This is all the active feedback, a small section of players actually supply. We take this player feedback and filter it through our metrics to determine what actual actions are needed.
- 1:53: All of these metrics are important when you add the spacial component. Where do players clump up, where do players die, where are the complaints coming from. Using heat maps we can look at planets and see exactly where people are dying but more importantly what class of what level are dying.
- 1:57: We developed a content tool that we use on SWTOR to analyse any content with x,y, and z data points. We can then quickly share that data within the studio using the tool kit. We can generate heat maps that we can then click on to instantly go to that map and location. For an example, we will look at Tython. First we get player feedback that is useful but not normally clear enough to be useful by itself. We can then jump into the tools to pull up metrics to examine the areas in question (in this case a map on Tython).
- 2:00: We studied death a great deal. This is why we re-examined the death penalty system. This is why we added the medical probe, to give players that single player feeling of reloading after a tough fight to try again. We found that death penalties are generally negative and unnecessary. After metrics are drawn up, we add content data to compare the metrics to content. The data can tell us things such as, "are the players on level to the monsters they are fighting in general?" We tend to find that many places people are two levels under level. We can then take this information to alter the content to make it more fitting.
- 2:05: We also track whether the players are playing solo or in group and this helps us to find when players are or aren't grouping and whether the content is too difficult or not. Another example is looking for player feedback when they are confused. We can then look at the chat logs for key phrases such as "How do I.." "How does...", etc. We can see when people ask these questions in chat and what they are asking about to help us find tutorial points or places for clarification. These word clouds can even be placed over a heat map to make it that much easier to see where a player is standing when they ask a question.
- 2:10: We can also use the chat log for "bug," "broken," "bugged," questnames, etc. This helps us find issues that players don't take the time to do a bug report on. We can use these tools to check performance optimization. We use drawcall metrics from bot driven sampling systems then check this data to tell us how many rendered, active objects are in certain areas that a player's computer will have to generate. We can also do this with terrain and foliage. Knowing instantly where speedtrees are located vs. static trees to see where we might be over using or under using certain assets.
- 2:15: We can generate all these maps in 3D to let you find metric points without heavy filtering. This works well because the game is in 3D and we can look on multiple levels adding to verticality. Other applications include behavioral analysis. For example, given a perfectly symetrical t-junction, 68% of the players will go right. This is good to know when designing content. We also use fog of war to find what areas people never see. Without fog of war, our testers would pass through entire planets without a single person going to some areas. After adding fog of war our heat maps showed players going to every piece of the map seeing all the content.
- 2:25: Question: Your metric tool set is extremely complex and impressive. Was this integrated from the beginning or added in later?
- Answer: We didn't have this integreated in the beginning. We found as the game grew it was needed to look at the informaion in a better way. We examined the Hero data and slowly filled in the tool.
2:25: Q: How do you stop or catch leaks?
- A: With all the data tracking we have we can see a single ability used at a single location for every player. If a leaked video shows a player wearing a certain piece of armor or doing donuts on a speeder in a certain zone we can find them very easily.
- Talk Over
Seriously impressive. Gogo BioWare ^^