Life Expectancy Feedback Heuristic

[SteveP]

As the game evolves and particularly as the predator AIs improve, it will help to balance availability of critical resources such as weapons, food and clothing if there is some method to gather statistics as to player life expectancy. This could also be used to help tune the game to a player's ability level and experience. I would be very interested to see the life expectancy statistics. I suspect there is an abundance of wolves or they are just too efficient since they seem to kill me fairly efficiently after a few days when I try moving about ML.

The developers could use the feedback to manually tune resources or let the game apply heuristics. The latter approach would broaden the appeal of the game.

Steve

[octavian]

You mean something like the Steam global achievement stats?

[SteveP]

You mean something like the Steam global achievement stats?

Well the mechanics might be similar. Tracking the game and recording the number of days survived at the time of death should provide an interesting set of data when aggregated. Other information that could be collected would include the manner of death such as by wolf, bear, freezing or starvation.

The global achievement is for specific objectives rather than recording mortality stats.

The output, if in user readable format, could include a histogram of the average days survived. In the case of a heuristic, the data could then be collected on the individual's performance and make the next run a little easier if he/she died from lack of supplies or something. I wish there was a tunable factor such as wolf aggression or detection range. A heuristic is a tunable characteristic of a game that is automatically updated as the game "learns" the capability of the player.

There are two goals: 1) provide overall game tuning 2) provide individual feedback to improve game experience. The statistics would also have appeal to the players since they can compare their performance to that of others.

Steve

[octavian]

Sure, but, is this how game development works? I don't know. It seems a bit, weird, to me. What are the guidelines? I mean, if most people die by wolf in under 20 days, what does that mean? Wolves need to be changed or what? And if you tweak the game based on the average whatever, you do it to what end? Make it easier for the average player? Harder?

Also if the game adjusted on the fly to suit you it would be, peculiar, because they're not delivering a preset experience anymore. So you can throw yourself at wolves for an hour, die 50 times, and then have an easier game? Just look at the leaderboards and hibernation, that's not a good idea.

[DrMembrane]

I'd just be interested to see what my average death was cause by and when it happened.

[SteveP]

Sure, but, is this how game development works? I don't know. It seems a bit, weird, to me. What are the guidelines? I mean, if most people die by wolf in under 20 days, what does that mean?

A heuristic is a game feature that adapts automatically. I'm not suggesting that they need to incorporate heuristics directly; it means that you can collect a TON of statistics to provide a meaningful interpretation of how to improve the tuning of the game.

I would expect to see the life expectancy start low and then increase with experience gained by game play until one is able to expect a reasonable game length 50% of the time. A skilled player should be able to last indefinitely on Voyageur but not necessarily on Stalker. Pilgrim should be easy enough to permit lots of exploration but still present opportunities to perish.

If a player dies of wolf attack, the immediate assumption might be that the wolves are too abundant and too ferocious and too skilled. That's not necessarily the case; there are many other factors to consider; is this player starving for example? or very low on food? that might be causing players to seek out conflicts with wolves in order to have a desperate last chance to evade starvation. I would collect statistics on a ton of stuff: clothing condition, clothing warmth, weapons, # conflicts, calories in inventory (if any), weapons and so forth. It's no surprise that a starving individual is forced outdoors where the scavengers will eventually take care of him.

Also if the game adjusted on the fly to suit you it would be, peculiar, because they're not delivering a preset experience anymore. So you can throw yourself at wolves for an hour, die 50 times, and then have an easier game? Just look at the leaderboards and hibernation, that's not a good idea.

The intention is not to make the game easy to trick; it is to maximize player enjoyment. That doesn't mean making it too easy. If its too easy, I don't enjoy it. Some people only want to win and don't care how; I don't think we should necessarily cater to them.

I really think we are struggling with tuning. Are we doing this manually by developers altering the configuration? How often does it happen?

Here is a link to the technical issue post regarding game tuning in general:

Tuning the Game

[SteveP]

I added some relevant comments on Tuning the Game post about stochastic feedback for tuning. Stochastic can be either manual (ad-hoc) or automated (heuristic).