For some time now I am thinking about adding a ranking system for GSPs to the FPS meter. But I am not sure how to do this right, so I want your input... Some time ago (when only the first version of the FPS meter was published) there was a ranking for single measurements, simply by the QI ("Quality Index"). In those days I have learned several things:
- GSPs will always try to manipulate a ranking in every possible way. Some will try to get to the top with "prepared" servers they never could (economically) sell, others even try a more direct way by sending emails telling that some GSP is "cheating" and it should get excluded.
- If one starts developing technical means detecting/evading this kind of manipulation (i.e. requiring a certain amount of players on the server) the GSPs will quickly adopt (i.e. by putting those players on the server without letting them play).
- The QI alone is not a good quantity to judge game servers (even if some of the obvious flaws e.g. due to the OB update of CSS get fixed). There are a lot of other important things that need to be assessed for getting a complete picture. Especially since the OB update for CSS FPS became more unimportant for most games (for CS 1.6 they are still important).
- Thus some kind of customer feedback would be required, which again could be easily manipulated by GSPs...
So up to now my attitude was better not to do this, as I do not know a solution for these problems. But there are ranking pages in the web, and in my opinion most are misleading. Most of them are more or less marketing platforms for GSPs, the GSPs can pay money to improve their ranking and to be presented more prominent.
So my questions are: Would a new, really neutral ranking system for GSPs be appreciated by the community? Has someone ideas how to do this "right", i.e. how to eliminate/reduce the possibilities for manipulation?
I hope this creates some fruitful discussion ;-) I do not want to criticize GSPs or those existing ranking platforms, IMO is this kind of behaviour quite natural. I am looking for a way to naturally evade this ;-)
I would think the best way of doing this would to take only one server submittion score per client IP(The IP of the person running the test) then lookup who the servers IP belongs to and set that as the GPS name. If they have it setup properly the name that the IP belongs to should be the GSP and if its not they can just request for you to change the name that displays to clients.
Another thing that could also affect the score that the GSP gets is the servers per IP, this can be found by querying the Master List Server.
Also as you said I hope this turns into a discussion it could be interesting.
First thing would be to make a new QI.
But you should take player count into account.
Second I think it should work this way.
1. GSP Register on the site.
2. PPL that use the FTP meter will then be able to set witch server provider they use.
3. You make a average out of it.
That way even if a GSP cheats they would have to have a large amount of servers to manipulate the result a lot.
I think GSPrating.com already does most of these things, and their system on ranking on several key areas for selling points is the best around. They require a customers server IP for user submission, and have a great "social networking" system to connect GSP's. As far as I know, they are 3rd party, and even if they are associated with a particular GSP, they really don't show it.
Yep...
And sad but true: There will be always companys that have contractors which post for them. There are different kinds of, like only positive, let the portal look active and so on and so on...
hm strange, "View new posts" never showed me the replies here... :-)
thanks for your input! I will answer only some points, that doesn't mean I don't appreciate or consider the others ;-)
Nisd Wrote:First thing would be to make a new QI.
But you should take player count into account.
The player count is already now taken into account. Probably for hlds the QI could remain (more or less) unchanged, but for srcds it has to be changed much. If the FPS are 67 or 67000 is no longer any difference. But the amount of jitter in time is still important...
There are probably still some pre-orangebox servers (like HL2:MP or some mods). Do you think they can be neglected? Or do I have to make separate cases (I could probably detect it by the server version)?
Nisd Wrote:Second I think it should work this way.
1. GSP Register on the site.
2. PPL that use the FTP meter will then be able to set witch server provider they use.
3. You make a average out of it.
Yes, that is pretty much my basic idea. But there are still many methods to abuse the system...
Nisd Wrote:That way even if a GSP cheats they would have to have a large amount of servers to manipulate the result a lot.
As loopyman pointed out, not a real problem for large GSPs. Also this works only well if a lot of people are actually contributing. So it's probably best to make a minimum number of different servers required for a GSP to show in the list?
What do you think of the following idea: A server only counts for the rating, if it runs long enough (like 4 weeks) and constantly has players on it. I would use some lower average usage of a typical war server for it. This way it becomes really difficult for a GSP to get multiple servers into the ranking that are not actually customer servers (it would still be possible with a small number of e.g. sponsored servers that are better than what they usually sell, but that's OK then, I guess). The draw back of course is that only long-term measurements would get into the ranking. That will of course decrease the number of measurements, but on the other hand it will improve the statistics and equalize the measurement conditions...
PS: I will probably drop the requirement for donating money to get the "premium" features soon, as the server seems to be funded by other means now...
10-26-2010, 06:21 PM (This post was last modified: 10-26-2010, 06:28 PM by Terrorkarotte.)
(10-26-2010, 05:12 PM)BehaartesEtwas Wrote: What do you think of the following idea: A server only counts for the rating, if it runs long enough (like 4 weeks) and constantly has players on it. I would use some lower average usage of a typical war server for it.
PS: I will probably drop the requirement for donating money to get the "premium" features soon, as the server seems to be funded by other means now...
Are you talking about average playes on it?
If yes: What about adding only the measurement request with >2 real players on. If you have X positive request for one Server only those requests will be used.
Other things:
For obox games high fps mean nothing. A relative stable server with a var <1ms and 100fps will be as good as a server with same var and 10000fps. How are you planning to take that in acount?
If you do not, you will help hosters in sellling high priced servers to the people which they do not need. Also you will help to keep the false impressions about fps alive.
Saying that the qualitiy with obox servers should be only about relative stability (net_graph 4: var) above the tickrate and not about who has the highest fps (or better penis?).
How are you planing to take the source tv issues in count? Even if the stats command displayes bad dropping you won´t noticing a thing on the server. The drops caused by source tv will make the server worse than it is if you only go for as high and stable as possible.
Also server fps mean nothing if you have a bad or not large enough connection. If you want to say something about quality you need to take that in account too.
But how can you measure packet loss, low RTT etc. for servers that are not in germany? You would need servers in all major hosting countries to provide any quality results.
(10-26-2010, 06:21 PM)Terrorkarotte Wrote: Are you talking about average playes on it?
If yes: What about adding only the measurement request with >2 real players on. If you have X positive request for one Server only those requests will be used.
partially, yes. the idea was to use only measurements that run for at least 4 weeks and that have also some player activity. for measurements that run over an hour or so it is not a real problem for GSPs to get them filled with real players. still that doesn't mean that those players are really playing and that the server is of the same type/quality as those really sold by the GSP.
Quote:Other things:
For obox games high fps mean nothing. A relative stable server with a var <1ms and 100fps will be as good as a server with same var and 10000fps. How are you planning to take that in acount?
yes, I pointed this out earlier. the QI needs to be modified to take this into account.
Quote:How are you planing to take the source tv issues in count? Even if the stats command displayes bad dropping you won´t noticing a thing on the server. The drops caused by source tv will make the server worse than it is if you only go for as high and stable as possible.
I am not really convinced this is true. If a server runs with 66 fps and sourctv causes a drop, this drop will be noticeable. you are right in case the server runs with higher fps and drops to something still above 66 fps. the QI probably should simply ignore fps > 66, then this solves automatically :-)
Quote:Also server fps mean nothing if you have a bad or not large enough connection. If you want to say something about quality you need to take that in account too.
correct, but there is no real way to do this from the fps-meter. important is the connection to the clients which are usually connected via DSL. also the distance to the data centre plays a very big role. I am not sure if that can be done properly... at least it would involve an additional tool to be installed on the client computer. but I could imagine people would not want to have a tool (even if it uses very low resources) running all the time in the background when playing...
Quote:But how can you measure packet loss, low RTT etc. for servers that are not in germany? You would need servers in all major hosting countries to provide any quality results.
11-03-2010, 12:19 AM (This post was last modified: 11-03-2010, 12:21 AM by BehaartesEtwas.)
First step: I want to change the graphs, so the orange box update is visually taken into account. would the community accept a y-axis scaling like in the attachment? I changed it so it actually resembles a time scale (of the time that passed between the frames). the top is 0ms, the bottom is 100ms. one horizontal line is 5ms. I still keep the labels in FPS, because the people will probably understand this more easily. what do you think? I expect quite many people shouting at me, because 1000 and 2000 fps no longer make a really visible difference, but that difference isn't a real difference, so... ;-)
look at the left y-axis scale. 66.7 fps what my servers are running at is now quit at the top of the graph, while it used to be much lower (cf. http://www.fpsmeter.org/p,view;1.html). the y-scale changed so that lower fps can more easily distinguished, while high fps look more alike. actually the scale is now a time scale (i.e. the inverse of the fps).
![[Image: 76561197967000252.png]](http://badges.steamprofile.com/profile/extended/steam/76561197967000252.png)
