d2jsp
Log InRegister
d2jsp Forums > Diablo II > Diablo 2 Discussion > Ladder Tracker > Character Statistics
Prev1171819202143Next
Add Reply New Topic New Poll
Member
Posts: 36,368
Joined: Mar 31 2010
Gold: 106,698.50
Aug 13 2020 07:25am
Quote (Rontu @ Aug 13 2020 08:58am)
of course there are
lets be honest, the best mechanisms to detect bots are ingame
the only information we have is experience points over time

so we have some parameters (which we tested against known legit and known bot players) to run our bot detection by analysing the past experience values


As you should. Testing against cases you can confirm is smart.
All I'm saying is that if you're only going on experience over time and simply setting a cap that shouldn't be human-possible, then you're also catching users who share their account with friends to level quickly, which actually happens when the race is taken seriously (mostly in the early ladder seasons however).

I'll think about some extra measures to alleviate this. Maybe some kind of a weighting that includes not only your primary check on experience over time, but also what name the character has, and if other characters at similar levels have similar names and uptimes.
Depending how granular the sampling is, which from the code you sent it might not be enough, but if it was you could look at experience over shorter timeframes knowing that bots will get more experience out of each game (perfect teleporting, no hesitation or time waste, fast aura switching and skill swapping, long range casting especially on teleport, and lastly shrine finding by other bots to nearly guarantee 50% more experience).
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 14 2020 07:14am
Quote (laztheripper @ Aug 13 2020 03:25pm)
As you should. Testing against cases you can confirm is smart.
All I'm saying is that if you're only going on experience over time and simply setting a cap that shouldn't be human-possible, then you're also catching users who share their account with friends to level quickly, which actually happens when the race is taken seriously (mostly in the early ladder seasons however).

I'll think about some extra measures to alleviate this. Maybe some kind of a weighting that includes not only your primary check on experience over time, but also what name the character has, and if other characters at similar levels have similar names and uptimes.
Depending how granular the sampling is, which from the code you sent it might not be enough, but if it was you could look at experience over shorter timeframes knowing that bots will get more experience out of each game (perfect teleporting, no hesitation or time waste, fast aura switching and skill swapping, long range casting especially on teleport, and lastly shrine finding by other bots to nearly guarantee 50% more experience).


Thanks a lot for your input, I really appreciate it. Lets talk about it :)

All I'm saying is that if you're only going on experience over time and simply setting a cap that shouldn't be human-possible, then you're also catching users who share their account with friends to level quickly, which actually happens when the race is taken seriously (mostly in the early ladder seasons however).
-> That's totally true, those people might get flagged as bots. Well I can say that this did not happen (to my knowledge) in europe this ladder - there was no account sharing to play close to 24/7 legit. So I guess this just hasn't popped up as a problem yet, as people seem not to play like this.

Maybe some kind of a weighting that includes not only your primary check on experience over time, but also what name the character has, and if other characters at similar levels have similar names and uptimes.
-> I understand and we also had very early thoughts about that. In my opinion there is no reliable connection between usernames and the player being bots. Checking similar charnames is problematic as guild members tend to use a prefix (DoH-Benny, DoH-Hit, DoH-Rontu) aswell as legit players like to use similar char names (RontuSorc, RontuSmite, RontuHdin). Even the combination of 6 players with same name prefix being online and leveling same time might be okay, as one could host (legit) a 6p baalrun with 1 killer and 5 leechers (which is very easy at throne).

Depending how granular the sampling is, which from the code you sent it might not be enough, but if it was you could look at experience over shorter timeframes knowing that bots will get more experience out of each game (perfect teleporting, no hesitation or time waste, fast aura switching and skill swapping, long range casting especially on teleport, and lastly shrine finding by other bots to nearly guarantee 50% more experience).
->I don't know, but talking much with DoH-Benny I get the feeling that high-quality-baalruns also tend to have similar experience output per run. Usually we also have that shrine search and all that stuff. So taking this into account might be very difficult.

Overall:


The challenge in this project is to find a good balance between "finding-max-amount-of-bots" and "making-sure-no-legit-player-gets-flagged". And there's just a fine line betwen that. The current measurements we have are very safe and also not taking too much effort. However I really like your idea of adding a parameter how long we have been into the ladder, I might include that :)

We have also thought into terms of sending players ingame questions to check if someone is really present or not - but this could be classified as spambot by blizzard or even be very annoying to the community, so this might be a bit dangerous.
Or we add some only human-solveable question on the website and players must be pm us something from their char ingame. But that's also not very intuitive...

To be honest, I think there is no way of tracking everyone. That's where the community gets important to me. Every person can go on the website and inform us (with given proof) about a non-legit player and we add manual flaggig (or unflagging). This seems to be the best thing we can do. Our guild can do that pretty reliable on europe, but not on the other realms and modes. That's where I'm dreaming we could all help together a little bit
:wub:
Member
Posts: 36,368
Joined: Mar 31 2010
Gold: 106,698.50
Aug 14 2020 07:30am
Quote (Rontu @ Aug 14 2020 09:14am)
Thanks a lot for your input, I really appreciate it. Lets talk about it :)

All I'm saying is that if you're only going on experience over time and simply setting a cap that shouldn't be human-possible, then you're also catching users who share their account with friends to level quickly, which actually happens when the race is taken seriously (mostly in the early ladder seasons however).
-> That's totally true, those people might get flagged as bots. Well I can say that this did not happen (to my knowledge) in europe this ladder - there was no account sharing to play close to 24/7 legit. So I guess this just hasn't popped up as a problem yet, as people seem not to play like this.

Maybe some kind of a weighting that includes not only your primary check on experience over time, but also what name the character has, and if other characters at similar levels have similar names and uptimes.
-> I understand and we also had very early thoughts about that. In my opinion there is no reliable connection between usernames and the player being bots. Checking similar charnames is problematic as guild members tend to use a prefix (DoH-Benny, DoH-Hit, DoH-Rontu) aswell as legit players like to use similar char names (RontuSorc, RontuSmite, RontuHdin). Even the combination of 6 players with same name prefix being online and leveling same time might be okay, as one could host (legit) a 6p baalrun with 1 killer and 5 leechers (which is very easy at throne).

Depending how granular the sampling is, which from the code you sent it might not be enough, but if it was you could look at experience over shorter timeframes knowing that bots will get more experience out of each game (perfect teleporting, no hesitation or time waste, fast aura switching and skill swapping, long range casting especially on teleport, and lastly shrine finding by other bots to nearly guarantee 50% more experience).
->I don't know, but talking much with DoH-Benny I get the feeling that high-quality-baalruns also tend to have similar experience output per run. Usually we also have that shrine search and all that stuff. So taking this into account might be very difficult.
Overall:

The challenge in this project is to find a good balance between "finding-max-amount-of-bots" and "making-sure-no-legit-player-gets-flagged". And there's just a fine line betwen that. The current measurements we have are very safe and also not taking too much effort. However I really like your idea of adding a parameter how long we have been into the ladder, I might include that :)

We have also thought into terms of sending players ingame questions to check if someone is really present or not - but this could be classified as spambot by blizzard or even be very annoying to the community, so this might be a bit dangerous.
Or we add some only human-solveable question on the website and players must be pm us something from their char ingame. But that's also not very intuitive...

To be honest, I think there is no way of tracking everyone. That's where the community gets important to me. Every person can go on the website and inform us (with given proof) about a non-legit player and we add manual flaggig (or unflagging). This seems to be the best thing we can do. Our guild can do that pretty reliable on europe, but not on the other realms and modes. That's where I'm dreaming we could all help together a little bit
:wub:


Yeah the name thing is touchy, however if the character is named eorikjffsf, it's far more likely that is a bot. People who end up pouring days and weeks of time into a character tend to name their character something they feel is cool.
And there's keywords you can look for, like known pvp modder groups who use botting to level their characters. IHPK for example.
As for the per game experience, there may be too much overlap between legit and non-legit runs, but at the highest end, regardless of what the legit people would like to claim, bots outclass them quite handily in how much experience is reliably gained per game.
At the moment I don't believe anyone is using the 15-20 second diablo/pindle/baal setup, but if they were, you could see 2.5 day 99's. That's not something any human can do.

Edit: you're probably right overall though. If a person gets caught by this limit, they can contact you, which like you said is vanishingly small of a group these days.

This post was edited by laztheripper on Aug 14 2020 07:32am
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 16 2020 10:07am
bump
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 16 2020 12:57pm
thanks a lot for 50 fg donation :wub:
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 18 2020 10:16am
Up
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 20 2020 04:00pm
Up
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 22 2020 10:28am
up
Member
Posts: 2,903
Joined: Aug 25 2009
Gold: 170.00
Aug 22 2020 03:29pm
its pretty nice to have this man, but can you stop promoting this "we are here for the legits" while you use bots and third party software to generate and create this list?
Member
Posts: 15,272
Joined: Aug 9 2009
Gold: 2,387.00
Aug 22 2020 04:37pm
Quote (FreeUsername @ Aug 22 2020 11:29pm)
its pretty nice to have this man, but can you stop promoting this "we are here for the legits" while you use bots and third party software to generate and create this list?


Check the site, if a bot filter is not extra benefitial for the legit players, I don't know what is.
Not sure why you care about any technical background (which I didn't even explain here - and will not).

Since we decide how we advertise our site, and we do not have to please everyone, we will not change anything in this respect.

This post was edited by Rontu on Aug 22 2020 04:38pm
Go Back To Diablo 2 Discussion Topic List
Prev1171819202143Next
Add Reply New Topic New Poll