Proposal for a player trust system

That's a great idea and I'll support it and sign it anyday

There's only one way I know to abuse this system: The Griever marks you as "fully trusted" so you become target to all the jerks he knows. But it's such a small deal that I wouldn't worry too much.
The good answer could be that if you distrust anyone, even if they trust you fully, their positive trust-web won't cover you (they will see you as trusted but their friends won't). Making you basically *impervious* to any form of abuse from those you don't trust (except distrust withing their web, which is good enough).

Rating only inside of an instance would mean that there are people that are abusive enough that I would never group with them, that I met in cities, that I would need to group with to rate them. Big no-no. What about scammers and trust sellers?

Another thing: I'd like one more level "totally unthrusted" - meaning the person is a paramount of idiotry - that allows me to "reverse see" his trust levels. Sometimes one is a small jerk (say -50) and I just want to remember he's on "probation" in my eyes, sometimes I define one as a poor soul to steer clear from, and I'd like to steer clear of his friends too.

I'm not a big fan of decaying positive trust either: there are people I've been playing with in many games that I would instantly mark as trusted - I *know* them. I would rather not have to re-rate them every so long to avoid my trust to decay. I like the decaying negative trust tho: it's a good idea.

I will weep in joy the day I see this implemented
A/Calimar

Ah, but "fully trusted" only goes one-way, meaning you only import trust information from people you trust, you do not export any of yours to them (unless they also trust you fully).

This has been mentioned by IndyCC as well. I'm not quite sure about it, because this may lead to griefing: The griefer simply puts trust in people he hates, and then makes a fool out of himself - Blammo, decent people just earned negative trust. Kinda not good, eh?

Indeed, in order for people to not having to re-rate people they like, I rectified the proposal to not have decay for positive trust a while ago. Though, some case can be made for decay of positive trust (see above) I'm sure people will be annoyed by the need to re-rate.

Edit: Oh, and by the way, if you see any way to exploit the system, please also explain how the exploit works. It's a bit hard to improve on the suggestion otherwise

A nice idea and it appears to be self healing. Even if a griefer decided to be trust worthy for a short while and then go ugly to "break" trust webs... you can still have created your own trust relationship to others therefor he just drops himself from the web.

My real question is from the tech side of things. How expensive will this be on server resources? If every person has their unique value of trust to another person... and that is calculated based on this web... aren't we talking about a exponential calculation? Its very possible for my char to have an unlimited number of entries. Every person that has me in their web will have a unique value for my char's trust. A global rating (which is very susceptable to abuse) gets around this since each char has just one value and that can be easily retrieved and stored.

Sure my comp classes were a while ago in college, but while your solution seems to be very elegant I'm not sure if its pratical in application.

It's obvious the OP put some real time and effort into this idea, and for that I applaud him. (or her)
However, ask yourself "What does this proposed system add to game value versus what it may inadvertently further muck it up?"
The answer of course is little to the prior and quite a lot to the latter.
It adds a level of complexity that is not warranted, and introduces an entirely new method to scam or otherwise negatively impact the playing population.

Regards,

Talesin

One side note that occured to me, but I wanted to separate from the above...

It would make sense (to me) to automatically assign "fully trusted" relationships to anyone in your guild. This would remove the need for new guild members having to set up the relationships (and maintain if we decay trust). After all, if you don't trust your guildmates... why are you in the guild?

This would also jumpstart trust webs as a huge chunk of players are already in guilds but if this trust web were implemented they'd have to create possibly separate groups.. It would also enable a guild to guild trust to be quickly created and validated.

I understand in terms of expense or complexity to impliment it is probably lengthy and could be mucky. Effectively it could replace the friends list and instead simply have a web of trusts and a web of untrusts rather than friends and blocked. This way you get the expanded friends list people are asking for and you dont have to always type in players names into the friends list (since people are also asking for drag and drop interface). I dont know if this is what the OP intended, but it certainly is an idea extension.

I think it adds a fun value to the game because you can more likely get groups you are more suited to and thus have more fun, so I think it adds a lot more than you give it credit for.

I really dont understand your scam coment, please clarify based on the arguments already posted that this system really cant be abused.

I like this idea. Only thing is that it assumes that a mass of people are competant and logical to feel out guilds before they join them by teaming with them or whatever instead of joining a guild that advertises in the local chanel of a random town.

well, if folks are just joining for a cool cape... then they'll be trusting folks like themselves. Others that care about being in a good group/guild will obviously be more careful. Heck if I joined up with a guild that was trusting everyone... it would be pretty obvious when the entire city of Ascalon suddenly showed up with "fully trusted" auras!

My first thought when I looked at the topic title was, "No way, it'd be abused all over the place." But I actually read his idea, and I've got to say I like it an awful lot. I don't think I'd make it an "aura", but rather place a trust indicator on the screen when in towns, to show your level of trust for whatever character you currently have selected. The other side of this though, is what do you trust that person for? Do you trust him to not drop in the middle of a mission, or do you trust that he won't try to sell items at ridiculously inflated prices or offer absurd trades to people who don't know any better? I think that the best idea would be to extend the friends/ignored menu. Add space to put comments on people, and an option to display friends of friends all XFire style. Still, I like your idea, and if it happens to be something that get's implemented, I wouldn't hesitate to use it.

I do that too, staying from people with un-RPGish names.

Expanding on that idea, it would be useful to have, in conjunction with a beefed up team recruitment feature, a place where you can enter a description about your character. Often times you can tell a lot about a person by his/her prose--the rationale behind college admission essay. People who can't write are probably not terribly intelligent.

bobrath raised the question of how expensive this would be on the servers. After some short cosideration I suggest this calculation:

The cost of calculating of the distributed web of trust is dependent on how many steps one wish to remove trust from the person influenced by it.

My original suggestion was:


Steps Contribution
1 ------ 100%
2 ------ 50%
3 ------ 25%
4 ------ 12.5%
5 ------ 6.25%
etc.


Let the number of steps removed a considered trust link is be P.
Let the number of players with registered accounts be N.

In the worst case scenario every person have rated every other person in the game to "fully trusted". This gives us the computational complexity (being linear to the number of contacts reached) of:

O(N^(P))

(Brute force, there might be better algorithms )

However, this is a highly unlikely case - It is more likely that every player has a select number of fully trusted contacts, averaging at say 20. Also, a reasonable number of steps to propagate trust is 4. In this case, the number of reached contacts in the general case would be somewhere in the range of 20^4 = 160 000.

Hmmm, 160 000. That is a sizeable portion of the player base (certainly several percent) and that is good, but it also presents a very real problem - How does calculating such a large web of trust compare to the complexity of running the game itself? That is to say, how many percent of ANets sever park will be busy with these calculations?

I would suggest putting up a few computers that handle all trust calculations as fast as possible. That might mean the calculations are not updated for a few days, but this will detract very little from the utility of the system in itself. The system as such is realistic, the question is only how often the trust information will be updated.

There are a few mitigating circumstances - Agglomeration of trust groups and that there is no need to calculate trust webs for inactive players. But essentially, this brings up the question of decaying trust. Might one prune the tree at inactive players? I think this is a reasonable thing to do, and it would certainly reduce the processing power needed to calculate the trust webs.

Mmmm. Long winded, but interesting for me to write

------------------------------------

Edit a few minutes later. And yes, there is a better way to do it. Since the system is distributed by nature, it is only natural to solve it in a distributed manner:

Let each client ask the "fully trusted" contacts for their local trust lists and propagation will be between clients, with the server only ever seeing (or rather caring about) the first step trust relations.

This seems so simple I'm astounded at myself for not thinking of it in the first place

The only backside of this scheme is that you will only get fresh trust information when you are logged on at the same time as your "fully trusted" contacts, but this is a light trade off, not worse than pruning the calculation tree at inactive players, as in my original suggestion.

So by my understanding, you're offloading the processing of determining trust levels to the individual workstations? Does my client request this information for each town I enter, or is it just pulled in at logon time? It also sounds like you're suggesting my client talks to your client if you're on my fully trusted list. Somehow that feels icky (sorry for the non-technical term there).


So the trust calculation occurs only in town (cause no one in their right mind wants to waste cpu cycles when they're out on a mission or quest). For larger population towns such as Lions Arch my system will have to chunk through each new person entering the town while I'm there and see if they're in my list AND what kind of glow to give.

Dunno, seems like that's a hefty strain to put on a gaming rig, of course it is in a "safe" zone so that might be worth the trade off.

Well, it might work to have the clients talk directly to eachother, but firewalls and somesuch would be a problem (sure, UDP shotgunning and whatnot can solve some of that, but it's still not elegant). No, the suggestion is that the clients communicate using jus the same method they communicate today - Via the server.

The work of analysing what trust indicator to display for a certain player would be a tough job, but the number of players in a given district is limited to a fairly small number (a couple of thousand at most), and finding the trust level of a given player in an ordered database is very fast. Moreover, the trust calculations can be put very low on the processing pecking order - It really does not matter if it takes one minute to "cach up" on a district.

I'm very confident that this kund of system would work, at least computationally. And since none of the posters have managed to spot a way to corrupt the system as such... It might just be the Alexanders cut

Wouldnt having any sort of thing on the client side lead itself to hacks and other stuff? Not that there would be a really huge advantage in hacking trust, but...

Also, given the difficulty that some people have in understanding the proposal, even with the explanation and several pages of clarification, isn't it highly likely that many people who dont even bother to read the manual would misunderstand the system also? Thus allowing for trust-scamming, etc.

I know they're only stupid people and we dont care about them, but when they then turn around and complain on the forums it's annoying to see their posts clutter up and distract from my elegant, innovative, and well-written (and completely ignored ;) posts :P


Edit: I do like the idea in principle, tho, it's very elegant. But it requires a lot of people to use it to fulfill its purpose (ie. to decide whether or not you're likely to enjoy playing with a random stranger) and it's a little too complex "For the masses".

It might be easier just to give people a score based on how much swearing, all-caps, and certain provocative "key phrases" they use in team and all chat, since to me that's usually a pretty good indicator :P

Surpisingly not - Any "artificial" changing of the scores of the people on the list would simply be equal to rating the person in question personally. This you can of course already do, so it's meaningless to tamper with the list, since the system already allows you total control over all your scores.

Of course, the majority of the player base is expected to not actually understand how the system works. But on the other hand, how many do you actually think understand how the other game systems work? I'd say that most players are not aware of hit locations, but the system works just fine anyway, eh?

Seriously though - It is fairly easy to explain how to use the system. In fact, I'd say it can be done in two or three sentances. The hard thing is explaining how it actually works, and why it's resistant to tampering.

Thank you! But I disagree with your opinion that the system would be too complex to operate by the average player. In fact, handling the system requires very little actual understanding beyond the ability to rate a player to one of three trust levels, and an understading that the trust you see displayed on many players are what your friends and their friends think of the person.

Basically - Right-klick, rate, squint near-sightedly at the trust rating displayed on other players

what about a trust rating ie like faction

complete a bonus with a group of all humans 3pts
complete a mission with a group of all humans 3pts
same with quests

4 henches 0 pts
2 henches 1 pt
1 hench 2pts

use an emote to symbolize trust in towns

dinglenut, if I understand you correctly, what you propose is another system entirely - Kind of a reward system for completing missions with people. It's not a bad idea, just a bit different from mine. If you don't want to, maybe I'll start a new tread about it? At least it'll be easier to explain than the distributed web of player trust

There would be soo many ways to cheat this system.