|
Post by Yojimbo on Apr 2, 2014 15:04:49 GMT
That seems like it might be an easier means than I thought up but I would still like a means to link a single pair of toons so that you can fill a party roll and earn the reward on the toon you originally brought to the run. This is only a concern in cases where the rewards are none transferable though so that is a factor if that is needed. Why is it only a concern in such cases? XR Tokens, which are the primary reason for this discussion, are (I'm fairly certain) transferable. The concern here is that if you're bringing bots to Limbo, and earning XR tokens on them, you'll be able to exchange them for XR items at a rate of twice, or even thrice, that of a player who isn't botting. You seem to think I mean a detection system is only of concern in case of non-transferable rewards I mean it is only of concern to have a means to forfeit the reward on one toon and xfer rights to it to another toon.
|
|
|
Post by CataclysmicDeath on Apr 2, 2014 15:54:38 GMT
Can I just ask, is there a way to implement a system allowing party members in game to 'vote' on whether a character in the party should be given bot status? And no that isn't my idea but the answer will determine if my actual idea has merit lol. Cata That won't prevent one (or a couple of players) voting all their characters as non bots to get them XR tokens which seems the primary goal of tthe whole bot detection thing. XR's are so uber that I can kind of understand the justification of attempting to make people play in 'real' parties to obtain them and reducing the ability of players endowed with multiple accounts of racking up absurdly high numbers of these items/tokens by dragging a bot army. Will keep the replay value of the XR areas and servers as a whole significantly higher than if people can get them all in short order and I'm speaking from the viewpoint of somebody who's run Oinos so much I've managed to get a copy of pretty much all of the BUR items in the game by doing this (in HC mode, had already got them all outside HC years ago). Yes, I know, which is why I said it wasn't my actual idea, I just wanted an answer to the question of if it is possible so that I knew whether it would be any use posting my actual idea. My idea is a little more complicated than this Cata
|
|
|
Post by Enius the White on Apr 2, 2014 16:39:36 GMT
Another way to discourage bots in Limbo is just by making Limbo very hard. A bot is of much lower value to a party than a real player. Very tough runs can thus make them non-viable altogether. Currently, there are no "Very tough" runs for a level 80 party, thus the extensive viability of bots. On even reasonably difficult runs (the most difficult currently available), a party of 10 players may complete a run 2-3 times as fast as a party of 5 dual boxers. The utility and efficiency of bots drops quickly as runs get to the upper end of difficulty. How many bots do we see on Nessus or Prince fights compared to Dusty, or Tia?
If Limbo is difficult enough, bots should be all but useless there (e.g. If 10 players can barely succeed, then 8 players and 2 bots will likely fail, and get no tokens at all). Balancing the party power required for a win very near the power of 10 players, will basically eliminate the desirability and viability of bots. There is then also a fairly broad difficulty "buffer" range, where a party with bots might succeed but at such a cost in time that the extra bot token(s) would be a comparatively inefficient reward. On an individual player basis, bots are currently always dropped before the loot split anyway, while consumables (e.g. bios), the risk of lost limbo (or slagged) gear, and even deletion in HC, falls exclusively on the dual boxer. In short, dual boxers are often the most passionate, dedicated end-game run formers.
As long as going with a party of 10 is significantly more efficient (tokens/time per player) than running with bots, then there is an argument for not worrying about it too much at this point. Players forming, and grinding through the inefficient tokens/hour "with bot" run path probably deserve their reward.
Inevitable, continued power creep is, of course, the scourge of this approach, as today's "very tough" Limbo may be tomorrow's "reasonably difficult" run.
|
|
|
Post by FunkySwerve on Apr 2, 2014 17:02:07 GMT
You read my mind elgrathforestwalker. The biggest delta in bot vs single toon player behavior, without getting very complicated, is probably how often the toon moves. A dual boxed toon will normally advance less frequently, sometimes only for those spawns where it's function is needed. Definitely far from perfect, but differentiating between an active dual boxer and inactive single toon player can actually be extremely difficult. This, assuming it's correct, is precisely the kind of thing it's useful for me to know - remember I don't play much these days, for better or worse. I like this metric, but I'm concerned that it'd be fairly easy to game it, by just twitching your bots every so often. It seems difficult to base bot-dom entirely on this, since even real players would go afk from time to time, and I'm hard pressed to figure out how often that 'should' be. Funky
|
|
|
Post by kaezar on Apr 2, 2014 18:44:10 GMT
You read my mind elgrathforestwalker. The biggest delta in bot vs single toon player behavior, without getting very complicated, is probably how often the toon moves. A dual boxed toon will normally advance less frequently, sometimes only for those spawns where it's function is needed. Definitely far from perfect, but differentiating between an active dual boxer and inactive single toon player can actually be extremely difficult. This, assuming it's correct, is precisely the kind of thing it's useful for me to know - remember I don't play much these days, for better or worse. I like this metric, but I'm concerned that it'd be fairly easy to game it, by just twitching your bots every so often. It seems difficult to base bot-dom entirely on this, since even real players would go afk from time to time, and I'm hard pressed to figure out how often that 'should' be. Funky What about the action thing? I thought in nwn we could resume all actions a player can take in about a half-dozen options: choose a target for attack, move, cast a spell, use an item, use a feat and use a skill I think. Some of those demand you to choose a target after. If we made a count of actions we could probably separate the bots by number. By going a bit further we could eliminate from the count actions that were equal to the last action, or last two, or whatever many actiions, to avoid somebody jamming the heal pot funcion key on a laptop and logging actions this way. I realize it is not perfect, but I can't think of anything better. Take care Kaezar
|
|
|
Post by Yojimbo on Apr 2, 2014 19:24:14 GMT
I think each action can have a certain "weight" to it so using a targeted spell or ability is worth 2 points moving and heal pots 1 point etc a dwindling return on repeated actions might not be bad either.
|
|
|
Post by Twilight Semner on Apr 2, 2014 21:44:27 GMT
I think actions are a great metric for measuring activity and determining whether or not a character is a bot. As are incoming and outgoing attacks. I kind of feel that if heal pots could be problematic, why not exclude them from the count, since they're probably the easiest to fake?
I don't feel that dwindling returns for tracking actions would help in the slightest. A system of diminishing returns inherently favors those who take less actions. If anything, you would want to have increasing returns for consecutive actions in order to better distinguish bots from human players. Of course, I'm not saying we should do that, just saying it makes more sense than the alternative.
|
|
|
Post by madzapper on Apr 2, 2014 21:53:20 GMT
I really believe the way to deal with this is to require people who have multiple accounts to report their list of accounts, and link their accounts (or give them a way to link their accounts) and make the link permanent unless removed by DM. Only allow one of those accounts to have non-bot status on a server at a time. It can be used in many ways from loot splits, party formation, etc. It's healthy, truthful, and responsible playing. I don't know why anyone would object as the overhead for the user really isn't tremendous. Give messages when entering areas of concern that bots must be clearly identified. My opinion is go ahead and allow an "unbot" to remove bot status from a character if there are two people playing from one location on linked accounts. Log it, shout it, don't be opaque about it, but be transparent and let everyone know that it is happening. Deal with the liars and cheats as necessary. Assuming all of that is not possible or desirable. (i.e. based on a well played bot should be able to get rewards), here are a few situations that I think can cause some real issues with determining bot-status. Inactivity: How long is too long without activity? 3 minutes? 5 minutes? 7 minutes? * Whole party waiting for timers. (note some people may have already been inactive prior to end of battle due to dying, etc) * Whole party waiting for resters/buffers/spell slot changes. * Members waiting for other party members to (slowly) dispatch of ultra KB, dangerous mob that can only be dealt with by select party members. * BillyBob has to walk his dog urgently, can you wait? * Veteran BillyBob needs to bio and gets intercepted for 10 minutes by knocking door. * spam heal pot, harm pot, str, wis, dex, etc. * any number of buffing spells cast at regular/irregular intervals * Gaming accessory with programmable macros and delay timers* play toon A for 3-5 minutes, play toon B for 3-5 minutes * cast harmless spell on self versus buffing * cast harmless spell on neighbor bot (i.e. GR) * New to the run versus pro at the run * High level 5x Demi iterations on a toon versus immortal toon on the same run. * Toon designed for that run versus one not designed for that run. (think called-shot zen-ranger last map of Rona) * I have two Internet connections in my home AND I have very comfortably played multiple accounts using my tethered phone as well. Once a character is determined to be a bot, are they flagged forever on the run as a bot? How long will it be before someone is like, "Geez, mate, I'm going to get flagged as a bot if you don't stop taking all my kills." or "C'mon, man, you know they make adult diapers for that." or "Anyone got any spare heal potions?" Bots don't usually do a lot of things, but without rules on them, they don't have a lot of reason to. I can honestly say, for example, that my bots rarely rest. It won't be a good metric, though, for determining if a character is a bot. Implementing a list of objectives in a game is likely to be taken as a challenge. Just sayin'.... I mean, this list can/will go on. If the metrics are too few, they'll likely be easily circumvented. If the metrics are too many, they'll likely be overly complex and difficult to reliably trap. Tracking any metric in real-time will require server processing and is going to have storage requirements. BTW, if BOTH solutions were implemented, bots wouldn't have to be checked as they would already be flagged as bots. If after all of this, this is still the thing to do, I think the following really needs to be addressed: For each "build" relative to each OTHER "build" in the party, there is an appropriate set of actions that make sense for that specific area. "build" being class level breakdown, level, demi-iterations, dex vs. strength, gearing, etc. Once you have that list, you can probably determine if a user is a bot. Anyone can say I am going overboard with that statement, but I believe nonetheless that it is probably an accurate statement. This problem is difficult. If you don't go to these type of extremes, I honestly believe that the accuracy rate will not be great. I'm out of time, so I'm not proofreading this, but I hope it at least gets looked at and provokes a few ideas. Cheers, Madzapper
|
|
|
Post by FunkySwerve on Apr 3, 2014 1:20:50 GMT
I'm short on time, so I'm doing a single post to respond to a number. Sorry. I don't feel that dwindling returns for tracking actions would help in the slightest. A system of diminishing returns inherently favors those who take less actions. If anything, you would want to have increasing returns for consecutive actions in order to better distinguish bots from human players. Of course, I'm not saying we should do that, just saying it makes more sense than the alternative. I'm pretty sure yohimbo meant diminshing returns of repeating the same action, so as to hinder things like auto-movement via macro being used to evade the metric. In response to those asking about the purpose of the system: The system isn't for determination of tags, but for 'per partymember' loot rewards. The entire point is to avoid the creation of an incentive to drag additional useless bots purely for more loot. By way of clarification, I want a system to flag bot vs. non-bot. At present, the only use for that system planned is to distinguish the two types for handouts of XR tokens. Other uses might or might not follow, depending on how effective it is. A more general reply: I think the way to resolve the weighting questions ultimately will be to set up logging of whatever factors we flag that I can track in a low-overhead way, and to analyze runs with a number of known bots. The main goal here is just to flag what factors will most likely be indicative of bot status. Funky
|
|
|
Post by rhodar on Apr 3, 2014 1:50:57 GMT
Just a quick note to express my hope that the algorithm will be "smart" enough to distinguish between the play of a bot and the play of a player who moves slowly, carefully, and only after a certain amount of voice-carried discussion of strategy.
Not to say "growing pains" are intolerable; there's a difference between "temporary inconvenience" and "play faster or go away."
|
|
|
Post by magecat on Apr 3, 2014 3:35:18 GMT
You read my mind elgrathforestwalker. The biggest delta in bot vs single toon player behavior, without getting very complicated, is probably how often the toon moves. A dual boxed toon will normally advance less frequently, sometimes only for those spawns where it's function is needed. Definitely far from perfect, but differentiating between an active dual boxer and inactive single toon player can actually be extremely difficult. This might be true, but using it as a threshold seems to be fairly easy to circumvent. What's to keep a player for moving their bot a few steps back and forth between spawns, or have them run around in circles while the party is resting? That being said, while on its own it's certainly not sufficient, it's still a great indicator when taken in conjunction with some other factors. Also, remember that there is an autofollow wand in the game that is painless to get. With a sufficiently capable primary, you could just latch your bot onto them.
|
|
|
Post by Yojimbo on Apr 3, 2014 12:57:35 GMT
I don't feel that dwindling returns for tracking actions would help in the slightest. A system of diminishing returns inherently favors those who take less actions. If anything, you would want to have increasing returns for consecutive actions in order to better distinguish bots from human players. Of course, I'm not saying we should do that, just saying it makes more sense than the alternative. I'm pretty sure yohimbo meant diminshing returns of repeating the same action, so as to hinder things like auto-movement via macro being used to evade the metric. Yes that is correct only for repeating the same action
|
|
|
Post by FunkySwerve on Apr 3, 2014 17:28:26 GMT
This might be true, but using it as a threshold seems to be fairly easy to circumvent. What's to keep a player for moving their bot a few steps back and forth between spawns, or have them run around in circles while the party is resting? That being said, while on its own it's certainly not sufficient, it's still a great indicator when taken in conjunction with some other factors. Also, remember that there is an autofollow wand in the game that is painless to get. With a sufficiently capable primary, you could just latch your bot onto them. I'm actually considering adding a second, more advanced widget allowing automated combat and healing, as loot somewhere. I'll be setting that status and checking for it and autofollow as a prefatory check in the bot check. Thanks for the reminder. Funky
|
|
|
Post by sabregirl on Apr 6, 2014 14:12:44 GMT
Well I think the obvious first check would be kills, as a general rule a bot doesn't get very many if any kills unless it's actively played. And for the support characters a metric like HGX's "helper" could also be used since bots don't usually Pt2 or rez too much. Spells cast would work as well but I don't know how easy general detection would be. The question would be cutoffs - perhaps use a percentage of the party's total? Since bots may occasionally do the things listed above, but unless they're actively played it won't be much of the party's total.
-S
|
|
|
Post by kingcamaro on Apr 6, 2014 15:51:55 GMT
It is very difficult to do a detection over so many different types of characters. IMO it will be best to detect based on class. So let's say for a bard, curse and songs, epic usage, movement, skill usage (taunt, persuade etc), even healing done with healing circles. For tanks, target actions, skill usage, damage dealt and received. For casters would be spells cast, movement etc
Basically check the classes are doing what they're supposed to. Some fights I'm sure some casters will be useless at while tanks may be useless for others, so perhaps a way to omit some things from the detection. If Jim spams 50 potions of healing while taking little to no damage, then it's not recorded as being active.
Id think overall, check if PC is being targeted by NPC. If so, then start "detecting" to see if other PCs are reacting accordingly.
|
|