According to NS2Stats.com, the situation is the other way around; the marines have won most public games and the aliens have won most competitive. Why does UWE promote NS2Stats if it's then apparently so inaccurate compared to the official stats? How can a website that saves the actual statistics of every game on so many servers be inaccurate though?
NS2Stats has a smaller sample size. Although it's definitely a great tool and an awesome community project, the stats are less reliable because they're only taken from servers which run at least one mod (NS2Stats). This one doesn't affect balance, but other mods the servers might run could. Beside that, there could be a skill gap between the players that prefer to visit servers with NS2Stats and the players on the rest of the servers. I can imagine that better, more K/D oriented players would rather play on servers with NS2Stats enabled, so they can flaunt their stats every once in a while.
Either way i think they should take NS2Stats into account since that seems to be more accurate to the actual situation than the stats that Hugh presented in this article. The marines are definitely winning more than the aliens in public now, and i've seen no indication of an 80% win-rate for the marines during competitive games. It's way closer to 50-50 there from the dozen matches i've seen with the teams i follow.
Just remember that you are only one person that has played an X amount of NS2 games. Your personal experience is minimal compared to the total number of games played and might not be representative of the average.
Think of this, if you flip a coin there's a statistical probability of 50% that it will land on either side (presuming the toss is accurate and the coin is symmetrical). Yet, it's entirely possible that in a 100 throws, it always lands heads-up. That's statistics, probability only describes the average chance something will happen, it doesn't mean it has to.
Am I the only one who enjoyed the period in b238 and b239 where marines had a much lower win rate? (I mainly play marine) The challenge was so much more fun.
Only if you choose to look at it that way.
I looked at it like this:
If we lose, its because the game isn't balanced.
If we win, it must be because the opposing team is awful.
In such a scenario, what meaning does any outcome have?
This is especially depressing for a frequent comm.
Either way i think they should take NS2Stats into account since that seems to be more accurate to the actual situation than the stats that Hugh presented in this article.
So you're saying, the tens of thousands more games played on public than NS2Stats games using the UWE official stats count...are inaccurate?
It's a great achievement that you can get a game such as this to near enough 50/50 in pub play. Other competitive games like cs rely on 100% mirror matchup. SC2 relies on units that do and take consistent damage and somehow NS mixes the two genres.
I can definitely see that it's easier to dodge skulks since Gorgeous. The similar win rates for 240 and 241 kinda shows how many people actually use or benefit from wall-jump since that was the only change between builds.
Either way i think they should take NS2Stats into account since that seems to be more accurate to the actual situation than the stats that Hugh presented in this article.
So you're saying, the tens of thousands more games played on public than NS2Stats games using the UWE official stats count...are inaccurate?
We're gonna have to make a new rule now. Every time Hugh mentions a "parameter," he has to explicitly define it so that certain people will understand.
For clarification, a parameter is the data from an ENTIRE POPULATION. This means that UWE has ALL the data from EVERY game played in Natural Selection 2.
In contrast, a statistic is data from a sample of a population. How accurate this statistic is depends on the sample size and sampling methods. NS2stats is a perfect example of this, and a lot of bickering on these forums result in questions about the sample size and sampling methods. (Specific servers, playercounts, competitive vs pub, etc.)
Obviously by definition a parameter is a gazillion times better than a statistic.
For clarification, a parameter is the data from an ENTIRE POPULATION. This means that UWE has ALL the data from EVERY game played in Natural Selection 2.
No its not, its how they measure distance in the special olympics.
I think he missed all the writing after the first line or two.
I did read it, and I just read it again now, that writing doesn't seem to say that to me, but maybe it's just not being clear. If it is being balanced on the small details then great.
This gives me an idea for a new novelty account OnlyReadsFirstLine.
Yes, this parameter-statistic distinction is ill founded is this context.
It's like flipping a coin once and saying that what you measured is the coin bias parameter, because somehow you have the data from the entire coin flip population. The thing is you need to flip the coin an infinite number of time to measure the parameter exactly.
Otherwise these two statements are somewhat contradictory:
For example, all games played on Descent may yield higher marine win probabilities. Or all competitive games might have lower marine win rates. Games played in Europe may have higher alien win rates than those played in Asia. Ad infinitum.
This is the ‘headline ratio’ that is utterly crucial to lower-level game balance. It is also crucial to the user experience: Over time, the user experience will tend towards the global win probability, and their innate sense of their chances as a particular race will tend towards equality between each.
It's easy to understand why an European player will play on local servers and will not experience the global win probability but its local and possibly biased one.
I also agree with ScardyBob that the "impossible theoretical winrate" theory doesn't make a lot of sense.
I don't think that complaints have anything to do with balance really, unless I hear it from the pros.
A lot of this bluster is really about knowledge of the game and how to play it, and to do this you have to know the game inside and out. How even the seemingly small decisions can effect the whole game.
Really don't think you're able to determine how balanced the game is from taking just the wins/losses. From my experience playing on the pub servers, you get 1 or 2 good players that will massively determine the outcome of the game depending on whichever team they choose, and majority of the time they will alternate between going alien then marine etc meaning that games are obviously going to be around the 50/50 mark.
As CheezyPeteza mentioned, you need to look at each individual aspect of the game. That being said I don't think balance is too far off at the moment, ARCs need taking a look at along with a team having numerous Exo suits.
Reading this description of how you balance the game really disappointed me.
I have never written a description of how to balance Natural Selection 2 and don't pretend to have the knowledge or skills to do so. The post you are referring to is a discussion of two important elements of balance: The probability of victory in a randomly sampled game, and the quality of experience that comes from all the little variables that feed into the balance whole.
It is small details you need to look at to balance the game while keeping the overall figure in the back of your mind.
You wrote a post telling me my blog post was not good, and that it disappointed you. Then, you concluded by restating in drastically simplified terms the same argument I make in my blog post. Is that not hypocritical and somewhat mean to me? Remember the bit about forum posts damaging developer morale? Do you think your post could be an example of that phenomenon?
What you didn't show in the blogpost is that Wins for Marines jumps way up in percentage at 10+ minutes.
This is true. Why is it relevant to the blogpost at all? I also did not show: Probability of win changes based on time of match, region of play, time of day, number of players on server, tech paths chosen, and a myriad of other factors.
Did you have a specific problem with the variable I chose? Why is your variable a better example than mine? Why should the blogpost use a specific variable that you like, and not one that the writer finds is easiest to express the argument he is trying to make? If I had used your variable, would other people have stopped suggesting alternative variables?
Does 'knowing' something is wrong with the game preclude collecting information about the game?
For something as fundamental as spawn rate I would say yes.
Ok. To make sure we're clear: Because some individuals 'know' what is wrong with the game, Unknown Worlds should turn off Sponitor completely until that 'known' thing is changed. The corollary of course is to ignore all forum, social media and anecdotal feedback until that 'known' thing is fixed.
I would like the input of others on this point. Who else agrees that Unknown Worlds should stop collecting all other information about Natural Selection 2 until one issue that some individuals 'know' is wrong with the game is fixed?
Why does UWE promote NS2Stats if it's then apparently so inaccurate compared to the official stats?
NS2Stats.org is perfectly accurate, you just have to be able to interpret a statistic. This means understanding the sample size, confidence interval, confidence level, and other factors. Such a discussion is beyond the scope of this reply but you can find a detailed analysis of the confidence intervals of NS2Stats.org in my reply to another thread, here: http://forums.unknownworlds.com/discussion/comment/2093948/#Comment_2093948
[Could you then talk about what actually goes deep in the game. All the talk I see is very general level stuff.
If I had more time in a day, I would. You see talk about general stuff because Sponitor is not public (unfortunately) and it takes time to query, graph, and write posts like the monster we are discussing.
Fair enough. Apparently they do sometimes post the overall winrates too. Nevertheless, if you ever read David Kim's posts, I'm fairly sure you're aware that they do often separate their information depending on leagues and such and discuss those much more in detail.
What the purpose of this statement? You told me Blizzard does not post winrates. When it was pointed out they do post winrates, you tell me Blizzard posts their winrates, and their detailed winrates! Ok, and? What do you expect me to say? "Yes, they do!"
You kind of scratch the surface of quality of gameplay and then end up drifting back towards the 50/50, finishing the post with a conclusion that you recovered the slip after the release 50-50 (which was heavily 'balanced' by the onos imbalances).
Either my or your English teacher will be sad: Either my expression was poor or you are choosing to draw conclusions that aren't in my writing. Never in this blogpost was any attempt made to discuss any specific balance gameplay aspects. What on Earth does the Onos have to do with my post. I never mentioned the Onos. Or any other lifeform. Your fixation on my optimistic and factual final paragraph perplexes me greatly.
I think the overall wins are fine to aknowledge, but they shouldn't be 50+ % of each balance post and debate.
You appear to agree with my blogpost. This confuses me, because you have spent much time and energy in this thread writing posts that tell me I'm wrong. You, like @CheesyPeteza, say that there are a myriad of variables that affect balance. You, like @CheesyPeteza, conclude that those myriad of variables are important. You, like @CheesyPeteza, therefore spend quite some time agreeing with me, while telling me I'm wrong.
But you, like @CheesyPeteza, seem to have a strong aversion to the possible value of the probability of victory in a randomly sampled game being 0.5.
Would this not be more useful, and more pleasant for this forum: Instead of telling me that there are a myriad of balance affecting variables in Natural Selection 2 (i.e. restating what is in the original blogpost): Why should the probability of victory in a randomly sampled game should not be 0.5? This opens up many further questions for you to consider:
Why should Unknown Worlds only discuss aspects of balance that are not going well?
Why can Unknown Worlds not point out a positive aspect of balance (probability of victory in a randomly sampled game)?
I guess what I'm asking for is you to go past the overall winrates quickly and go much, much deeper in the quality of gameplay section - that's where the magic happens.
You can ask, but you won't get it. You have spent time writing forum posts telling me my blog post doesn't fit what you want in a balance discussion from Unknown Worlds. I discussed first principles, you wanted reams of balance data and analysis. I gave one specific example of a balance variable, you didn't like it and wanted others. Have you considered accepting that this blog post is not what you want, that the value in it is not where you think value is, and that other people may find value in examining the first-principles and definitions of balance, and what Unknown Worlds wants to do about it?
__________________
As you can all see, I'm being thorough about arguing why this discussion is valuable and refuting suggestions that because it is not thorough discussion of individual balance variables, it is worthless. I probably have the energy to do this once or twice more in this thread.
with all due respect, hugh, i understand that you have to put out pieces that reflect the game state in a positive and improving light. however, for some of us, we do want to know that you are technically aware and working on addressing those deeper aspects of balance that may not be quite so good. we don't want to get the impression that uwe is going to try to hit that magic 50/50 and then call it a day. that's all i think the posters are grumbling about. while the posts may sound like complaining, i think they more likely represent an interest in the game's long-term health.
[quote="Strayan (NS2HD);2096409"] [quote="Res;2096151"]What you didn't show in the blogpost is that Wins for Marines jumps way up in percentage at 10+ minutes.[/quote]
This is true. Why is it relevant to the blogpost at all? I also did not show: Probability of win changes based on time of match, region of play, time of day, number of players on server, tech paths chosen, and a myriad of other factors.
Did you have a specific problem with the variable I chose? Why is your variable a better example than mine? Why should the blogpost use a specific variable that you like, and not one that the writer finds is easiest to express the argument he is trying to make? If I had used your variable, would other people have stopped suggesting alternative variables?
.[/quote]
In the blogpost you said:
[quote]Conclusion: Since build 240, Natural Selection 2 balance has improved significantly and the game is fundamentally balanced between alien and marine.[/quote]
There are certain variables that affect probability of win percentrage much more so than other variables. Time is one of them.
You can't say a game is fundamentally balanced when Marines have a much greater chance at winning as a game goes on, forcing Aliens to try and end a game early.
You do seem to acknowledge part of this fact:
[quote]This headline figure is a blunt instrument that does not account for subsets of play.....[/quote]
Given the above though, you still can't say the game is fundamentally balanced as a result.
I think there has been some communication problem here. As I and I think others have read the post from yourself and come away with the impression you are using a 50/50 win ratio to balance the game, but you and others are saying in the comments that this is not what you are saying. So clearly we have misunderstood you.
A customer told me today that he is going to start a software unappreciation society for one of our products as it is so bad, I chuckled... NS2 is one of the good ones, don't let these things get you down, people love the game that's why they are so passionate about every little thing and moan all the time. It means you are doing something right and should be happy. If they didn't care they'd say nothing and play another game.
Win rate per build image explains the older builds I may need to get some more games in with the new builds. 25-35 min games seem to be the most fun and the map changes after on most servers so it keeps people playing. Also game understanding (players knowing maps, guns, and ability not so much skill) has a lot to do with balance too, along with how they are distributed between the 2 teams.
Is it easy to check what the win/loss ratio is for more specific parts of the game? I am rather curious as to what the win/loss % is for each time in your "match time histogram" for 241 specifically. We all too often see global win/loss numbers when bringing up balance topics, while that is important and interesting I would be really interested to see exactly at what parts of the game does the tide start swinging 1 way or another.
Conclusion: Since build 240, Natural Selection 2 balance has improved significantly and the game is fundamentally balanced between alien and marine.
This. The average person isn't going to care if a parameter isn't a statistic or whatever (saying that over and over doesn't change anything) or if the randomly sampled probability of something something is 0.5. No offense but while the game is enjoyable (somewhat mitigated recently by skulk nerfs =/) I for one certainly wouldn't call it "fundamentally balanced", neither do I believe this 48/52 winrate statistic parameter indicates at all that it is.
The conclusion given in the blog post based on the data presented is just incorrect according to many peoples opinions. What you are saying doesn't match up with what a lot of your customers are seeing in game. We know that a detailed look at the data will show the game isn't balanced nearly as well as this makes it out to be, and you know that too.
It feels like you took the conclusion you wanted "NS2 is balanced" and then picked out an easy statistic parameter (damn it) to show that conclusion. It's all backwards!
Once the details are out regarding where the game is and isn't balanced then we can clearly see who is right and who is wrong. Why not make that a priority since balance is so important, rather than hiding it under the rug and parading the least useful parameter (!) in terms of balance that exists.
As the move from 239 – 240 shows, subtle changes can have big positive effects.
Some of those "subtle changes" seemed like breaking a leg (or four of them) just to make a pain in your arm go away. Maybe they improved on one parameter, but some of them sure have big negative effects on the fun factor too.
For clarification, a parameter is the data from an ENTIRE POPULATION. This means that UWE has ALL the data from EVERY game played in Natural Selection 2.
In contrast, a statistic is data from a sample of a population. How accurate this statistic is depends on the sample size and sampling methods. NS2stats is a perfect example of this, and a lot of bickering on these forums result in questions about the sample size and sampling methods. (Specific servers, playercounts, competitive vs pub, etc.)
Obviously by definition a parameter is a gazillion times better than a statistic.
That's not correct, a parameter is a number that allow to change the shape of the probability distribution of a random variable, like the mean of a Gaussian or the width of a uniform distribution.
In the case of NS2 you've got a discrete probability distribution with two states, 0 and 1 (marine loss, marine win) with probabilities (1-p) and p respectively. The distribution is characterized by this single real parameter p, that you want to estimate. This parameter is determined the game internal variables and mechanics, but has nothing to do with the sample size. Ideally it should be equal to 0.5.
In order to estimate this parameter you use a parameter estimator. What you get from NS2 stats or UWE internal system is a binary sequence of win loss (e.g. 010101000111100) of length N (the number of games played), a good estimator of p is just the sum of this binary sequence divided by N. You can show that it's an unbiased estimator, i.e. the mean of the estimator equals p, and you can compute its standard deviation (the error on your estimate) in function of N. The standard deviation of the estimator goes down like square root of N, and so only in the limit of infinite games recorded you can compute exactly the parameter p. Having the all population or only a random subset doesn't change anything to this picture (it just decreases N).
Yes, this parameter-statistic distinction is ill founded is this context.
It's like flipping a coin once and saying that what you measured is the coin bias parameter, because somehow you have the data from the entire coin flip population. The thing is you need to flip the coin an infinite number of time to measure the parameter exactly.
Yeah, this bothered me a bit too, but the distinction is more an issue of semantics rather than concept. If your talking about the state of NS2 as of today, then the Sponitor winrates are a parameter. However, if your interested in using Sponitor to predict future winrates, then its a statistic of a population that includes an infinite number of NS2 matches.
I would argue that b241 represents the best balance NS2 has ever had. There are still improvements to be made, of course, but the devs are definitely moving in the right direction.
I personally prefer 30+ minute games to shorter 10-15 minute matches. I don't see match length as a problem in this game. If anything I've been having more long matches in 241 than any previous build, and I'm loving the fact that matches take a while to conclude themselves.
I guess what I'm asking for is you to go past the overall winrates quickly and go much, much deeper in the quality of gameplay section - that's where the magic happens.
You can ask, but you won't get it. You have spent time writing forum posts telling me my blog post doesn't fit what you want in a balance discussion from Unknown Worlds. I discussed first principles, you wanted reams of balance data and analysis. I gave one specific example of a balance variable, you didn't like it and wanted others. Have you considered accepting that this blog post is not what you want, that the value in it is not where you think value is, and that other people may find value in examining the first-principles and definitions of balance, and what Unknown Worlds wants to do about it?
I guess this is what I was looking for. It's not my call on what goes to your posting, so it's obviously your call.
I think part of the reason why I keep pestering so much about these things is because there has been a great deal of uneasiness about how UWE understands the NS and how NS2 is going to develop on the parts that aren't quite finalized yet. That kind of uneasiness easily drives away who might otherwise contribute and commit to building the community. I have no idea whether any of those uncertain contributors are still around, but I believe the deeper and more convincing you are with your posts, the better the chances are that you have very capable and committed people who help to build NS2 into something more meaningful and longer lasting than your average AAA title.
That is all. Hopefully most people will find the balance report sufficient.
Wasn't the skulk just fixed? I know the speed had changed to be bad, but that was a bug which got fixed. I'd say the better hit detection/hitbox against skulks is just better, since before people in pub servers could pretty much run straight at a marine and get a kill, but now playing a skulk really feels more natural imo, all that ambushing and shift-walking on the ceilings trying to sneak up on rines really seems like it was supposed to be that way.
Well, ofc I wasn't happy at it at first since I had also gotten used to the old harder to hit skulk, but now I've learned some more better strategies and do almost exactly as good as I did before, and playing marines is much more enjoyable as it was before (atleast for people like me having only a decent fps, only slightly noticeable ingame).
Wasn't the skulk just fixed? I know the speed had changed to be bad, but that was a bug which got fixed.
The momentum bug on landing was fixed but skulk acceleration was nerfed by a little over 50%. That huge nerf and the hitbox change was how "balance" was achieved.
Yes, this parameter-statistic distinction is ill founded is this context.
It's like flipping a coin once and saying that what you measured is the coin bias parameter, because somehow you have the data from the entire coin flip population. The thing is you need to flip the coin an infinite number of time to measure the parameter exactly.
Yeah, this bothered me a bit too, but the distinction is more an issue of semantics rather than concept. If your talking about the state of NS2 as of today, then the Sponitor winrates are a parameter. However, if your interested in using Sponitor to predict future winrates, then its a statistic of a population that includes an infinite number of NS2 matches.
Looks like a deep conceptual mistake to me, I agree it doesn't really affects the conclusions in this case, but it makes me cringe a little.
Very vaild concerns. Long games can be fun... But we have to 'balance' (Geddit?) the fun long games provide with the effects they can have on the broader game. A median game time of around 10-15 minutes and mean of 20 minutes would probably be ideal to mitigate a variety of issues long games cause.
Count me amongst those who love long games! We just have to all be calm and recognise the issues with too many of them.
@Strayan: I absolutely understand that to much of something good devalues it. But one thing bothers me: (Those epic 45-60 minute games aside!)
You wrote in the blog post, that even 24-36 minute games are to long for modern games. Leaving us with say, 15-20 minute games that are desired. Is that right? Are you really sure about that? (Warning, anecdotal evidence following!)
Because nearly all the 15-20 minute games I remember, are absolutely skill-stacked games that were decided after 5 minutes but dragged on to 15-20 minutes. Before they really ended. The good games were both teams struggle to get the edge and are even in player skill, are nearly always at least 25-30 minutes long. Sure, you also see the good executed rush ending a game at 18 minutes while both teams were relatively even the rest of the time. But this are the exceptions. In most of this good, fairly balanced games both teams can't end it so fast and need until late game, before the game is decided.
Btw. games that end after 6 minutes (mostly by a rush) aren't even this of a problem here, because I haven't invested much time into that round and it feels like that the enemy deserves the win for the good executed rush.
To conclude what I'm afraid of: While I agree, that you shouldn't increase the appearance of those epic 45-60 minute games, you also shouldn't increase the appearance of those 15-20 minute games, because most of them aren't really fun. In my experience they are mostly lopsided and were decided much early but dragged out to the 15-20 minute mark.
I think you have the data from the Sponitor to check, if I'm right with my experience. Maybe look at how one-sided the extractor-amount is in such games compared to the longer ones. How the K/D-difference is. Or other markers that can indicate if those were desirable games. Or if it were games in which you can tell after 5 minutes what team will win.
I will close with a thank you, for the very good blog post and a "Good job!" to UWE for achieving this kind of balance with 241.
Lets say you put 1 marine vs 1 skulk - The skulk would probably win 50% of the time. Now lets put 7 skulks vs 7 marines - The marines would probably win around 70% of the time. The reason for this has been discussed on these forums before so I won't go into why this is the case, but the main point to take away here is that aliens are unable to close the distance on larger servers because they die long before they ever get into melee range.
@invTempest: This sounds logical. But it could also be balanced by the spawn-rate. If the spawn-rate of eggs, alien respawn-rate and marine spawn-rate would be dynamically coupled to the amount of players in a team, you could compensate the problem you mentioned: "melee-range and numbers" and the problem of "aliens don't need to build, making them overpowered in small games" too.
I will just make up some numbers that will probably not work, but just to demonstrate the mechanic:
alien eggspawn per second (per hive) = 16 - alien player count
marine respawn time (per IP) = 2 + marine player count
With the right numbers and / or formulas, this dynamic respawn-time system could at least increase the balance on 12-24 player servers. Heck, it could even be extended to be a better auto-team-balance system as we have now!
with all due respect, hugh, i understand that you have to put out pieces that reflect the game state in a positive and improving light. however, for some of us, we do want to know that you are technically aware and working on addressing those deeper aspects of balance that may not be quite so good. we don't want to get the impression that uwe is going to try to hit that magic 50/50 and then call it a day. that's all i think the posters are grumbling about. while the posts may sound like complaining, i think they more likely represent an interest in the game's long-term health.
Don't worry, we're not "calling it a day" just because we've achieved a pretty good global win/loss ratio. We will continue to improve and update the game.
While some may consider looking at wins and losses a very simplistic way to examine balance, the reality is that when you exclude everything else, there can only be two outcomes in a game like this. Either you win or you lose. To what extent, with what method, and in what amount of time is secondary.
You have to start somewhere, and this is the lowest common denominator. Once you can balance overall wins and losses, then you can start moving further up the tree. Now you can look at wins/losses in games of certain player sizes, or of certain time lengths and possibly make adjustments. However, changes can't upset the 'trunk of the tree' or else you end up with a broken tree. Then you have to start all over again from the bottom.
Frankly, that UWE actually created spark monitor to capture all this data deserves a freaking medal. I've seen many titles where they use anecdotal evidence to make balance decisions, or they just drop one post-launch balance patch and abandon the title to work on the next project.
Comments
You probably play on a certain selection of servers. Perhaps modded. That can significantly skew the stats one way or another.
NS2Stats has a smaller sample size. Although it's definitely a great tool and an awesome community project, the stats are less reliable because they're only taken from servers which run at least one mod (NS2Stats). This one doesn't affect balance, but other mods the servers might run could. Beside that, there could be a skill gap between the players that prefer to visit servers with NS2Stats and the players on the rest of the servers. I can imagine that better, more K/D oriented players would rather play on servers with NS2Stats enabled, so they can flaunt their stats every once in a while.
Just remember that you are only one person that has played an X amount of NS2 games. Your personal experience is minimal compared to the total number of games played and might not be representative of the average.
Think of this, if you flip a coin there's a statistical probability of 50% that it will land on either side (presuming the toss is accurate and the coin is symmetrical). Yet, it's entirely possible that in a 100 throws, it always lands heads-up. That's statistics, probability only describes the average chance something will happen, it doesn't mean it has to.
Only if you choose to look at it that way.
I looked at it like this:
If we lose, its because the game isn't balanced.
If we win, it must be because the opposing team is awful.
In such a scenario, what meaning does any outcome have?
This is especially depressing for a frequent comm.
So you're saying, the tens of thousands more games played on public than NS2Stats games using the UWE official stats count...are inaccurate?
I can definitely see that it's easier to dodge skulks since Gorgeous. The similar win rates for 240 and 241 kinda shows how many people actually use or benefit from wall-jump since that was the only change between builds.
We're gonna have to make a new rule now. Every time Hugh mentions a "parameter," he has to explicitly define it so that certain people will understand.
For clarification, a parameter is the data from an ENTIRE POPULATION. This means that UWE has ALL the data from EVERY game played in Natural Selection 2.
In contrast, a statistic is data from a sample of a population. How accurate this statistic is depends on the sample size and sampling methods. NS2stats is a perfect example of this, and a lot of bickering on these forums result in questions about the sample size and sampling methods. (Specific servers, playercounts, competitive vs pub, etc.)
Obviously by definition a parameter is a gazillion times better than a statistic.
No its not, its how they measure distance in the special olympics.
I did read it, and I just read it again now, that writing doesn't seem to say that to me, but maybe it's just not being clear. If it is being balanced on the small details then great.
This gives me an idea for a new novelty account OnlyReadsFirstLine.
It's like flipping a coin once and saying that what you measured is the coin bias parameter, because somehow you have the data from the entire coin flip population. The thing is you need to flip the coin an infinite number of time to measure the parameter exactly.
Otherwise these two statements are somewhat contradictory:
It's easy to understand why an European player will play on local servers and will not experience the global win probability but its local and possibly biased one.
I also agree with ScardyBob that the "impossible theoretical winrate" theory doesn't make a lot of sense.
Nice article otherwise.
A lot of this bluster is really about knowledge of the game and how to play it, and to do this you have to know the game inside and out. How even the seemingly small decisions can effect the whole game.
As CheezyPeteza mentioned, you need to look at each individual aspect of the game. That being said I don't think balance is too far off at the moment, ARCs need taking a look at along with a team having numerous Exo suits.
I have never written a description of how to balance Natural Selection 2 and don't pretend to have the knowledge or skills to do so. The post you are referring to is a discussion of two important elements of balance: The probability of victory in a randomly sampled game, and the quality of experience that comes from all the little variables that feed into the balance whole.
You wrote a post telling me my blog post was not good, and that it disappointed you. Then, you concluded by restating in drastically simplified terms the same argument I make in my blog post. Is that not hypocritical and somewhat mean to me? Remember the bit about forum posts damaging developer morale? Do you think your post could be an example of that phenomenon?
This is true. Why is it relevant to the blogpost at all? I also did not show: Probability of win changes based on time of match, region of play, time of day, number of players on server, tech paths chosen, and a myriad of other factors.
Did you have a specific problem with the variable I chose? Why is your variable a better example than mine? Why should the blogpost use a specific variable that you like, and not one that the writer finds is easiest to express the argument he is trying to make? If I had used your variable, would other people have stopped suggesting alternative variables?
Ok. To make sure we're clear: Because some individuals 'know' what is wrong with the game, Unknown Worlds should turn off Sponitor completely until that 'known' thing is changed. The corollary of course is to ignore all forum, social media and anecdotal feedback until that 'known' thing is fixed.
I would like the input of others on this point. Who else agrees that Unknown Worlds should stop collecting all other information about Natural Selection 2 until one issue that some individuals 'know' is wrong with the game is fixed?
There are no statistics in this article, only parameters.
NS2Stats.org records statistics, not parameters.
NS2Stats.org is perfectly accurate, you just have to be able to interpret a statistic. This means understanding the sample size, confidence interval, confidence level, and other factors. Such a discussion is beyond the scope of this reply but you can find a detailed analysis of the confidence intervals of NS2Stats.org in my reply to another thread, here: http://forums.unknownworlds.com/discussion/comment/2093948/#Comment_2093948
If I had more time in a day, I would. You see talk about general stuff because Sponitor is not public (unfortunately) and it takes time to query, graph, and write posts like the monster we are discussing.
What the purpose of this statement? You told me Blizzard does not post winrates. When it was pointed out they do post winrates, you tell me Blizzard posts their winrates, and their detailed winrates! Ok, and? What do you expect me to say? "Yes, they do!"
Either my or your English teacher will be sad: Either my expression was poor or you are choosing to draw conclusions that aren't in my writing. Never in this blogpost was any attempt made to discuss any specific balance gameplay aspects. What on Earth does the Onos have to do with my post. I never mentioned the Onos. Or any other lifeform. Your fixation on my optimistic and factual final paragraph perplexes me greatly.
You appear to agree with my blogpost. This confuses me, because you have spent much time and energy in this thread writing posts that tell me I'm wrong. You, like @CheesyPeteza, say that there are a myriad of variables that affect balance. You, like @CheesyPeteza, conclude that those myriad of variables are important. You, like @CheesyPeteza, therefore spend quite some time agreeing with me, while telling me I'm wrong.
But you, like @CheesyPeteza, seem to have a strong aversion to the possible value of the probability of victory in a randomly sampled game being 0.5.
Would this not be more useful, and more pleasant for this forum: Instead of telling me that there are a myriad of balance affecting variables in Natural Selection 2 (i.e. restating what is in the original blogpost): Why should the probability of victory in a randomly sampled game should not be 0.5? This opens up many further questions for you to consider:
Why should Unknown Worlds only discuss aspects of balance that are not going well?
Why can Unknown Worlds not point out a positive aspect of balance (probability of victory in a randomly sampled game)?
You can ask, but you won't get it. You have spent time writing forum posts telling me my blog post doesn't fit what you want in a balance discussion from Unknown Worlds. I discussed first principles, you wanted reams of balance data and analysis. I gave one specific example of a balance variable, you didn't like it and wanted others. Have you considered accepting that this blog post is not what you want, that the value in it is not where you think value is, and that other people may find value in examining the first-principles and definitions of balance, and what Unknown Worlds wants to do about it?
__________________
As you can all see, I'm being thorough about arguing why this discussion is valuable and refuting suggestions that because it is not thorough discussion of individual balance variables, it is worthless. I probably have the energy to do this once or twice more in this thread.
This is true. Why is it relevant to the blogpost at all? I also did not show: Probability of win changes based on time of match, region of play, time of day, number of players on server, tech paths chosen, and a myriad of other factors.
Did you have a specific problem with the variable I chose? Why is your variable a better example than mine? Why should the blogpost use a specific variable that you like, and not one that the writer finds is easiest to express the argument he is trying to make? If I had used your variable, would other people have stopped suggesting alternative variables?
.[/quote]
In the blogpost you said:
[quote]Conclusion: Since build 240, Natural Selection 2 balance has improved significantly and the game is fundamentally balanced between alien and marine.[/quote]
There are certain variables that affect probability of win percentrage much more so than other variables. Time is one of them.
You can't say a game is fundamentally balanced when Marines have a much greater chance at winning as a game goes on, forcing Aliens to try and end a game early.
You do seem to acknowledge part of this fact:
[quote]This headline figure is a blunt instrument that does not account for subsets of play.....[/quote]
Given the above though, you still can't say the game is fundamentally balanced as a result.
I think there has been some communication problem here. As I and I think others have read the post from yourself and come away with the impression you are using a 50/50 win ratio to balance the game, but you and others are saying in the comments that this is not what you are saying. So clearly we have misunderstood you.
A customer told me today that he is going to start a software unappreciation society for one of our products as it is so bad, I chuckled... NS2 is one of the good ones, don't let these things get you down, people love the game that's why they are so passionate about every little thing and moan all the time. It means you are doing something right and should be happy. If they didn't care they'd say nothing and play another game.
Is it easy to check what the win/loss ratio is for more specific parts of the game? I am rather curious as to what the win/loss % is for each time in your "match time histogram" for 241 specifically. We all too often see global win/loss numbers when bringing up balance topics, while that is important and interesting I would be really interested to see exactly at what parts of the game does the tide start swinging 1 way or another.
This. The average person isn't going to care if a parameter isn't a statistic or whatever (saying that over and over doesn't change anything) or if the randomly sampled probability of something something is 0.5. No offense but while the game is enjoyable (somewhat mitigated recently by skulk nerfs =/) I for one certainly wouldn't call it "fundamentally balanced", neither do I believe this 48/52 winrate statistic parameter indicates at all that it is.
The conclusion given in the blog post based on the data presented is just incorrect according to many peoples opinions. What you are saying doesn't match up with what a lot of your customers are seeing in game. We know that a detailed look at the data will show the game isn't balanced nearly as well as this makes it out to be, and you know that too.
It feels like you took the conclusion you wanted "NS2 is balanced" and then picked out an easy statistic parameter (damn it) to show that conclusion. It's all backwards!
Once the details are out regarding where the game is and isn't balanced then we can clearly see who is right and who is wrong. Why not make that a priority since balance is so important, rather than hiding it under the rug and parading the least useful parameter (!) in terms of balance that exists.
Some of those "subtle changes" seemed like breaking a leg (or four of them) just to make a pain in your arm go away. Maybe they improved on one parameter, but some of them sure have big negative effects on the fun factor too.
the BUS!?
That's not correct, a parameter is a number that allow to change the shape of the probability distribution of a random variable, like the mean of a Gaussian or the width of a uniform distribution.
In the case of NS2 you've got a discrete probability distribution with two states, 0 and 1 (marine loss, marine win) with probabilities (1-p) and p respectively. The distribution is characterized by this single real parameter p, that you want to estimate. This parameter is determined the game internal variables and mechanics, but has nothing to do with the sample size. Ideally it should be equal to 0.5.
In order to estimate this parameter you use a parameter estimator. What you get from NS2 stats or UWE internal system is a binary sequence of win loss (e.g. 010101000111100) of length N (the number of games played), a good estimator of p is just the sum of this binary sequence divided by N. You can show that it's an unbiased estimator, i.e. the mean of the estimator equals p, and you can compute its standard deviation (the error on your estimate) in function of N. The standard deviation of the estimator goes down like square root of N, and so only in the limit of infinite games recorded you can compute exactly the parameter p. Having the all population or only a random subset doesn't change anything to this picture (it just decreases N).
I personally prefer 30+ minute games to shorter 10-15 minute matches. I don't see match length as a problem in this game. If anything I've been having more long matches in 241 than any previous build, and I'm loving the fact that matches take a while to conclude themselves.
I guess this is what I was looking for. It's not my call on what goes to your posting, so it's obviously your call.
I think part of the reason why I keep pestering so much about these things is because there has been a great deal of uneasiness about how UWE understands the NS and how NS2 is going to develop on the parts that aren't quite finalized yet. That kind of uneasiness easily drives away who might otherwise contribute and commit to building the community. I have no idea whether any of those uncertain contributors are still around, but I believe the deeper and more convincing you are with your posts, the better the chances are that you have very capable and committed people who help to build NS2 into something more meaningful and longer lasting than your average AAA title.
That is all. Hopefully most people will find the balance report sufficient.
Wasn't the skulk just fixed? I know the speed had changed to be bad, but that was a bug which got fixed. I'd say the better hit detection/hitbox against skulks is just better, since before people in pub servers could pretty much run straight at a marine and get a kill, but now playing a skulk really feels more natural imo, all that ambushing and shift-walking on the ceilings trying to sneak up on rines really seems like it was supposed to be that way.
Well, ofc I wasn't happy at it at first since I had also gotten used to the old harder to hit skulk, but now I've learned some more better strategies and do almost exactly as good as I did before, and playing marines is much more enjoyable as it was before (atleast for people like me having only a decent fps, only slightly noticeable ingame).
The momentum bug on landing was fixed but skulk acceleration was nerfed by a little over 50%. That huge nerf and the hitbox change was how "balance" was achieved.
Looks like a deep conceptual mistake to me, I agree it doesn't really affects the conclusions in this case, but it makes me cringe a little.
@Strayan: I absolutely understand that to much of something good devalues it. But one thing bothers me:
(Those epic 45-60 minute games aside!)
You wrote in the blog post, that even 24-36 minute games are to long for modern games. Leaving us with say, 15-20 minute games that are desired. Is that right? Are you really sure about that? (Warning, anecdotal evidence following!)
Because nearly all the 15-20 minute games I remember, are absolutely skill-stacked games that were decided after 5 minutes but dragged on to 15-20 minutes. Before they really ended. The good games were both teams struggle to get the edge and are even in player skill, are nearly always at least 25-30 minutes long. Sure, you also see the good executed rush ending a game at 18 minutes while both teams were relatively even the rest of the time. But this are the exceptions. In most of this good, fairly balanced games both teams can't end it so fast and need until late game, before the game is decided.
Btw. games that end after 6 minutes (mostly by a rush) aren't even this of a problem here, because I haven't invested much time into that round and it feels like that the enemy deserves the win for the good executed rush.
To conclude what I'm afraid of: While I agree, that you shouldn't increase the appearance of those epic 45-60 minute games, you also shouldn't increase the appearance of those 15-20 minute games, because most of them aren't really fun. In my experience they are mostly lopsided and were decided much early but dragged out to the 15-20 minute mark.
I think you have the data from the Sponitor to check, if I'm right with my experience. Maybe look at how one-sided the extractor-amount is in such games compared to the longer ones. How the K/D-difference is. Or other markers that can indicate if those were desirable games. Or if it were games in which you can tell after 5 minutes what team will win.
I will close with a thank you, for the very good blog post and a "Good job!" to UWE for achieving this kind of balance with 241.
@invTempest: This sounds logical. But it could also be balanced by the spawn-rate. If the spawn-rate of eggs, alien respawn-rate and marine spawn-rate would be dynamically coupled to the amount of players in a team, you could compensate the problem you mentioned: "melee-range and numbers" and the problem of "aliens don't need to build, making them overpowered in small games" too.
I will just make up some numbers that will probably not work, but just to demonstrate the mechanic:
With the right numbers and / or formulas, this dynamic respawn-time system could at least increase the balance on 12-24 player servers. Heck, it could even be extended to be a better auto-team-balance system as we have now!
Don't worry, we're not "calling it a day" just because we've achieved a pretty good global win/loss ratio. We will continue to improve and update the game.
You have to start somewhere, and this is the lowest common denominator. Once you can balance overall wins and losses, then you can start moving further up the tree. Now you can look at wins/losses in games of certain player sizes, or of certain time lengths and possibly make adjustments. However, changes can't upset the 'trunk of the tree' or else you end up with a broken tree. Then you have to start all over again from the bottom.
Frankly, that UWE actually created spark monitor to capture all this data deserves a freaking medal. I've seen many titles where they use anecdotal evidence to make balance decisions, or they just drop one post-launch balance patch and abandon the title to work on the next project.