[2021] Subs in randoms

Ranked is the closest in terms of player behavior to Randoms, while Coop has much lower stakes and is played a bit differently.

This doesn't mean we'll go directly to Random Battles from here. We'll look at the data we got and see if it paints enough of a picture for us to be able to predict general trends for introduction into Randoms. It's not always just about checking whether current balance state is ok, but also about how behavior changes over time, with different balancing factors being applied in the timeframe. Eventually, if everything goes well, subs will see Random Battles, but we'll also be looking at what accompanying content should be in that patch, as well as the neighboring ones. It's all important.

We do have a "feelings" objective, but we need to quantify it. The same goes for gathering that type of qualitative data - we gather it, but we need to match player responses to quantitative data describing their behavior to make it actionable.

Just getting a response that says "i don't like it" is not actionable, because we don't know which of the multitude of factors created the sentiment. If the cross-reference shows that the player consistently dies first and has more miles on average than other people who died in a similar time frame (over many battles), then one of the conclusions you could make is that the ship is too fast. It's just an example, but the more data we gather, the higher the chance that we will have enough different behavioral trends that cross each other to be able to narrow the problem down to a relatively small area.

We also routinely see player behavior not matching their answers. This also increases the need to have a lot of data over time, so that we can be sure that the current behavior pattern (which contradicts the qualitative answer) is their actual play pattern and not something that is seasonal (as in: just happening now, influenced by outside factors). This is very normal for sentiment analysis.