Thinking Fast & Slow – Experienced Utility in Game Design

Utility is one of those words which mean something different to psychologists than it does to us normal humans. People aren’t who researchers would like to assume we are. We know what made us happy, or hurt us. However, our memory is subject to how we process the world – something we’re not rational about. So it’s our memory of events which matters, not what actually happened that determines if we like something, or not. That understanding ought to impact how we design games.

First, some background on memory – or experienced utility to use the psychologist’s term. From Introduction to the Principles of Morals and Legislation by Jeremy Bentham, “Nature has placed mankind under the governance of two sovereign masters, pain and pleasure. It is for them alone to point out what we ought to do, as well as to determine what we shall do.” The term utility, in the psychological sense, includes pleasure and pain in the its scope, and our recollection of them. It’s completely subjective, and verity has nothing to do with it.

Don Redelmeier and Daniel Kahneman did a pain study on… a really painful medical procedure. They tracked two different measures of “utility”: pain at any given moment and a global retrospective rating. They charted the results for a variable length and variable pain intensity procedure. Summarizing their results as two rules:
* Peak-end rule: The global retrospective rating corollated to the worst level of pain, and at its end.
* Duration neglect: Duration had no effect whatsoever on the ratings of total pain.

1. If you’re in maximum pain at the end of a procedure, you will feel the whole procedure was more painful.
2. The average level of pain maters, if it matches the pain at the end.
3. How long a procedure lasts doesn’t matter to your perception of pain-level.

Applying that to medical procedures, if you’re goal is to minimize pain, lower the peak intensity and gradual relieve pain at the end, rather than reducing duration of the procedure. If you’re goal is to reduce the actual amount of pain experienced, reduce the duration, even if doing so causes patients to have a more awful memory. Reduction of the actual amount of pain experienced is the rational goal, but not the one that participates will select given the choice. They’ll opt for a duration increase of 50%, because their memory of the pain will be less severe.

So, what the hell does this have to do with games?

Here’s what getting utterly stomped on looks like in Hearthstone (click for larger):

Priest go BOOM!

Quick recap: The player on the top has utterly crushed the bottom player (me) with a score of 34 to 2. (The first to 30 wins.) Interestingly, in this particular case, I had been doing fairly well, up until the last two turns, at which point I got utterly stomped. Lopsided defeats like this are painful, and do happen fairly often.

Or do they?

Is it possible I’m just remembering wrong? Humans do widely overestimate the frequency intense events…

Hypothesis: By applying the psychological concept of utility, we can reduce the intensity of the negative memory formed by loosing so badly and quickly. In doing so, we can increase near and long-term player retention.

Goal: Reduce the intensity of the loosing player’s “pain” by obscuring the magnitude of their loss. That should result in an increase of immediate replays after substantial losses and corresponding increased long-term customer retention.

Specific actions to be taken:
1. For the looser, do not display a negative (or zero) number for the player’s health (visible in the lower right hand corner of the player’s character.) Assume that players will be lazy, and not do the math, especially because they (initiatively) know that it won’t come out in their favor.
2. For the looser, do not display the damage of the killing blow. This extends the logic from action #1 to it’s logical conclusion.
3. For the victory, do display the negative value and killing blow damage. Those are positives, so hype the hell out of them, possibly even more than they are already. (Supported by the psychological concept of loss aversion.)
4. Prevent end-game “burst” damage. High mana costs cards (the blue crystal in the upper-left) do more damage – that’s desirable and to be expected. However, when a player takes 15 hit points, or more, of damage in a turn, that’s “pain.”
5. Change Archmage Antonidas (the minion in front of the winning player) from “Whenever you cast a spell, put a ‘Fireball‘ spell into your hand.” to “Whenever you cast a spell, put a ‘Fireball‘ spell into your hand*, at the end of your turn.*” Using the existing card, a player can do at least 12 points of damage in a turn, even if they don’t already have a ‘Fireball‘ card in their hand. That is “burst” – arguably too much.
6. Review logs for other combinations which have the same issue. For example, a Shaman could do more than 25 hit points of damage in a turn, but that has been fixed.
7. Ramp down the effect of repeatedly cast spells. A gradual step down will result in a less painful memory. Applying that here, each subsequent ‘Fireball‘ spell should do one less damage than the previous one. Even though the result would have been the same, in this case, the intensity of the negative memory would have been reduced.
8. A more substantial ramp down of 50% would also potentially cause players to reconstitute their decks instead of every Mage deck always having two Fireballs in them. At a minimum, one would expect less back-to-back casts of the same spell on the same target. If those would be desirable outcomes, for other reasons beyond the scope of this document, then that would also bear testing.

How to measure success:
1. Time until next play. Desired outcome: Down for the looser. Remain unchanged for victor.
2. Time until next play of the same class. Desired outcome: Down for the looser. Remain unchanged for victor.
3. Is there a deck composition change. Desired outcome: none, just want to watch for unexpected results.

Why is this important:
It’s critical that players not come to the conclusion that Hearthstone is a “pay to win” game, otherwise huge numbers will never play. Legendary cards, indicated by the orange pip under their name (visible on the Onyxia card in my hand) are extremely rare. For example, after expensive play (261 wins, and more losses than that,) I have five legendary cards, two of which I spent money for. Getting defeated so roundly by a legendary card drives home the concept of paying to win. Spend $50 and you’re statistically likely to get 2.

Blizzard must be ever vigilant at managing the player’s memory of the performance of legendary cards, because, for better or worse, they’re going to be assumed as being bought – even if they weren’t.

(Lots of assertions in this section which bear further study and supporting proof, but that’s for another day.)