Diminishing Returns for Balance

Linear scaling stats has its problems. If a player spreads their points out between each of their statistics evenly or accrues small bonuses in one stat or the other, the effect is perfect... but min/maxers who find the best stats and jam all of their points into them can snap the balance in half.

One way to mitigate this effect is to make every point of a stat after a certain threshold have less effect on that stat than the last one. You can alter the coefficients as necessary for each attribute, and if you do a smooth enough job nobody will notice. The game works as it should and everyone's happy.

Of course, diminishing returns has its troubles. First, it's a lot more work, and takes a lot longer time to playtest and get just right. Second, if it's done poorly it just makes the game mechanics confusing; explaining to the player that, above a certain point, each point of Dexterity only provides 36.74% of the benefit of each previous point makes your game sound arcane and confusing. And if you don't explain it to them, astute players will feel cheated when they realize that they pushed all of their points into Strength and are getting no appreciable benefit from half of those stat points.

The second problem is it doesn't always solve the problem of min-maxing. There are usually at a handful of stats that are the most useful in any game, it's just that one of them tends to be more useful than the others. Diminishing returns encourages players to dump points into the "best" stat or stats until the penalty is hit, then dump more into the second best. If the system is not well balanced regarding stats in the first place, the net effect is that instead of focusing excessively on one, players will focus excessively on a small handful.

The third problem is inconsistency with player vision and roleplaying. You might want to play a very strong, hardy warrior who is very dumb, but games that take this trope too far make such a character impractical, as you'd have to sacrifice every other stat to make such a warrior, including other necessary ones like dexterity. Systems which abuse the trope result in characters with very similar stats, often leading to stagnant gameplay, the exact problem such systems were designed to avoid.

Like most attempts at balance, it's difficult to find the right setup that makes everyone happy, but some form of balance needs to exist in order to keep the game playable.


    open/close all folders 

    Alternate Reality Game 
  • Ingress:
    • The player level cap is 16, but the level locked Weapons and Resonators can only go up to Level 8. Instead, the benefits of Level 9 and beyond other than increased recharge range and efficiency and increased XM capacity manifest in the form of one-time Field Kits that provide various items as rewards.
    • Shields and Links can only provide up to 95% damage reduction; any higher becomes "excess" shielding that is ignored in damage calculations and only exists in case other sources of shielding are knocked out.
    • With regards to non-Shield Mods: As you stack more of the same mod, each successive one will have a reduced effect; for example, the first Common-level Multi-Hack will add 4 extra hacks to the Portal, but a second one will add only 2 extra hacks.

    Fighting Game 
  • A lot of fighting games use this to keep combos under control, causing each hit of the same type to cause less damage than the last until a certain amount of time has passed.
    • Some games will cause the first hit after an arbitrary number to automatically whiff even if the enemy is still well within its hit box, giving the opponent enough time to recover and counter.
  • In the Super Smash Bros. series, each game has a mechanic known as "stale-move negation", that causes an attack to become weaker the more frequently it is successfully used (as in, it hits an opposing hitbox). In Melee, it wasn't much of a factor, as while it caused attacks to deal reduced damage, it didn't affect the knockback they dealt (thus your finishers will still KO at nearly the same percent they would fresh regardless of how often you used them). In Brawl, it's much more severe, with knockback now being affected; the knockback power of a move will decrease rapidly with each successive use (for example, a move that could KO around 100% fresh will not even KO at 300% if fully decayed). With the freshness bonus also being introduced (a mechanic that boosts the power of a fresh move by 5%), there's a large disparity between the knockback power of a fresh move and a move that has been frequently used, making the strategy of not using your primary KO move(s) until your opponent is at KO percent actually viable (rather than just being an ideal tactic). It's a double-edged sword though; while this rewards players who plan appropriate use of their KO moves and punishes those who used them haphazardly, it causes a few moves to be overly powerful at comboing into themselves, which then causes the characters with such moves to be very hard counters to those that are especially vulnerable to them (Pikachu's down throw is the most infamous example, being a chain throw that causes massive damage, if not directly leading to an outright zero-death, on several characters because of this).

    First Person Shooter 
  • Borderlands 2
    • You can complete challenges to collect "Badass Ranks", and when enough are collected, you can permanently increase the attributes of every character you've created. However, when it's time to increase a stat, you can only increase one of a set of five (out of fifteen or so possible stats), and stats that you haven't increased are more likely to show up in this set than other stats that have been increased a lot. As well, the size of the increases fall gradually from 0.7% to 0.4%.
    • The other skills for all characters avoid this trope: every skill point gives the same bonus as the other ones. This makes maxxing out fewer skills much more effective than spreading them out, since a 40% boost to one thing is usually more helpful than a 20% boost to two different ones.
  • E.Y.E.: Divine Cybermancy's stats become increasingly expensive (level-up points wise) once they're past level ~30. At level 100, it can take a dozen points to level up one stat. Cybernetics cost more brouzouf with each upgrade - and require the expensive Necrocybermancy research past tier 6 - to compensate for the massive bonuses given by upgrading.
  • Many classes in Team Fortress 2 become less effective when there are more of them. This is especially obvious with Spies and Snipers, as they are precision-elimination classes and have the least pushing capability in a push-based game. Not that this'll stop you from having eight per team, of course. The Soldier is probably the only exception; being the Jack-of-All-Stats, a team of Soldiers is quite feasible at low levels of play.
  • In contrast to TF2, Overwatch actually accounts for this at the hero selection screen - it will warn a player when their team is has too many of a class (especially snipers) and point out roles that may be lacking (such as a healing character). Since it's a push-based game like Team Fortress 2, this can be valuable information for a player who wants to help turn the tide of battle for their team.

    Hack And Slash 
  • Diablo II uses this for many skill bonuses, such as Dodge giving you an 18% chance to dodge with the first point, but quickly tapering down to less than 1% bonus per point by level 20. This type of balance wound up turning many skills into "one point wonders." Just put a single point in the skill, and the "x all skills" bonuses on your equipment end up giving you just as much of a bonus as actually maxing the skill would have in the first place.

    Mecha Game 
  • To encourage certain allied units to be flat out tanking (namely Super Robots) rather than have them dodge-tanking, newer Super Robot Wars installments introduced "Evasion Decay": if any unit, allied or enemy, successfully dodges an attack, the next attack gains a cumulative bonus to its accuracy rate. This resets once the unit takes a hit.

  • Dragon Saga's archer classes have a skill that lets them juggle enemies in the air by holding down the 'Z' key. Predictably it quickly became one of the most hated exploits even after more broken ones were found. One attempt to fix it saw its damage gradually decrease to 30% over consecutive hits... with no other change, meaning that the only 'improvement' was that the victim survived longer in a helpless state.
  • City of Heroes originally let you add as many of a type of 'Enhancements' to a power as you liked, leading to builds that were far and away so much better than anything else that there was no reason to use any different kind of enhancement build. This led, eventually, to The Great Diversification, when Diminishing Returns for Balance was instituted, making every enhancement of the same kind give less return. The returns diminished so harshly that anything over three of a kind was useless. The net effect was to make the game overall more difficult and to weaken linear, straightforward powers that only benefit from one type of enhancement. Certain powers and builds became useless overnight.
    • It was very necessary as the Inventions System that came out a few years later gave many newer ways to enhance powers would have been absolutely gamebreaking if there were no diminishing returns. Inventions also had the side-effect of making many previously unfeasible builds very effective ones that can rival the most powerful of the pre-diversification builds. If anything, Min-Maxing was made even better.
  • Multiple damage or defense upgrades in EVE Online employ Diminishing Returns. The first such module has full effect, the second approximately 80%, and after the third a fourth becomes near-pointless.
    • Something similar happens with skills. Each skill typically gives a 5% increase per level to it's relevant modules, but the training times increase exponentially for each level, so you may need only an hour to train the skill to level 1 and unlock the first 5% boost, but it will take a month to train it to level 5 and unlock the final 5% increase.
  • World of Warcraft uses this for the duration of stuns and similar effects on other players to prevent "stunlocks". Defensive stats also use them to encourage tanks to invest in all forms (if available to them) rather than just one. Hard caps are also in effect in some cases, but Blizzard is trying to get rid of those or change things around so they cant be reached.
    • For instance, there is a hard cap for armor. At some point druid tanks, helped by a generous armor multiplier to compensate leather equipment and lack of shields, could easily push past it. Many healers also had a talent to increase a players armor for a short time after healing them with a critical effect. This was then addressed by several means such as changing the armor multiplier to only work on specific types of items and the healer talent changing to a plain damage reduction effect.
    • Trade skill leveling works in a similar fashion. Early on, it's ridiculously easy to cook spiced bread with inexpensive flour and spice you can get anywhere or make cheap potions with flowers common as dirt but as you increase your skill it becomes harder to increase it further. Ingredients needed to craft become rare and expensive and it often takes multiple attempts to actually gain a skill point, each attempt consuming the reagents regardless of a possible lack of skill gain.
  • Champions Online implemented this feature fairly early on to balance out the downright insane returns people were getting on Dex/Ego builds (things like critting on almost half your attacks for almost double damage). They instituted a "softcap" partially determined by the character's level that seems to be working well as a compromise; low level players see better returns early on (Dex became a feasible defensive stat BEFORE the late teen levels). but the higher level min/maxers can still push their stats high enough to get noticeably better performance from their characters.
  • eRepublik skill training works like this, in the case of strength when you start out you gain .5 every time you train, by the time you hit 4 strength this is down to .04 every time you train.
  • Star Trek Online has a skill point-based system with diminishing returns the more you rank it up. The only reason for maxing out some skill trees is to get the ability to train high-level, class-specific powers on Bridge Officers.
  • Ragnarok Online inverts this before its renewal patch for its offensive stats; while the cost to increase a stat went up as it got higher, you'd get bonus damage when you reach certain points (for example, every 10 str points gave a melee attack bonus). This bonus grows as you reached higher and higher stats, to the point that oftentimes that it was worth paying more for the higher stat (so you'd have less total stats), if it meant reaching a damage bonus. The renewal patch changes this behavior.
  • Final Fantasy XIV has status ailments occur in shorter intervals if you keep applying the same type to the same target over and over again. For example, Sleep will last for 30 seconds, and then 15 seconds if inflicted on the same target again, and it keeps growing shorter from there. Luckily, the diminishing mechanic also applies to enemies so they can't cheaply put you in a stun lock. However, players and enemies can have multiple status ailments of the same kind inflicted on them at once if they were hit by different attackers, so it's possible to have 20 players all inflict the basic Poison status with full strength on a boss character since each Poison effect is coming from different people instead of the same player multiple times.

    Puzzle Game 
  • Used in Puzzle Quest, where putting more and more points into each stat gave less and less reward per point.

    Real Time Strategy 
  • League of Legends uses hard and soft caps for some stats. Attack Speed and Cooldown Reduction hardcap at 2.5 attacks per second and 40% reduction, respectively, with no diminishing returns. Movement Speed has no hard cap, but two "soft caps" make it hard to increase past a certain point-raising your movement speed past one soft cap causes increases to have their effectiveness cut, and going past the second causes a steeper cut.
    • Dot A 2 also has similar hard and soft caps, but a certain character manages to disable the speed cap (which is 522). If an enemy hero has less than 50% of his health, Bloodseeker will get not only a speed boost but will also disable the speed cap, making it possible to have it's speed to go from around 370 to a whooping 1500.
  • Rise of Nations and its spinoff Rise of Legends (no relation) increases the price of a given unit as their population increases, to make it more expensive to mass single unit types.

    Role Playing Game 
  • The Fable games give you three sets of stats and let you choose how to level. Unfortunately, each level of a stat costs about 5 times the cost of the previous level. Since leveling a low stat is so cheap, almost everyone advances roughly equally.
  • The Elder Scrolls III: Morrowind is a treasure trove of this trope:
    • The most direct example is the patentable method The Elder Scrolls has used for leveling for a long time - you are awarded experience for misses and critical failures as opposed to successes. Your level is proportionately related to your ability to use a skill successfully, hence you will be able to learn something you have no skill in quickly, but master something you have a lot of skill in only through devoted use (since you will stop failing as often). There are only a few exceptions, such as some implementations of magic.
    • Alchemy, meanwhile, is broken beyond all belief and is the perfect example of what can happen when this trope is not enforced. Not only can mass herb mixing give you lots of Alchemy skill (snowball effect), it will also give your base stats lots of free level-up points AND eventually absurdly powerful and expensive potions as your skill increases. Lots of money, lots of experience, powerful stackable buffs, all while doing very little. Additionally, this leads to the legendary Fortify Intelligence Stacking trick and can ultimately result in a situation where the game crashes from mathematical overflow.
    • When you buy from trainers, it costs increasing value for each level. (The actual increase in price, however, is typically considered trivial compared to one's ability to make money.)
    • In what could be a direct call-out to the introduction to this trope, the Merchant skill is broken at 50. Up to 50, the price decrease of all merchants' goods makes sense as a better level should indicate you are better at buying and selling. After 50, things start getting more expensive again. Many mods have attempted to correct this issue.
  • The Mario & Luigi RPGs award points into every stat automatically on each level up, and then let the player apply a bonus to one stat by timing a roulette spin. If one stat is getting too high, or you've used the roulette on the same stat too many times, the numbers on the roulette get smaller. There's always a 3 on the wheel, however, so with good timing you can minmax any of your stats.
  • The Might and Magic games had an interesting take on this trope, where the stat system only offered bonuses at certain intervals, which progressively decreased in frequency as the number got higher. You received bonuses to relevant skills, for example, at 13, 15, 17, 19, 21, 25, 30, 35, 40, 50, 75, 100, 125, 150, etc., until finally stopping offering bonuses past 500. So, a stat of 15 and 16, for example, was functionally the same. This tended to make raising most stats higher than around 50 or so rather pointless, and rendered most minor stat gains worthless well before that.
  • The first Geneforge used the Might and Magic version of this trope—going from 10 to 11 in a skill, for instance, does nothing, whereas going from 10 to 12 provides a bonus equivalent to going from 9 to 10. From 10 to 20, the bonuses came from even numbers only, and above 20 bonuses were only gained at 23, 26, and 29. This incidentally meant that getting one point of a skill from a quest could just mean that you'd be capable of getting a bonus if you were willing to buy another level of the skill. Later games replaced this with the Dungeons & Dragons system of having higher levels cost more points—in that case, the only problem is that it encourages minmaxers to get their free quest-related bonuses as late in the game as possible, so as to maximize the number of skill points saved.
  • Stats in Demon's Souls tend to start giving diminishing returns when brought up to somewhere around level 30-50; Endurance is the most extreme, as at level 40 it outright stops giving stamina boosts.
    • Its spiritual successor Dark Souls have a soft cap of 40, but diminishing returns really hits hard at 50, in which the only reason you want 50 in that stat is for spell requirements (such as 50 Intelligence for White Dragon Breath and 50 Faith for Sunlight Spear). The exception is Carrying Capacity, which is tied with Endurance stat; while Endurance stops giving any stamina boost after 40, increasing it beyond 40 gives you one point of carry capacity per level, that is, a flat rate increase.
    • Dark Souls II both plays this straight an inverts it: all stats have diminishing returns and the soft cap is still at 40, but the points from 30-40 are worth considerably more (about 50% greater return per point) than the first thirty. Some advantages of a stat won't even show up at all below a certain value, for instance spells won't start getting extra casts until you get attunement above 20.
  • Fallout, Fallout 2, and Fallout Tactics do this with skill points. Raising a skill to 100 is done on a 1-for-1 basis. Raising it to 101 costs 2 skills points per rank. This quickly accelerates so that, by the time the skill is close to its cap at 300, it costs as many as 6 points per rank. Raising skills so high has its benefits, especially in Fallout 2, where certain very high skills (such as Science and Speech) unlock special gear, quest solutions, and even party members, if you're willing to specialize. Being able to blow someone's head off with a sniper rifle from across the map doesn't hurt, either, and you'll need a Small Guns above 200 to do it.
  • Fallout 4 has Armor rating (Ballistic Defense and Energy Defense) give diminished returns on damage reduction. For example, 60 Armor rating gives 30% damage reduction, but about 93 Armor rating turns into 40% damage reduction. And the difference between 560 Armor rating (T-45 Power Armor) and 1280 Armor rating (T-60 Power Armor) is 81% and 87% damage reduction respectively.
    • It is useful to note, however, that when you are decreasing something by an absolute percentage, every extra percent becomes more powerful. By taking the provided example, when you increase damage reduction from 30% to 40%, you're not increasing the defense effectiveness by 10%. In practice, you've increased your damage reduction by 14.3% (60/70 = 0.857, calculating from the damage you'd receive from a 100 damage attack.) Likewise, increasing damage reduction from 81% to 87% you are effectively increasing your defense by 31.6% (13/19 = 0.684, calculating from a 100 damage attack, again). Look at it this way, if you had 10 000 HP, it'd take 143 attacks at 100 damage each at 30% reduction to kill you, but 167 attacks at the same strength at 40% reduction, and that's about 14.3% more attacks total needed to kill you. Likewise, it'd take 527 such attacks at 81% reduction, while at 87% this number is 770, a 31.6% increase. Another way to look at it is that if you are at 99% reduction and increase it by another single percent, you are increasing your defenses infinitely, as you are now ignoring 100% of all damage. Some games call this "effective health". At 50% reduction, you have 100% extra effective health. At 66%, only a 16% increase in total reduction, you already increased your effective health by 200%. At 75%, your effective health increased by 300%. At 80%, it is increased by 400%. At 83.3%, it is increased by roughly 500%. 600% at 85.7%. 700% at 87.5%. 800% at 88.8%. 900% at 90%. And so on. This is a lot of math to basically say that diminishing returns in damage reduction percentages are warranted, otherwise you'd easily reach invulnerability and later be HEALED by each attack.

    Tabletop Games 
  • The Role-Playing Game Role Master has this for the skill system. Every ten points in a skill up to +50 gives you +5. The next ten give you +2, up to +70. Then you get +1, then +1/2. Magic items and stat bonuses, on the other hand, are linear, so +10 is +10 whether you're adding to 10, 100 or 1000.
  • Dungeons & Dragons (both 3e and 4e) uses a variant of this in their point-buy system for stats - The modifier value derived from your stats remains constant for easy math, but the cost of each stat grows faster the higher you get.
    • This only applies for new characters - points gained from leveling (and from items or spells) have no diminishing returns, although the cost of most magic items grows quadratically with the benefit.
    • Older editions also increased the Wish cost for raising stats: if your stat was anything between 1 and 15, then a Wish spell would improve it by +1, but if your stat was already 16 then each Wish would only raise it by 1/10 of a point; if it was 20 or more, each Wish would increase the stat by 1/20 of a point! This was back in the day when the Wish spell aged the caster by five years, and required a System Shock roll (basically a Fortitude save) to avoid instant death (with success merely leaving you bedridden for a week).
  • Similarly, the New World of Darkness tends to encourage evening out one's portfolio past a certain level — each new dot in a category typically costs its new rating multiplied by a flat rate, so buying the fourth dot in an Attribute from the third costs twice as much as buying the second dot in an Attribute up from the first. Similarly, the character-creation rules make the fifth dot in a Trait (out of a normal maximum of five dots) cost twice as many of starting points to buy.
    • Averted in character creation, where costs are linear, except for the highest level. This leads to the problem where it is mathematically superior to start with an extremely minmaxed character, and spend xp to cheaply flesh out secondary abilities, than make a balanced character and pay a premium to increase your abilities later. Shadowrun 4th Edition has the same setup and suffers from the same effect.
  • Skills in GURPS start to give noticeably diminishing returns around level 14 (90% chance on normal tasks) and additional levels become a total waste of points at level 24 (90% chance on "impossible" tasks) except in the most extreme settings where techniques can have penalties of -30.
    • Earlier editions of GURPS had the point cost of any skill double with each level, so increasing anything beyond +7 or so (depending on skill difficulty) would be ridiculous. Later editions made the cost linear with level, allowing for easy minmaxing with opposed skills (like swordfighting).
  • In some games the basic mechanics are inherently designed to create this. For example, any Tabletop RPG that uses a bell curve (e.g. 3d6) as opposed to a flat statistical probability (e.g. 1d20) for its core mechanic will get a version of diminishing returns: in a bell curve the numbers in the middle of the probability space are created by more of the possible permutations of the dice, while the numbers on the edge of the probability space are created by relatively few permutations. This means that changing a very low or very high number grants a smaller benefit than raising a number in the middle of the range.
  • Swashbucklers of the 7 Skies employs similar ability improvement to the aforementioned The Elder Scrolls video game series: you only gain points to improve your abilities when you fail. Characters who max out an ability and then focus exclusively on it are going to advance very slowly, if at all, while those who dabble in many things or throw themselves into scenarios where they've got no real skill are going to develop faster.
  • 7th Sea has an unusual version of this. Each game session a player starts with a number of hero points equal to their lowest stat. If unused, these hero points become experience points at the end of a session. Depending on your sessions this can increase your net gain somewhere between 50 and 100%. So the character who put two in all his stats will gain experience points faster than the person who has a wit of 1 and a finesse of 3. (and this lets him earn his third points faster, causing the divide to grow even faster)

    Third Person Shooter 
  • Splatoon has this manifest in several ways:
    • Having multiple copies of the same ability on year isn't very helpful, because every duplicate ability will provide less of a boost. For example, having one "Ink Saver" ability gives you a decent boost, adding as second one gives you a much smaller boost, and adding a third is mostly pointless because the boost is so small as to be negligible.
    • Certain weapons can be very effective on their own, but become a liability if the entire team is using them. For example, having a roller on your team is good because it can ink a lot of ground quickly and splat any enemy it runs over. Having a team of nothing but rollers is an uphill battle, because rollers have very little range and very poor vertical mobility. Likewise, having a charger on your team is good because they can splat enemies and provide support from a distance. Having a team full of chargers makes it very difficult to cover ground or combat any enemy that gets close.

    Real Life 
  • This phenomenon is called the bottleneck. It's especially visible in computer setups: if you have an awesome CPU and lots of RAM but a piddly video card, games won't run very well. Conversely, if you have an awesome video card but a weak CPU and little RAM, your games still won't run very well.
    • This is a problem with computer systems as a whole: just because you increased the performance of one part doesn't mean you'll see the same gains. For example, if going from a dual core processor to a quad core processor cut 10 seconds out of a job, adding four more cores may only cut out 5 seconds, while costing 4 times as much.