Yeah I'll stick with the guys who have access to the complete set of metric, thanks.
Why doesn't Frontier invest in dedicated servers?
- Thread starter magnaram
- Start date
Yeah I'll stick with the guys who have access to the complete set of metric, thanks.
Aight. I do PvP a lot, I play OPEN only and never heard about those surveys you're refering to...
But I appreciate you provided the sources on which you might feel appropriate to base your estimations on. The major issue with those surveys is though (other than I havent heard about them
), we have no idea of how many people are playing the game overall, so it's hard to tell how representive they are. But that variable only Frontier posess and not willing to share.If I can ask you something, I'd like you to start with the source supporting your claims next time
Technical requirements and cost.
Always makes me laugh when people talk about wings or fleets of 20-30 ships... it rarely happens most of us only play in small groups anyway. It wouldn't add anything massive to the game nor would it change the game. Other things like space legs, atmos planets and the stuff we have in Beyond Chapter 4 from next week are all things players would do or use on a daily basis more than big battles or big fleets in one area.
I've visited PvP systems or CGs many times and seen loads of players in and around stations anyway. I've only had instancing problems with one friend and that was down to the quality of his Virgin Media connection... seemed to be fixed opening ports and changing configs. All my other friends instancing and grouping has always been fine.
Though I reckon Frontier should pull CQC into the main game as a holo match from stations that could have a bigger player limit.
Great, here is what they said:
"While it's true that the PvP crowd do tend to be more vocal and in previous betas have given more organised feedback, we're well aware that the majority of players don't get involved in PvP."
https://forums.frontier.co.uk/showt...ents-UPDATED?p=4808539&viewfull=1#post4808539
Dev said quite specifically that Open was the largest SINGLE mode - with the inference it made up less than 50% of the total (so Open>Solo Open>Private Group but Open <Solo + Private Group)
F
Aight. I do PvP a lot, I play OPEN only and never heard about those surveys you're refering to...
But I appreciate you provided the sources on which you might feel appropriate to base your estimations on. The major issue with those surveys is though (other than I havent heard about them
If I can ask you something, I'd like you to start with the source supporting your claims next time
It's not just the survey, Frontier agrees with me as well...
i rather tend to see partial reveals of information by closed sources as an indicator of their specific agenda more than of the whole picture. but i can understand someone taking it as sacred scripture. specially if it fits the bias.
Yeah I'm well aware of that, and don't dispute it. Although I haven't mentioned PvP at all in this thread, so not sure why you're bringing it up with me.
Give me a shout when someone's got some actual metrics from FDev on mode use.
Mr. Sandro Sammarco only said the majority of players play OPEN. I've seen the stream myself. The rest is your interpretation.
By <50% of the playerbase who filled in the survey.
As there was no attempt made to get a random statistically-valid sample, you can't extrapolate from that to the playerbase with any reliability.
In the UK professional polling companies - who make every effort they can to get as close to a true random sample as possible, apply advanced statistical techniques to correct for likely deviations from true randomness, carefully choose question wording and order to avoid introducing bias there, etc. - tend to get no better than a +/-5% margin of error on their answers on the rare occasions their polls can be tested versus reality. (Which means, even had this been a poll conducted by one of them, that "49% Open" could really be anywhere between 44% and 54%, which could be entirely consistent with Frontier's "most play in Open" statements anyway)
Guess what the margin of error is on sticking up an internet poll somewhere and letting anyone who stumbles across it fill it in?
Well... we don't have to guess. We can just look at some of the other survey answers:
"12% Emergent Gameplay Deliverer" ... maybe if you survey a CG supercruise instance in Open? But given that half the players in the survey said that they didn't play in Open, that would meant that about a quarter of the population of Open are going round attacking peaceful traders. Really? You think that's anywhere near reality? The peaceful traders and explorers I see in Open certainly don't behave as if about a quarter of the people they meet are going to try to kill them!
"39% of the respondents said they'd been playing more than 2 years" ... at the time of the survey (August 2017), the game had only been out for about 4 years. And it wasn't a "quick starting" game where it's hyped to the hills and everyone buys it on release day (certainly it wasn't on PC). Linking up with the "primary platform" question you can conclude that a majority of the PC-platform players completing the survey have been around >2 years. PS4, which makes up 22% of the respondents, having only been out a few months at that point ... and XBox not having been in at the beginning either, there's very little room for any PC players to have bought Elite Dangerous in 2017! I think Frontier might have been rather more worried and rather less profitable if their PC sales in 2017 had actually been "almost none".
"10% said that they always participated in Community Goals". Now this, we have exact numbers on to compare with. In and around August 2017, the least popular CG [1] was the Sigma Summit with 3411 [2] participants, so even if that one was quiet because absolutely none of the "Sometimes / Occasionally" people took part ... that implies a total player base of only 34,110 players. Er, no.
"Over 800 respondents (4%+3% = 7%) said they played CQC at least once a week". Yeah. I remember trying to get a CQC match in August 2017 and it often involved a lot of communication between the regulars on Discord to be on at the same time. If 7% of the actual player base played CQC at least once a week? It'd have been permanently busy!
(Frankly I can't even believe that there were even the 800 people who completed the survey playing at least one match a week back then - I'd have seen more of them! - never mind extrapolating up to the actual player population. But I'm sure no-one would lie on an anonymous internet survey, or fill it in more than once, or misinterpret a question, or whatever...)
I mean, you can believe that despite the obvious breaks with reality on the other questions, the errors happened to all cancel out and the Modes question was spot on. You can also with equal validity believe that 95% play in Open, or 95% play in Solo. The survey doesn't really help decide between the two.
(One thing you can actually do to sample the Modes question is sit in supercruise in Open, and compare the traffic implied by the station traffic reports to the people you actually see. For the systems I visit I get "more than half of players are in PC Open" doing that comparison, but there's no reason that the systems I'm in are representative of players as a whole ... and actually quite a few reasons why they're probably not!)
[1] I've ignored Colonia CGs for this.
[2] This also hints at how useful a "stale" survey might be - 3,500 is now a pretty good total for weekly CG participation in the most popular CG types. CG participation has fallen by about 75% over the last year ... but other indicators of total player activity have remained relatively stable. So even if the survey was accurate back in 2017 - which it clearly wasn't - it certainly won't be now. This applies equally to any non-recent FDev statements about the composition of the player base, of course.
Open is the most popular mode out of three. If you combine Solo and PG it's a different story though.
This is all pretty irrelevant. The claim was that most people PvP. It's not true.
That's the post our discussion is based on:
Regardless if my posts, sources, claims, etc. are accurate or not, it's pretty clear that most people don't PvP. If you want to believe that or not is another story.
Last edited:
It’s aprox. 45% Open, 35% Solo and 20% PG.
That is play time, not players.
Most players probably use all three modes.
.
It's not the question about capacity or performance. It's about effort and responsbility.
.
If in 10 or 15 years FD doesn't operate the BGS any more, how much effort will they still invest to give us a a working implementation of the BGS? It's much easier and cheaper to disconnect it from the game and let it run on "default values".
.
In contrast, if they go ahead and give us the necessary files to set up a working BGS, it also won't be the same. In this scenario you'll end up with many people having their own BGS running, some of them manipulated for their own purposes. (Not necessarily bad purposes, mind you. A lot of them might just bend and manipulate it to match their own roleplay and story telling. )
.
So you won't have one galaxy and one BGS any more, but many different variants of the galaxy. Which isn't bad, but definitely not the same.
.
.
Uh... man. I can't raise my eyebrows as far as would be appropriate. They'd hit the back of my neck...
.
Let's first cover the last part: you are right that it goes from client to client. But the clients are not "nodes". It's always direct connections from client to client. Routing issues are the very same as for other internet connections. So would what you describe actually be true, it would only be single clients which have issues.
.
Unfortunately that's not really true for ED. We know that the game actually set up "hosts" internally, although they use different nomenclature. We don't know all the details there, but according to all we know by playing around (PGs, firewall settings, etc. ) the rough outline looks like this:
- Missions still come from FDs centralized mission server. That's actually client/server and nothing else.
- All BGS information also is provided by FD. It again, when seen on the most basic level, is client/server. Mind you, there is much more about it, as it's actually a distributed structure and quite complex, but it follows the client/server logic.
- Login data and player files also come from FD. See BGS.
.
So up to now, we're still far purely in client/server land. Now we go for the P2P section:.
- The game seems to evaluate clients, their computer and their connection. When several people of the same region are in the same system, the one with the "best rating" is declared to be the "master", where the rest are "slaves" to.
- The information about which system will be "master" or "slave" again still comes from FDs infrastructure and thus is in client/server country. But it manages the P2P part of the game.
- NPCs are created and managed by the "master". Only if a "slave" looses connection, it starts managing its NPCs itself again.
- Every client still is responsible for the own ship movement position and damage management. They communicate the results among each other, but manage it individually.
.
And that's why I say that client/server really wouldn't be that much of a change on a basic level. One client already now always is the "master". Which means it already now in several (but not all) aspects already now does the job of a server. You mention features like invite, accept, reject, boot, etc. which indeed are not present. But they are features, not defining criteria for a client/server infrastructure.
.
What's really missing would be:
- Movement plausibility checks and correction on master/server side.
- All damage management done on master/server side.
- The server being disconnected from being a game client and running standalone.
- A complete rework of the distribution logic running on FDs servers.
.
It's the last one, which would be the one big thing. It would have to turn the complete logic around. Instead of distributing based on location (both in game and in RL) and then trying to match according to secondary criteria, it would have to manage server instances, assign them the in-game location and then distribute clients accordingly.
.
I mean, similar things have already happened, even without being designed into, to other P2P systems.
.
Somebody here remembers the classical honeypots in Emule and the likes? That were just nodes set up at advantageous conditions and modified to best performance. This gave them so high performance and thus so advantageous rating that they for a while operated as de-facto servers. They were run by intellectual rights owners and the collected data allowed them to du a lot of suing and cashing in.
.
For FD the effort would, as already explained, be significally higher. But generally speaking, a conversion of P2P to client/server can be done with acceptable effort, while trying to convert a client/server system to P2P usually requires a completely new implementation.
.
Of course, all of this is pure theory. All I wrote is oversimplified and none would be optimized yet. There's a boatload of extra work to be done to get it to run smooth. I am fully aware of that. While ED is a good example of a "P2P" structure, which actually could be switched to a client/server model with comparatively limited effort, it would still be a lot of effort which are better invested elsewhere. And that's before talking about costs and who would pay for it.
.
.
As you emphasize this part, I will only comment on this.
.
First of all, the mere fact that less than 50% of all players gave feedback does not matter. Generally a 5% sample or even smaller sample sizes are already considered sufficient. So the bold part can easily be ignored and discarded. The last sentence I quoted, on the other hand, is very true and important. We don't know which part of the player base did vote.
.
By all we know it could be that all the players who voted were former players, who just found the poll and wanted to create a mess, while no active player ever voted. But based on the fact that the voting was not announced to players by E-Mail, but was announced on Reddit, I dare to say that it was mostly seen and answered by people actively interested in the game.
.
Also note that generally only 5% to 10% of a games population ever visits the games forums, let alone Reddit. That means that the poll participation was an order of magnitude (!) higher than was to be expected. The most obvious explanation for that is word of mouth. Some players who were active spotted the voting and pointed it out to others. This resulted in more players voting. (Of course, that's just a theory. If you know of any other mechanic, which would explain why so many more players voted than was to be expected, please point it out. )
.
As long as word of mouth is the best explanation, my question would be if word of mouth might have a different range, based on play style and preference. So among which people do you think that word of mouth works best and has the biggest range? Options are:
.
- Organized player groups who fly in open.
- People who use private groups to fly with a handful of friends.
- People who only play in solo and never see another player in the game.
.
Once you managed to answer that, i think we can also assume that the voting results might be biased in favour of that group.
.
Cost, most likely...
But P2P performances leave to be desired. I get disappearing wingman syndrome in a wing of two... and the other player is living within 20km...
Any other MP game ? No issues ever.
But P2P performances leave to be desired. I get disappearing wingman syndrome in a wing of two... and the other player is living within 20km...
Any other MP game ? No issues ever.
Only if the sample is random. You can have as large a sample as you like that's not random and it won't matter.
In 1936, the Literary Digest infamously sent surveys to around 10 million people, and around 2.5 million of them responded (around 6% of the total voters). Their prediction was a landslide victory for one Presidential candidate with 57% of the vote and ... who went on to win just 2 states with only 36.5% of the national vote.
Their survey had well-studied statistical flaws - which led to the closure of their publication. But even they took more steps to ensure the quality of the sample than this survey did.
You set up a nice artifical trilemma there, but there are way more types of players than that. I fly almost entirely in Open ... but I'm not really associated with any organised player group. Mobius is a Private Group but a huge one rather than "a handful of friends". Half the online people on my in-game friendslist are currently playing in Solo, but I must have met them at some point! Certainly of those three groups, the Open-only organised groups are most likely to have seen the survey. Now, what about all the other types of players you didn't mention?
The problem is that word of mouth has far more potential biases than that. If the survey got promoted heavily on PvP Discords, it would significantly overestimate PvP. If it got promoted heavily to Exploration groups ... it would significantly overestimate Solo (just because they don't instance with others doesn't make them unsociable - even single player games have communities!) and Exploration. Where did the survey actually get promoted? I have no idea - I didn't see it until after it was over.
Let's be clear: if Frontier hadn't actually *said* that more than half the player base play in Open, I would have been very surprised by any claim that they did. Since they did say that... it's probably true but there is endless room for debate over whether they meant half the players (by most-played mode), half the time played, half of the players online at particular sampling point(s), or half of something else entirely. (It's very unlikely to mean "half the players visiting the current CG system", for example, which is the metric people will actually see in-game ... no-one cares what mode some explorer 40,000 LY away with no other players for ten jumps is in ... but it's probably Solo for the high-res screenshots)
The survey is junk either way, though, and sheds no real light on the matter.
That's not a point in favour of the survey having any accuracy. You've admitted that the only people who will have seen it will be extremely atypical players.
How large do you think the (at least semi-active) game population is, by the way? You either have a very different definition of "order of magnitude" to me (I would say it means ~10x) or you think the player base is considerably smaller than I do.
[1] There is endless room for debate over whether they meant half the players (by most-played mode), half the time played, half of the players online at particular sampling point(s), or half of something else en.
.
Hmm. What you wrote reads as if you want to correct me. Although, if you really read, it's what I wrote.
.
.
Sure I simplified. But it shows the problem. And again, I am confused. You basically say what I also wrote: it's impossible to draw any real conclusion from the poll. And still you make it sound like I was wrong.
.
.
And confusing once again. It is x10, yes. The bold part was about <50%. Of course it's not quantified how much less than 50% it is. But 50%, at least in my book, is an order of magnitude bigger than the 5% I mentioned. (Although of course it's not one order of magnitude more for the more optimistic number of 10% of all players being on the forums. But that's not what I'd expect to be the case here. )
.
I must have misread you somewhere then. Apologies for the confusion.
Ah. I think I maybe see it. When I said "By <50% of the player base who filled in the survey" I meant "of those members of the player base who filled in the survey, less than 50% of those players said X" and you seem to have thought I was saying "less than half the player base filled in the survey" - which is of course true, but nothing to do with the points I was making originally.
So your reply then made no sense to me and it seems my reply has made even less sense to you.
We'll have to agree to agree, though neither of us is sure on what
.
Ah. Oki. I associated the <50% differently than you meant it. Now I get it. Thanks a lot for the clarification.
.
And yes, we can agree to agree.
.
TheMegaTurnip
Banned
Don't even try OP.
Most on here will try to blame either your internet or indeed claim FDev's P2P networking is fine.
You're barking up the wrong tree.
The game needs dedicated servers more than anything else I agree. If they can do it for the mission boards, they can do it for instancing. It really is that simple.
Most on here will try to blame either your internet or indeed claim FDev's P2P networking is fine.
You're barking up the wrong tree.
The game needs dedicated servers more than anything else I agree. If they can do it for the mission boards, they can do it for instancing. It really is that simple.