smoke wrote:

 Of course, asm is a just a bit of a jerk in RL.

You know asm in RL???

smoke wrote:

Take for example, KINSKI and LIONHEART, since it's public anyway -- see the conversation in this game from earlier today, where I was a less than gracious winner: http://www.wargear.net/games/view/52930

Here was my investigation:

2- Kinsky and LIONHEART: Same IP address, although they have a decent mixture of who wins.  My guess would be they possibly have preconceived alliances but maybe not the same person (ie family or something).

Same IP address, including that game referenced by smoke.

Yertle wrote:

smoke wrote:

 Of course, asm is a just a bit of a jerk in RL.

You know asm in RL???

Little bit. Someone I'm close to knows him very well, he's a good guy, lot of potential. I'm guessing he's getting his act together, which includes cold turkey on wargear.

smoke wrote:

Yertle wrote:

smoke wrote:

 Of course, asm is a just a bit of a jerk in RL.

You know asm in RL???

Little bit. Someone I'm close to knows him very well, he's a good guy, lot of potential. I'm guessing he's getting his act together, which includes cold turkey on wargear.

AHHHHHH! Super great to hear that he's alive!  Risky and I thought something may have went horribly bad in Greece!

Pass the word along that WarGear says hi!  But tell him to keep out if that helps him!

I am glad to hear that he is alive as well; I shared your worry Yertle (and Risky).

Yertle wrote:

smoke wrote:

Take for example, KINSKI and LIONHEART, since it's public anyway -- see the conversation in this game from earlier today, where I was a less than gracious winner: http://www.wargear.net/games/view/52930

Here was my investigation:

2- Kinsky and LIONHEART: Same IP address, although they have a decent mixture of who wins.  My guess would be they possibly have preconceived alliances but maybe not the same person (ie family or something).

Same IP address, including that game referenced by smoke.

 

I did not mean to give these names publicly, but as it's done, now I would say...

Could be they are not the same person, but LIONHEART is lying in the above mentioned game when writing "like i said the same names crop up in most of the games and kinski is in more as we are the same timezone unlike all the US players mentioned". They are not only sharing the same timezone, they are sharing the same IP so they must know each other. Same person or not, they are both in my ennemy list and will definitely stay there.

Yertle wrote:

smoke wrote:

Yertle wrote:

smoke wrote:

 Of course, asm is a just a bit of a jerk in RL.

You know asm in RL???

Little bit. Someone I'm close to knows him very well, he's a good guy, lot of potential. I'm guessing he's getting his act together, which includes cold turkey on wargear.

AHHHHHH! Super great to hear that he's alive!  Risky and I thought something may have went horribly bad in Greece!

Pass the word along that WarGear says hi!  But tell him to keep out if that helps him!

Will do.

Yertle wrote:

says hi!  But tell him to keep out if that helps him!

ditto, been missin him.

I think we need the site needs to be more proactive in catching and banning players using multiple accounts to cheat rather than relying on people to report suspicious behavior. The obvious way to do this is via game log file analysis. I am thinking of an automated weekly process which takes the game log entries for the week, analyses them and then comes up with a list of players who's behavior is suspicious for an admin to look at. Within the log file there would be the following data fields:

Timestamp
Player ID
Game ID
Seat
IP address
Action (including player ID of the player attacked)

Red flags
Players sharing the same IPs
Players playing in the same games together with the same account winning consistently
Players playing in the same games together with one account infrequently attacking another
Players playing in the same games together with very short breaks between turns if they are in adjacent seats

Unfortunately there is going to be no guaranteed way to catch cheats - the results of the analysis above will need to be double checked by someone to see if there is genuine cause for suspicion. But I think it is a good first step to improve detection of multi account abuse before it becomes a problem.

I don't want to be in a position where players are banned from playing with friends, family of work colleagues as this is contrary to the spirit of the site. I think in most cases where there is suspicion a warning will be sufficient (i.e. don't start games with pre-conceived alliances). In very obvious cases of cheating then the ban hammer will be used instead

Right now banned players are not removed from rankings. The way that this needs to be fixed is to run through all games they played and if the banned player won then that game is set to 'Terminated' instead of 'Finished'. Then once this is done the ranking stats are recalculated for all the games which will effectively result in them not receiving any ranking points (and any ranking points lost to them being given back). This is on my ToDo list to action - it's easy to automate.

tom wrote:

Right now banned players are not removed from rankings. The way that this needs to be fixed is to run through all games they played and if the banned player won then that game is set to 'Terminated' instead of 'Finished'. Then once this is done the ranking stats are recalculated for all the games which will effectively result in them not receiving any ranking points (and any ranking points lost to them being given back). This is on my ToDo list to action - it's easy to automate.

This is a good solution.  I'm glad to hear that it's not too hard to do.

Not that it's a big deal, but I'm thinking there will be slight discrepencies because ranking points are based on when games are played.

Take the example where a player loses 25 points to a cheater while their rank is 1500.   Then they go on to lose 25 more to a legitimate player because their rank is 1475, bringing their score down to 1450. Isn't it possible that had they lost to the legitimate player while their rank was 1500 they would have lost 26, not 25?

Like I said, it's not a big deal.  Global ranking scores fluctuate pretty wildly anyways and I think it's pretty safe to say that this won't affect a 50-day moving average at all. 

BTW, I can't wait to see Moving Averages included in the data.  I think somewhere between 50 to 100-day MAs will really show where people stand.

tom wrote:

I am thinking of an automated weekly process which takes the game log entries for the week, analyses them and then comes up with a list of players who's behavior is suspicious for an admin to look at.

Sounds like Common Opponents on Profiles would be good to have too

As for the Red Flags, they sound good as long as there is a combination of at least 2-3 of those flags before suspicion is really raised.  For example in a Real Time game players are going to be making very short breaks between turns, so it would pretty much have to be a combo of all 4 flags before you start getting sound evidence.

Once flagged, it's just a matter of a human looking at suspect games a bit closer.

Depending on how many red flags appear, perhaps 3 or 4 members would volunteer to help out.

I'm sure Toto would be happy to put himself at the top of that list.

M57 wrote:

Not that it's a big deal, but I'm thinking there will be slight discrepencies because ranking points are based on when games are played.

 

Don't worry about that part, the game rankings are calculated by resetting to zero then looping through each game chronologically by the time it finished. So it will be 100% correct.

tom wrote:

M57 wrote:

Not that it's a big deal, but I'm thinking there will be slight discrepencies because ranking points are based on when games are played.

 

Don't worry about that part, the game rankings are calculated by resetting to zero then looping through each game chronologically by the time it finished. So it will be 100% correct.

That sounds like it would take a while, especially as the number of games played continues to increase.

It takes about 20 minutes... but it gets longer and longer!

Yertle wrote:

Sounds like Common Opponents on Profiles would be good to have too

Whether I am a cheater or not it would be fun to have a place on my home page that lists my top 5 opponents and my record against them in public games --> Win/Loss/Tie  where Tie is someone else won.  

That this helps find multiple accounts (or the other way around) is a bonus.

It looked to me like they're family or friends, that's how they play. So I didn't want to just accuse of cheating publicly, when maybe what they need is a stern lecture on going into a game as a team.

Quite likely that is it.  I know that when I first started on ToS, I played only private games with family & friends.  I realized that when I played a public game with someone I knew, I was much more likely to make a treaty with them than another player.  It wasn't even that we had some truce from the start, but just that I was kind of reluctant to approach a total stranger with some deal, but I felt more comfortable talking to a friend.  I am more careful to not do this in public games now, but it's still a temptation.  This was one reason I proposed having a formal 'treaty system' for the site - I thought it would make treaties more likely between strangers.   But I guess it got shot down for that same reason, since more treaties in general was not seen as desirable.

tom wrote:

Red flags
#1 - Players sharing the same IPs
#2 - Players playing in the same games together with the same account winning consistently
#3 - Players playing in the same games together with one account infrequently attacking another
#4 - Players playing in the same games together with very short breaks between turns if they are in adjacent seats

Those red flags are all good indicators, but any two players in the same household are going to match 1 & 4, so I think those should only be used for confirmation, and you'd need to match one of the other criteria.

In calculating #3 (which I think is really the best indicator by far), you should probably only count 3v2 attacks so that a player doing 1v2 against someone, to suicide their troops doesn't get credit for that as a real attack.   Also, you probably should ignore cases when there are only two players left, as that could skew things.  In fact a robust solution might take into account the # of remaining players, as A attacking B only 1 in 20 times might be unusual when 3 players are not yet eliminated, but not when there are 12 players.

WITCH!!!!!!!

Alpha wrote:

doot em.

This is the internet. Don't you dare bring up minor catch phrases from six months ago.