The psychological phenomenon of loss aversion refers to how players feel losses twice as powerfully as victories.

For example, Bite Fight's PvP is a simulation based on probabilities related on character skills, and players voice this feeling many times per week in the community forums.

If you don't want to create a pay-to-win game, but you do want to let the worst players win often enough to feel good about it, how can you do that?

The question has two parts:

  1. How do you handle it technically? Do you use some kind of math techniques or a memory based simulation to avoid many loses in the row for a certain player?

  2. How do you handle it from the community point of view? What do you do about complaints like this on public forums?

  • 11
    \$\begingroup\$ It's sad that so many people want to close this question without even providing any advise how it could be improved. This website could really use more gamedesign questions and less "debug my code" questions. \$\endgroup\$
    – Philipp
    Apr 26, 2016 at 8:36
  • 7
    \$\begingroup\$ @Philipp Indeed. When I saw the down votes, it reminded me a question I opened in Meta site that this SE lacks of game design questions. And if Psychology is not part of the game design, then what is? We need to understand that game design cannot be strict questions and there might be a good possibility that it could be slightly more "too broad" than "debug my code" questions. \$\endgroup\$
    – Tasos
    Apr 26, 2016 at 11:26
  • 3
    \$\begingroup\$ Coming at game design as a programmer, I find these problems harder & discussions about them more valuable than stuff like "how do I make X code do Y thing." \$\endgroup\$
    – Pikalek
    Apr 30, 2016 at 15:33

3 Answers 3


I see this question has a number of close votes for being too broad or opinion-based, but I think a reasonably sourced overview can be provided within the scope of a StackExchange answer - I'll take a stab at that here.

Sid Meier talked about this problem in his 2010 GDC Talk "The Psychology of Game Design (Everything you know is wrong)" (this gave me an error in Chrome but I was able to view it in Edge - your mileage may vary) Some of the aspects he covers:

  • The Winners Paradox: although failure in games is in some sense much more acceptable than failure in reality, players are accustomed to winning games if they put enough time in.

  • Players tend to perceive good results as a well-deserved consequence of their actions, skills, and strategy, and bad results as vindictive punishments or glitches.

  • Humans have a very distorted view of probability. If you show players 3:1 or 4:1 odds in their favour, they perceive this as overwhelming advantage, and so losses (even though they should occur 20-25% of the time) feel shockingly negative.

  • Breaking random events into a number of smaller stages can help meet players halfway here. (This is why in many versions of Civ battles play out in multiple rounds). By rolling the dice multiple times for incremental results, probability drives the results toward the expected value, and you see fewer surprising outliers. When the player does lose, they can see it coming as a trajectory of several losses in a row, rather than a single bolt of lightning out of the blue.

1) Do you use some kind of math techniques or a memory based simulation to avoid many loses in the row for a certain player?

Quite often, yes. Sid Meier talks about doing just this in Civ in the talk linked above. XCOM: Enemy Within also does this on its lower difficulty levels:

"On Normal and Easy modes, there's a bad streak breaker," says Gupta. "It's stronger on easy. On Easy, it's very hard to actually lose soldiers, and the fewer soldiers you have the stronger this effect becomes, so if you start losing it builds to compensate. Of course, the aliens themselves have weaker stats on Easy and Normal too, and on Classic or higher difficulty settings this effect doesn't exist. Believe me though, on Easy, if you miss three times in a row you're not going to miss your fourth shot. It can be a 1% chance to hit and you're not going to miss that shot."

Even Tetris (in most incarnations) uses a shuffled bag of pieces to ensure you can never go more than a certain number of rounds without getting that one piece you need.

These types of systems provide a safeguard against a player getting "dice humped". Often, what we want as designers is a little variety/unpredictability/surprise, and not 100% unforgiving randomness. ;)

2) How do you handle it from the community side of view?

There are two major branches to this:

Listening to the community & maintaining open dialogue. World of Warcraft provides an excellent long-term case study of responding to community feedback - doing things like introducing ramping probabilities on certain rare drops so that players who've missed them many times in a row are more likely to find one on the next roll (similar to the example above from XCOM, this is effectively building the Gambler's Fallacy into the game probabilities, for the cases where it improves the player experience).

Even if you don't remove every pain point (some friction is necessary for there to be challenge), developing a reputation for attentiveness to balance issues can improve perception of fairness.

Cultivating a community that is accepting of failure. Dwarf Fortress's community rallies around the mantra "losing is fun". Dark Souls is lauded for its unforgiving difficulty (even enshrining it in the marketing tag-line, "prepare to die"). They have some things in common:

  • They encourage their audiences to self-select. Someone looking for a mindless reward-fest is not going to buy these games in the first place. Identifying as someone "tough enough" to handle these games is part of the player fantasy they advertise.

  • Loss in these games is not capricious. Even though a dragon could wander into your undefended fortress and slaughter all your dwarves in no time flat, that dragon wasn't just a bad die roll. It had a history in the world - it came from somewhere, it's going somewhere, and the fact that you got mowed down in its way is largely your own fault for picking that site and not defending it well enough. Everything in these games' worlds follows consistent rules, and a player who wields them well can handle even the toughest challenges the games throw at them. That doesn't mean guaranteed success for skilled players, but it means their skill makes a difference. When you structure your game this way, loss becomes an honest signal "there is something you can do better" rather than "the designer thought you should lose a bit more often."

  • 2
    \$\begingroup\$ Impressive. Thank you for the detailed answer. I will keep the question open for 1-2 days more to promote the discussion, but after that I will choose the "Best answer" \$\endgroup\$
    – Tasos
    Apr 26, 2016 at 11:28

There are good points in DMGregory's answer. I especially like the one where a win/loss is split into multiple minor wins/losses, which is taken from slot machines - when in doubt copy slot machines, because they're the ultimate game where (almost) all players lose, yet so many continue to play. Let's add some more points:

Use gamedesign to allow victories, even in defeat.

Ideally, an overwhelming one-sided victory should be very rare. So in a hockey simulation the loser can try for a consolation goal, in a strategy game they may want to at least take out a certain unit. In Titanfall there are super weak bots in every match so the player could at least get some kills. You can also leave the definition of victory open, such as in 8 player free for all deathmatch, where first place is so unlikely to achieve people don't expect to get it. That way weaker players can set their own targets, such as not being the first to die.

Use statistics to allow players to feel a sense of accomplishment after a loss

At the end of the game you offer statistics, like in HotS where you can see who made the most kills, dealt the most damage, the most other kind of damage, absorbed the most damage, died the least, contributed the most xp, healed the most, assisted with the most kills, etc. That way even if the match was lost, it is rather likely that you'll be best in some statistic. In Starcraft there is even a mark next to every statistic in which you played better than your average.

Players must stand a chance in PvP

Matching new players against the top players in pvp isn't fun for anyone involved. The challenge needs to match the players' level of skill, so in PvP you need a system that guesses a player's strength and (a) prevents totally unfair matchups and (b) provides a minor advantage for players suffering from a loss streak. (b) is key, because losing 10 times in a row can make players quit your game permanently.

Provide the option to train basic skills without losing

If there is any skill involved, players need to have an opportunity to train that skill without losing all the time. For example, people who have never played an fps will not stand a chance in a pure PvP game like the recent Star Wars themed Battlefield game - they would likely die more than 100 times before making their first kill. Even in a 100% multiplayer game, there needs to be an option to train against non-human opponents, no matter how dumbed down and basic it is. In Battlefront they solved this with a simple mode where they just spawn waves of weak enemies, in other fps games they solved it by adding a single player campaign, or by allowing to play games with or against any number of bots.

  • 1
    \$\begingroup\$ And then there is the very delicate matter of excuses. If i lose a game and my ping is not visible, i can always blame my connection, or perhaps my low-end GPU. If the ping otoh is visible to others, i cannot blame it if it's too good, but if it's apparently bad, i'll shout and aim at it. Show or not show technical hard facts, that may in fact well have relevance for the outcome? A separate discussion. \$\endgroup\$
    – Stormwind
    Apr 29, 2016 at 22:08

A lot of great stuff covered in the other answers. Here's my take on (the perception of) win/loss probabilities. In the case of PvP, consider different ways to track & show win loss scores either directly or in some sort of cooked format (I.E. player rank). Specific examples:

Puzzle Pirates ranks player performance relative to all of the other players locally & server wide. The developer wanted to keep numeric scores out of the game, this is expressed on the standings in the form of a descriptive word (E.G. Proficient, Master, Grand-Master, Ultimate). All of the puzzles have some degree of luck involved. Some such as sword fighting can be influenced by equipment. But I would argue that in all of them, player skill is the most important factor. Since the game makes this known, losing a wager is mostly on the player. This game allows player's wagering resources on matches & allows unranked competition, reinforcing the idea that the players opt into losing ranking &/or resources.

FIBS is a backgammon server that rates players using an open formula. If you win, your rating goes up. If you lose, it goes down. If you beat someone who is better than you, you will get more points than beating someone who is worse & vice versa. Your gain/loss is dependent on: difference in rating, the match length and your experience.

Often showing the stats, especially in a raw form, may not be good enough as humans often misunderstand probability (E.G. confusion about the Monty hall problem). You might be able to adjust for that by providing some context. For instance, the roguelike Brogue provides the following sort of information about combat:

The eel has a 63% chance to hit you, typically hits for 55% of your current health & at worst, could defeat you in 2 hits. You have a 70% chance to hit the eel, typically hit for 30% of its current health, & at best, could defeat it in 3 hits.

Notice that it bases everything on the current situation, reducing math for the player thus lowering the chances of death by math error. Also it gives both average & best case damage. It's not perfect - it doesn't account for simultaneous damage from different sources & the player has to choose to read it, but in generally, I think it does better than raw stats alone.

@DMGregory touched on filtering/tuning random numbers. Extreme runs of good & bad rolls from a vanilla RNG can & will occur given enough attempts. When you consider the number of rolls made in an MMO game, it is almost guaranteed that someone will experience such runs of luck. If you want to correct for this you need to filter the RNG output to remove these runs. My recommended reading on that topic:


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .