The Big Flaw of balancing games with 50/50 winrate
Handschuh
Join Date: 2005-03-08 Member: 44338Members, NS2 Playtester, NS2 Community Developer
Join Date: 2005-03-08 Member: 44338Members, NS2 Playtester, NS2 Community Developer
I don't want to deviate the other threads - so I post a new one here.
I think there is a big flaw in ns2 if you try to balance the game with a winrate of 50/50 and think "it's perfect". (Since we're about to have 2 Hiveskills anyway)
From what I've read for months is that somehow "50/50" is being the holy grail for balancing in ns2. That is in my opinion flawed in many ways.
More often than not you notice that marines are stomping the aliens (= NO FUN) and I'm not talking especially about bigger hitboxes (which I dislike) or that good players might rather choose marine.
I'm talking about the consequences of baserushes.
Aliens have a decent amount of winning games when they completely fucked up by their round - just by tunnels / baserushes they still win.
And that is in my opinion fine. Not watching the lanes, etc. should be punished.
But considering that these Games exist and you still try to force a 50/50 Winrate by balancing around it will leads to stomping marines, because if you want that the aliens have a equally high fighting chance against marines - when no baserushes happen - then you need to adjust this value from 50/50 to something more alien-favorable like 55/45. But please don't be fixated on the numbers, since it just shows the direction.
If you take a generic player which has 2k Hiveskill for aliens and marines (because he always wins the same amount on both sides) - and his winrate on aliens is not that his alienplay is somehow smart, but rather "tunnel-smart" then this will lead to the fact that this player is being stomped by marines but still holds his hiveskill because of his tunnel-base-rushes
In my opinion I'd rather have fair games and still loose more often as marine / win more often as alien due to baserushes. So when Baserushes don't actually happen you still have more fair games / fair fights.
I think there is a big flaw in ns2 if you try to balance the game with a winrate of 50/50 and think "it's perfect". (Since we're about to have 2 Hiveskills anyway)
From what I've read for months is that somehow "50/50" is being the holy grail for balancing in ns2. That is in my opinion flawed in many ways.
More often than not you notice that marines are stomping the aliens (= NO FUN) and I'm not talking especially about bigger hitboxes (which I dislike) or that good players might rather choose marine.
I'm talking about the consequences of baserushes.
Aliens have a decent amount of winning games when they completely fucked up by their round - just by tunnels / baserushes they still win.
And that is in my opinion fine. Not watching the lanes, etc. should be punished.
But considering that these Games exist and you still try to force a 50/50 Winrate by balancing around it will leads to stomping marines, because if you want that the aliens have a equally high fighting chance against marines - when no baserushes happen - then you need to adjust this value from 50/50 to something more alien-favorable like 55/45. But please don't be fixated on the numbers, since it just shows the direction.
If you take a generic player which has 2k Hiveskill for aliens and marines (because he always wins the same amount on both sides) - and his winrate on aliens is not that his alienplay is somehow smart, but rather "tunnel-smart" then this will lead to the fact that this player is being stomped by marines but still holds his hiveskill because of his tunnel-base-rushes
In my opinion I'd rather have fair games and still loose more often as marine / win more often as alien due to baserushes. So when Baserushes don't actually happen you still have more fair games / fair fights.
Improving gameplay for aliens would be easily accessable by including more tutorials or giving this info for lowskilled players in the HUD...
- When to use PRes (Lerking/Fading instantly and not wait) [HUD should give you that Info like any other game...]
- Having a good lifeformdistribution
- Biting RT's
- Not Rushing like Lemmings
- Real Explanations how movement in ns2 works
- Laning -> Ambushing
- When to use PRes (Lerking/Fading instantly and not wait) [HUD should give you that Info like any other game...]
- Having a good lifeformdistribution
- Biting RT's
- Not Rushing like Lemmings
- Real Explanations how movement in ns2 works
- Laning -> Ambushing
Comments
But other than this technicality I agree with the overall sentiment.
I still disagree with you because everything you mention can be countered... with skill/gameknowledge/mapawareness including the evil baserushes ( with laning / scanning / having someone rdy with sg in base). The same is with the growing aliens... you could just push to end the game once and for all instead of waiting (but that is what "turtling" - is teaching you
I just think that having "winrate" determine the hiveskill will lead to more stomping marinerounds on average which are no fun. Of course the counter does also exist "sneaky phasegate with hiverush" - but the chance of being noticed as well as the costs are much riskier
As for balancing the game without taking tunnel rushes into account, I agree, a 50/50 balance is not all that leads to a fun game, it does however still corroborate. But by balancing around base rushes, you essentially force aliens (depending on the strength of base rushes) to use this weapon. If they don't use this powerful weapon, they won't win. The issue is though that the marines will be superior to the aliens leading up to the base rush, since they will have been buffed to compensate. This is of course not very fun for an alien player.
All this of course depends on how strong base rushing is, and how many games are won due to a good tunnel. Depending on how many it is, the balance without base rushing will be even more skewed. Personally I don't feel like the problem is that big though (if it even exists), since I feel like not a huge percentage of games are won due to good rushes through tunnels.
I dont think you need to balance around games without baserushes. That is wrong as well since good players will also do baserushes.
I think there should be at least a nudge in favor for the aliens to have more fun in average so the fightingskills of the marines or balancing is somewhat equal. Right now I think the more "baserushers" are on alienside the weaker is the alienteam to not be stomped.
That is why 50/50 is still bad for the aliens. It needs to be a bit higher.
So you're saying marines would lose less if they just played better?
I mean... you're not wrong. Marines could definitely prevent it a lot of the time. But they don't, at least in pub. What I'm describing is something that I actually see happening out there in games.
In regards to the other current threads, shuffle is the principal topic of concern.
- Fades you need armor 2 and Shotguns and you can fend them off
- If you have tons of onossavers this gives Marines good chance because they're only skulks... what you need is a hiverush, maybe use that to tech up even more... if you win because skulks alone can fend off marines for like 20 minutes because they're all saving for onos? Than you should have lost anyway... but there are a lot of aliens who loose because they don't use their PRes.
- Lerks: If there are superaggressive Lerks who fly like in Pearl Harbor - get early Shotguns and sneakkill them
Bad earlygame can be recovered from both sides... marines can get Shotgun to get 1-Shot-Kills for every skulk
But still - this is not the topic - I'm talking here about the gameplay issue 50/50winrate
I used win rate averages in this other thread because it easily conveys the information I want to share to make my point. Rejzor may have lost 8 games in a row, but aliens aren't losing every single game like he is saying.
When it comes to statistical game balance, "Who wins when" graphs are much more useful. They are not as easy to communicate or read.
For example, here are the who wins when graphs for b318 and b319. Keep in mind that this sample has balance mods included. I did not have time to find the mod id's to filter out the balance mods.
B319
Even then, a lot of information is lost.
I would say that it does not determine skill, but the likelihood of an individual to win in a typical team environment. This should correlate to their skill but it is not their skill. Hive is useful for balancing games but it is not a direct measure of skill. Hive is relative to who you play with. Being useful for balancing games is not the same thing as a player's skill. Hive isn't designed to make fun rounds, but to give each team an equal chance of winning.
In terms of this thread, it is what Nous said.
You're arguing about a theoretical, ideal balance, that may be achieved in comp play, but rarely if ever occurs in pub.
I'm talking about working with what's actually there in your typical pub game, and getting shotguns most certainly will not guarantee one shot kills against skulks here. People play a lot less ideal in pub and you need to keep that in mind when balancing.
That's where the aliens-growing-out-of-control issue comes from.
Implied in the other post I wrote in your poll thread is a key idea: views like the one I'm quoting here suggest that the different balance models we've been discussing roughly correspond with different goals / desired outcomes / balance philosophies. For me, that's an indicator that we'd be better off with a selection of officially supported shuffle models rather than trying to construct one that works for the whole community.
I'm a better marine than an alien in practice, although not by much. Still, it's enough of a difference for me to recognize and appreciate. I also prefer playing marines a bit more; go figure.
But I would repeatedly play on a server which used the aggressive balance algorithm even if it meant playing my preferred team more rarely. That's because I value the idea of a game where the outcome is largely unpredictable and in which the teams are as balanced as is statistically possible (subject to hive limitations) far more than I value the experience of playing on my preferred team. I also think it's how both players and communities get better, and how amazing players are produced over time. It's just a better fit for my attitude, especially since I don't have much time for gathers or leagues these days. It's hard to get a balanced game in a pub, so in my eyes the aggressive approach would be invaluable, and I'd probably stick to a server that ran it.
In the post linked above I also described a rough approximation of a team preference system. I'm not saying it's the ideal way to introduce team preferences; it's just what I came up with after reading the thread. But following from that idea and what I've written above: even if the server I just described didn't allow team preference stats to influence shuffle results, I'd still play on it. I'd probably even prefer it, since team preference adjustments can only really walk us slightly further from the absolute statistical balance the aggressive model seeks (even if it's impossible to ever really get it). Maybe I'm weird, but based off of the poll results, a number of people exist in the general ballpark of my sentiment.
That said, I totally understand your feeling. This is a game, not everyone approaches it as if it were the quest to become the ultimate tryhard. And for most people, an increased likelihood of being able to play with friends and/or be on their favorite team means more than anything offered by a statistical model which seeks to maximize "evenness" each round. Something approximate is good enough for them. It's almost certain that a majority of active players match this rough description.
But that difference is precisely why we should formalize these conventions into a small handful of standard options as opposed to wasting time on the debate about which model is best. The models simply seek different outcomes, and of the most popular proposals, each desired outcome has good arguments in its favor.
Example: at the start of the round, a coin toss determines which team is instantly declared the winner. That's perfect 50/50 balance, but obviously no fun whatsoever.
I firmly believe that fun should be the highest priority of any changes, and that balance tweaks can be made without impacting fun.
My interest in 50/50 balance - as an idea and as an attempted reality - extends no further than through considerations of the new shuffle options that moultano's player skill improvements would enable. As in: after using the offset to establish respective marine and alien skill values, I'm then interested in optimizing player distributions per round to minimize inter-team skill differences while also minimizing intra-team standard deviations. Other people are interested in less aggressive uses of the offset. None of this has anything to do with in-game mechanics, and it certainly isn't intended to be anti-fun. It's just a different approach to balance that values a different outcome.
But your caution against presuming that overall game design should relentlessly and blindly seek 50/50 global winrates is otherwise well taken.
It is not just better players, but worse players too. Any players that are significantly distant hive skill from the server average will.
High skill player may play more often and be further from the average, but there are more low skill players by sheer quantity.
To be clear, I'm referring to this option in moultano's poll:
Question: How would you like teams to be balanced with the introduction of per team skill values?
Answer: Fully use them. Race win-rate will always be 50/50 (subject to the limits of hive accuracy). Better players will generally be forced to play on their worse race and vice versa.
In other words, the aggressive algorithm described in the other thread would have two values to work with for each player. The "classic" algorithm would instead work the way current shuffle does. This is captured by the following portion of moultano's blog post:
The skill values this learns without the offset are backwards compatible, so balancing algorithms will not need to be modified. Despite the fact that each player will still only have one skill value for balancing, there is still a benefit to doing this. A player’s skill will no longer randomly walk based on their team assignment. It will always be the average of their marine and alien skill, rather than being effectively a weighted average of their skill for the teams they’ve played on recently. Additionally, by collecting this data, and returning it from hive, we can open it up to modders to balance teams more aggressively if they desire.
The offset is meant to establish the difference in each player's marine vs. alien winrate, and by establishing that difference, we increase confidence that either value accurately reflects a given player's ability on a particular team. The question is: what do we do with that data?
An aggressive sort allows for the calculation the narrowest possible standard deviation / skill average per team. Here, "narrowest" captures the very thing about the sorting algorithm that can be described as offering novel utility, whereas the use of a single skill value for balance partially obfuscates each player's per team skill in order to preserve certain valuable effects, like not forcing players to play on a certain team due to an overly strict approach to minimizing standard deviation while ensuring average skill balance.
In response to people who are arguing to preserve the existing shuffle mode, I have only three points:
1. I agree. There's demonstrable value in the classic shuffle mode, and it is worth preserving.
2. However, there's also demonstrable value in the aggressive shuffle mode (based on both the votes (thus far) in moultano's poll and the personal preferences I described here). I think there's clear utility in it. Clear drawbacks, too. Hence: choice.
3. In addition to advocating for three official shuffle modes, I think there's clear utility in establishing a team preference flag which would allow server operators to take preference data into account in order to impose constraints on any of the proposed shuffle modes, the aggressive mode included.
I understand what you're saying, but I don't know that your chosen example fully captures your sentiment. Typical debate about weapon balance can at least be approximately settled by appealing to global variables, statistical consideration of outcomes in the course of rounds played, etc. Weapons have objective traits that merit direct contextual analysis, but it's much harder to statistically isolate specific variables that make a winning player, hence the need for abstraction in the form of winrate. Skill only represents winrate, and winrate is already an abstraction of what's actually a multidimensional range of skills wherein greater proficiencies in each of those dimensions roughly correlates with an increased likelihood of winning.
The aggressive shuffle model simply accepts the increased resolution offered by two skill values - one for each team - and uses that increased resolution to optimize balance results via brute force. The downside is as just you and Nordic have described: any players that have a significantly distant hive skill from the server average will be subject to team limitations.
There are preference-honoring ways to mitigate these limitations which are agnostic with respect to the particular shuffle mode being used.
But setting those possibilities aside and limiting our consideration to the proposed shuffle modes alone, a better comparison is probably more along the lines of: a given weapon can be more or less effective depending on the specific attributes of the person wielding that weapon; in order to preserve the general fairness of a gunfight, we'll be ensuring that each combatant only has access to weaponry that results in a personal effectiveness rating that trends toward the average effectiveness of all combatants in the gunfight. It's not about correcting the weapon imbalance, or even commenting on it. It's about constraining the imbalance in the player if doing so optimizes expected outcomes for the round.
It's easy to overstate the negative effect this would have for really good players. First off, many if not most great players are pretty great at both teams, after all. Secondly, teams aren't weapons. A weapon is just a tool with specific attributes. But there's a huge range of possibility within a given team. Any player could undoubtedly improve a subset of their aptitudes (per team) in order to influence their team-specific winrates via those improvements.
For example, the intricacies of gorge spit alone are characterized by a skill range between the most incompetent and very best spitters. Has anyone truly hit the ceiling, achieved the perfect spit accuracy with utter reliability? No. The perfect spit eludes us. The same holds true for almost anything in the game: metaskills like gamesense, your positioning, your understanding of the economy and timing, all the way down to the most micromanaged aspects of movement, juking, ambush patterns, and so on. There are always small, iterative improvements to make at virtually every dimension of the game, even when you're amazing at something, and especially when you're not.
But more to the point: I'm of the opinion that if the offset does in fact capture something essential about the difference between a player's skills per team, then games sorted with aggressive shuffle will also provide a useful mechanism for making certain players better at certain roles, even when they're not given much choice in the matter. I also think that this is a Good Thing (TM).
It takes a certain disposition to find the idea of aggressive shuffle appealing, granted. But it's also why I think that the sorting algorithms on the table actually represent distinct ways of considering what's of value in a match, or at least deep attitudinal differences toward what constitutes fun. That's why I resist simplifications which argue that your version of fun is superior to mine. For me it's much less about being on my preferred team; I accept that I often won't be on my preferred team in a game as intrinsically asymmetric as this one. The games that I find most fun, post facto, are those which wound up being challenging and dynamic, in which both teams were roughly composed by collections of players of comparable effectiveness at their roles. There's no perfect formula to produce that outcome in a game like NS2, which is a strong argument for allowing servers to define their own approach via a handful of shuffle modes.
Whether these modes are official or included in shine, each would be making use of the same hive data. And based off of your poll, there's interest in each mode. I'd rather see each mode get use than debate which would be the most fun.
So if official shuffle modes don't happen, the next best thing would be the same result, but via shine. In which case substitute "official shuffle modes" for "shuffle modes that are commonplace thanks to shine, the most popular mod offering shuffle functionality that's currently used on Natural Selection 2 servers".
Hope that's specific enough for you. :P
If the intention of shuffle is to have 50/50 winrate, then consider tweaking it.
Shuffle just tries to make balanced teams afaik.
When you said 50/50 winrate @Handschuh, I was under the impression that you were talking about the balance in gameplay discussions, not the shuffle discussions.
So I don't know what you're talking about anymore.
But I'll just add this; asking for 50/50 CHANCE between factions (ie. by enforcing equal teams), is different from aspiring to a 50/50 winrate.
But since everyone seemed to talk about shuffle it sounded like this "50/50 Chance" leads to "50/50 winrate".
Correct me if I understood it wrongly, but that's how it is (intended?) to work isn't it?
But yeah gameplay balance was originally the focus....
This kinda sounds like the noob tube idea
Easy to use, but severely limiting in it's usage
Gives newbies a chance to kill that godlike fucker who keeps killing them, while not giving it super powers in the hands of that very same fucker
BUT, this is Natural Selection 2, where units cost time and resources. So cheap noob tube kills are not on the agenda as per anti-rage policy
Then you are addressing the problem with the wrong solution.
I very much think that, overall, a 50% winrate should be targeted, though I obv wouldn't care about a 5% deviation from that. But yes, I do agree that only caring about a 50% overall winrate is stupid, as the individual mechanics and strategies need attention, especially if you want the game to be fun. So, the solution should would to address specific problems, not change literally everything around a few strats/mechanics that you don't like.
With that noted, however, although I don't know how comp play goes, the concept of alien base rushes "skewing" the winrate sounds like a pub-related problem, obv caused by a lack of lane-awareness. Perhaps we should discuss ways to get teams/comms to be more aware of vulnerable lanes or something? This would likely change the overall balance, but afterwards we could discuss which other strats/mechanics would need balance-attention.
I think that sounds about right, but it also doesn't capture the full picture.
There's a non-trivial difference between a desperate rush made against the marine base when the aliens are down to a single hive vs. a rush made against the marine base when aliens are on 2 hives, but there often isn't a significant difference between the alien team's aggregate strength in either scenario. This is usually the case when comparing rounds where aliens have 3 harvesters on one hive (the tech point harvester + 2 naturals) vs. when they have 3 harvesters on 2 hives (2 tech point harvesters + an intermediate natural). Marines are obviously dominant in the former scenario. But marines can still be very dominant in the latter scenario assuming that they're encroaching from one of the alien naturals and have map control otherwise. It's not uncommon; aliens often find it trivially easy to secure a second hive (which makes sense; preventing this outright requires massive marine aggression from the start of the round) even in the cases where doing so does not significantly alter the alien team's prospects to win.
In other words, the frustration caused by a sneaky tunnel rush probably has less to do with the alien team's relative strength prior to the rush, and more to do with the marine team's strength prior to the rush, how that strength is represented on the field, and how that strength is still undercut by a fundamental vulnerability that often feels cheap and unfun when exploited.
Marine teams just aren't incentivized by the current tech progression to aggressively pursue a second base even though they're strongly incentivized to aggressively contain alien expansion. (An aside: I'll be the first to admit that this has changed somewhat with the current exo meta that's dominating pub play, since marines can more easily do just about everything once exos are on the field. That's something that I hope will revert somewhat when changes to exos are made. Increasing the cost to 55 would be a great start.)
But to really capitalize on a second base as such, a marine team needs a command chair, an observatory, and at least one infantry portal. That's a non-trivial res investment for what amounts to an insurance policy, and it's an insurance policy that's most relevant during the midgame when marine armor, weapon damage, and arsenal upgrades are crucial to compete with lifeforms. That's why you most often see second marine bases pop up after the aliens have attempted and failed a base rush on marine main; it serves as a wake-up call to the marine team, assuming they recover.
Preventing every single tunnel rush against a dedicated alien team requires really coordinated lane control, multiple static exos on the field (lame), a very cautious commander willing to scan repeatedly, or intelligent marine players who check naturals / patrol lanes or zones. Oftentimes, those same intelligent marines would be more optimally used on the front lines. Nevertheless, this is non-optional because the marine base has multiple points of fragility (destroy the CC outright = win, destroy the power node = probable win, destroy the obs + main phase = probable win) and it takes only one successful tunnel rush to win. It's a constant psychological 'check' played against the marine team unless they commit to that insurance policy.
Taking nothing away from the coordination required to pull off a successful tunnel rush, it is nonetheless the case that the frustration factor stems from the marine team's status prior to the rush moreso than the alien team's status prior to the rush. Having been a part of a number of sneaky tunnel rushes, I'll admit that my enthusiasm for my team's success usually amounts to little more than a "meh" when I reflect on the fact that we only succeeded because the game design encourages a chronic and reliable weakness in marine defenses.
Why? cause i think they made things right and there system reflected real skill and not only a winrate.
After a 6 month break im really shocked that the only skill of some players is still: placing tunnels near the base.
Thats what they doing, every round, till 6 month and more: Placing tunnels near base.
And thats how ppl learn howto play aliens these days: "We lost the early round like always and need a tunnel near base". Im pretty sure, many ppl believe thats the only way to win as aliens. *facepalm*