Escorts: A proposal

[nehal]

There is, I firmly believe, a fundamental problem with Escorts in TOME. I'd like to discuss here an analysis of Escorts on two fronts: First, the issue as a matter of game behavioral issues; Second, the issue as a matter of game design. Once I've laid out my case, I'll suggest some potential changes.

I think it valuable to preface this with at least a brief introduction to give some insight into my attitudes and feelings towards gaming: I've been playing roguelikes for...longer than I can really remember. The first one I recall really spending time with was Omega, and thereafter years were lost to ADOM. I've played countless others(haven't we all) but those were the two that really grabbed me before I found ToME. Outside roguelikes I've recently become enamored with Dark Souls and its brutal-but-fair take on gaming: Every single death is unmistakably your fault and after a few hours you are sleepwalking through areas that you had previously thought were abusively difficult. I'm also a big fan of DotA2 - which I came to by way of HoN - which I came to by way of LoL, my entry into the genre; Amusingly, it is also a game where every death(or loss) can be seen as being tied intrinsically to player skill and experience - albeit not always yours. I bring these up because they highlight two topics which we'll revisit: brutal-but-fair and whose-fault-is-it.

Before I get entirely off-track, lets talk a little bit about what is acutely wrong with Escorts.

Rainman, the game

The average Escort quest goes something like this: Load a level, ignore the popup, move from chokepoint to chokepoint with the escort's AI intentionally trapped behind you and the monsters 'in front' of you. Eventually get the Escort to the portal and choose your stat/skill/category.

The memorable type of escort goes something like this: Load a level, you're in the corner of an 5x5 room with an escort next to you and a dozen enemies next to the both of you. Shrug your shoulders and focus on getting out alive.

Alternately: Load a level, keep the escort behind you through the first chokepoint, but then they see an enemy and decide to bolt in a random direction at which point they run into other enemies, causing them to bolt in yet another direction while you hopelessly try to keep up with them and the enemies they've got chasing them.

The problem as it stands here should be apparent. While the basic concept seems fun and enjoyable: "protect NPC" - the actual behavior of this system in the game is innately non-enjoyable due to environmental factors entirely outside the player's control. Historically, the AI had a kamikaze "portal or die trying" attitude which led them to charge full-speed ahead, often into imminent danger. Currently they have a panic-first-portal-later attitude which leads them, ironically, to often charge full-speed ahead into imminent danger. Ultimately, one ends up with the feeling that they are playing _against_ an RNG, rather than playing _with_ an AI. Successful escorts often feel less like a job-well-done and more like glad-nothing-happened. Failed Escorts feel completely outside one's control.

Additionally, you'll often end up in situations where the Escorts you save are onesones you kind of don't need - either due to stat or skill mismatch with your character; and the Escorts you lose are the ones that would provide you with valuable skills or stats.


That being said, we also need to look at the game design principles that make the above feelings and behaviors unsound.

Making Losing Fun

Players, especially roguelike players, don't mind losing - but only so long as that loss fulfills one of two(if not both) important requisites:

They have to know it was their fault - this is where we reach the 'DotA problem'. One thing a player absolutely needs to be able to happily accept defeat is to know it was their fault. DotA(and its ilk) create terrible suffering in their players by blurring the responsibility. Without getting philosophic about DotA, the problem comes down to the fact that due to the sheer amount of moments and events in the game - directly involving the actions not just of the player but of their teammates and opponents - it becomes extremely difficult for a losing player to identify their critical mistakes withing the game. This leads to the genre's most defining characteristic, a vitriolic community full of ill-tempers and rage. The problem is that no matter how badly a player might be playing - if you give them a plausible scapegoat they will take it. While 'The RNG hates me' is a scapegoat - there's always an internal recognition that no, you can't blame something as dumb and chaotic as an RNG. But when presented with a 'logical actor' like an NPC AI or another human player - it becomes very easy to blame then for the 'overall' failure even if the player has made mistakes. This leads to the feeling of being unfairly punished by being anchored by something 'worse' than they are.

When escorts behave illogically, in the player's eyes, they then become a scapegoat that the player feels has robbed them of a victory. Sometimes this may be true, sometimes it may not, but in either case most players will walk away feeling robbed by powers outside their control.


They loss has to have equivalency - this ties in with the other principle I noted earlier: brutal-but-fair. Dark Souls is not a kind game. You'll often be thrown into situations that will kill you instantly without mercy but, with a few more deaths and some time spent exploring the situation, you'll start to learn what you've been doing wrong and upon later visits be able to handle it in your sleep. That's because the game while being difficult, balances that difficulty against a player's experience - as you gain more experience the difficulty goes down and and you feel progress and advancement - oftentimes to the point where a particular challenge _never_ causes you further issues. You're paying for any losses with the knowledge, paid back later, that your loss gives you more insight and better experience to help you succeed later.

Compare this, then, to Escorts - who will oftentimes place even an experienced player in a position to have no power to stop guaranteed failure. I wouldn't even say this is a rare occurrence, given the totally random nature of not just when the missions occur, but also the level conditions in which they occur. How many players properly get through a game with 8/8 successful Escort missions? Very few, I'd imagine, and should you be able to take a look at all of them I'd imagine the majority of their successes were based on luck rather than hard-won through experience and skill.

This is compounded by the fact that every escort you lose is, fundamentally, a total loss for that character. With eight escorts throughout the game we're looking at the potential for upwards of 16 stats points - which is 5 extra levels of stats; or 8 skill points which is, roughly, 8 additional levels worth of skills. If you fail an escort quest you've just lost a significant potential character advancement in an unrecoverable way. In the case of characters wanting to attempt to go AM, this is even more significant, as they may be robbed of Escorts they are able to betray.


Solutions?

The problem then is that we are not only robbing a player of important benefits, but that we're giving them no clear indication in many cases of how they failed, oftentimes because they didn't and it was just some goofy AI behavior. When looking at solutions, then, we need to look at ways to triage the problems identified by the above two issues.

Blaming the player

The primary option for allowing a player to unambiguously accept their own failure is a rework of how escorts behave to create a more 'sane' approach that players will be more willing to accept as 'valid' behavior. While the 'proper' solution would be to create a solid AI that assesses the 'best' solution, the behavior can likely generally be found in a combination of:
-Escorts should never leave player sight
-Escorts should never enter terrain that they(and therefore the player) has not explored
-Escorts should assess the threat of individual enemy types

Along with that, a level generation change should be considered:
-Escort levels will never start with an enemy within the same 'section'(room/hallway/clearing) as the player+escort

By minimizing the chances that a player watches an escort "do something stupid" or be put in an instant-loss scenario then the player becomes more forced to accept their own fault when an escort fails.

Giving something back

The perhaps easier solution is to provide some sort of alternative in the case that an Escort dies, the player still receives something back. While initial very new-player losses due to failures to understand retreat logic and escort behavior can be 'paid for' with that knowledge, eventually we must concede that over time most Escort failures will result from RNG issues. In cases such as that, providing something to the player even in result of failure will offset the feeling that they've been outright cheated. This is already implemented slightly into the game with the ability to betray escorts to gain access to Anti-Magic. You're giving up the standard rewards to gain something else.

I'd propose that whenever an Escort dies, they have some sort of significant item drop. Given the absolute rarity of escorts throughout the game, and the fact that they primarily front-load the earlygame, I don't think it is outright to absurd that escorts should drop either a purple or randart appropriate to the dungeon level.

The game balance comes from twofold randomness:
-First, that the item generated can actually be used by the player.
-Second, that the item won't likely be useful for long enough to have a prolonged effect on the success of the player.
Mostly the latter point. Most Escort rewards have an ongoing effect on the course of the game and, in some cases, can provide lasting late-game changes when a player is able to grab a useful category or repeatedly skill up a class-relevant skill. Giving items with a short-term utility provides a consolation prize that will offset any frustration the player feels upon failure. It also provides a 'useful' mechanic for new players such that they can kill Escorts to pick up useful items that can be used to provide them a leg-up to learn the early-to-mid game.


Given the timeline of things, I clearly don't expect these sort of things in 1.0 - but I think they'd help with retention of newer players by giving them an easier route to learning the earlygame, as well as assuaging longterm feelings of players on their thousandth Escort mission the RNG has made impossible.

[TheRani]

I'd be thrilled if my temporal alternate timeline self escorts would stop zapping me with Dust to Dust while I'm trying to fight the things that are trying to eat them.

[Kurayumei]

I agree, nehal. So often I would be helping an escort by fighting a bunch of monsters at a chokepoint and then they run off and get themselves killed. I thought I was supposed to be escorting them? Why did they run off? It is really annoying. Would being able to give escorts instructions or control them within a limited range be an easy way to help avoid this problem? In that case if they die, it's entirely the players fault.

[bricks]

Wow, long post. I'll admit I skimmed it. The comparisons to DoTA, HoN, and Dark Souls are kind of lost on me, I'll admit.

If there is a design problem, it's that failing an escort mission feels like a loss, instead of winning it feeling like a bonus. No doubt this is due to the permanence and utility of the talent rewards. I still advocate changing the rewards to charms so players don't "build" themselves around picking up obscure talents.

[omeg]

Any game mechanic that causes player to lose (lose escort, life or something else) and the player can't do anything to prevent that is a bad game mechanic. Unless you play roulette or something.

[supermini]

If there is a design problem, it's that failing an escort mission feels like a loss, instead of winning it feeling like a bonus. No doubt this is due to the permanence and utility of the talent rewards. I still advocate changing the rewards to charms so players don't "build" themselves around picking up obscure talents.

I agree, but on the other hand, you have to move that utility somewhere else. Conditioning and celestial/light are extremely effective to the point of necessity for some classes/races (or at least appearance of necessity). Premonition, chant of fortitude, arcane eye, mobility tree...The effect they have in the long run is not negligible.

You could have trainers in the game world that could give you the locked category for a certain price, or require you to do a quest to unlock it. Lets say we have 1 trainer per city.
If maintaining randomness is a requirement, you could have random trainers spawn each game at level 20(to prevent scumming). Escorts could give you a completely random item (like chests do now) instead.

Maybe it's something to consider for after V1.

[bricks]

Or you could fix those classes, whatever those happen to be.

[supermini]

Or you could fix those classes, whatever those happen to be.

It's not an issue of fixing anything.

Escorts rewards are extremely useful. Some are so useful (conditioning on ghouls and celestial/light on archmages, and chant of fortitude for magic using classes is what comes to mind) that missing out on any of these is a huge disappointment, which goes back to your original point that losing it feels like a loss rather than winning it feeling like a bonus.