General Turn Discussion

[Zeviz]

<delurk>
This is my first post in this forum, and I haven't had much interaction with Memphus, so while I am not entirely independent, I don't have a particular reason to favor this team. I am just speaking as a math major here.

All Sooooo's statements are correct. Further, the recent streak of victories make cheating LESS likely. The reason for that is that streaks are more common than people think. There've been plenty of experiments where people were asked to write down a random sequence of numbers, and these sequences are compared to real random sequences. In most cases, real random sequences are more streaky. So if Memphus was cheating, he'd throw in a few losses because "it's impossible to win all the time".

Also, you aren't looking at the right sample size. The sample size isn't "all battles in this game". It is "all battles in all competitive games of CIV described online". How many SGOTMs have there been? Did anybody ever report this kind of winning streak in any of them?

However, the sample size is irrelevant here anyway: This could be the only CIV game in the world, and Memphus could get these results without cheating. People tend to forget that 1 in 100,000 odds doesn't meant that 100,000 players have to play for somebody to get that result. The odds simply mean that if 100,000,000 people play, you'd expect to see the result about 1000 times. However, the first person to ever play the game would be just as likely to win as the 100,000th person.

The point of the above is that while you can use statistics to help choose whom to investigate (and I agree that Memphus warrants investigation, even if only to prove your vigilance), you absolutely CANNOT use statistics as evidence.

There are only two ways to get real evidence:
1. If we are absolutely sure that replaying attacks will generate same result, replay Memphus' attacks yourself.
2. Ask every CIV modder if there is any way to cheat in PitBoss while the game is connected to Civstats.

PS Looking at further combat results will be pointless:
Option 1 - SANCTA starts losing most battles - explanation will be that Memphus was cheating before, and is now cheating in the other direction because of scrutiny.

Option 2 - SANCTA continues the winning streak - explanation will be that Memphus is still cheating.

Option 3 - SANCTA starts losing "correct" number of battles - explanation will be that Memphus was cheating and stopped.

Any result that will happen in the future will only "proove" to people that Memphus was cheating. So the only meaningful investigation you can conduct is by the two methods listed above.

 

[Love]

Can someone do a summary on the current situation? Because I am really too lazy to read all these long posts.

 

[Memphus]

Summary

sancta wins when attacking
other teams get mad
Sulla defends sancta
most recent battle was to unlikely for Sulla to believe
Sulla now questions sanctas integrity ( but directly Memphus)
other teams have decided we are cheating

and that is where we are at and why the game is paused.

 

[Krill]

most recent battle was to unlikely for Sulla to believe

Replace 'to' with 'too' and I'd agree...

 

[Whomp]

Sullla--Being a moderator at this site I take these allegations very seriously. I have no problem bringing the moderators into this game and even higher if need be.

DaveShack, also being a moderator, would be very available to whatever is needed to rectify this situation.

I'm having a hard time understanding how one can cheat in a pitboss game but I'll address it and will listen to any evidence on how this could be done.

In the meantime, I would suggest either the game be taken down until this can be discussed by the moderators or we continue playing and have another player execute the turnset with Memphus' (or anyone else for that matter) assistance.

Which is preferable?

 

[dsplaisted]

Right. Either there's a previously-unknown way to remotely hack the RNG to give the answers you want directly, or or a previously-unknown way to hack the RNG to tell you what the next number's going to be, and then either using it for the battle, or skipping to the next one. I don't see how either of those are plausible.

As I said before, disproving that we got two lucky streaks from the RNG in 50BC and 940AD, and were generally lucky (about 1% chance) of the single combats going the way they did, losing no units, isn't happening. 1AD was a "normal" good round of combat. So we play on, and pray that the next battle isn't a ridiculously streaky one, when these questions will come at us again.

Lurker's Comment: I am pretty sure that you can figure out the state of the RNG, and hence figure out what the combat results will be before attacking. For multiplayer games, it seems that the entire game state is sent to your computer, so if you poke around in memory you could figure out what the sequence of numbers the RNG will spit out will be (as well as all the other information you aren't supposed to see such as where your opponents' units/cities are, what they are building/researching, etc.).

If you can figure out the state of the RNG, you could then sim out the whole battle and figure out exactly what would happen if you attacked with your uniits in different orders. I'm not sure that would be enough to get the kind of success that team SANCTA has seen, however. If you had a way to use up values of the RNG, you would have more flexibility to get results. I'm not sure there are many ways in the game to do this, especially repeatedly. You can try spreading a religion to a city that already has one, and you can put a unit on auto-explore, although I doubt that will actually use up a random number if the unit has already moved that turn.

So it looks like it might be possible to cheat in this fashion. Even if it is, it seems unlikely that someone who was able to do so would do it in such a way that they won every single battle with above 21% odds. That just seems too detectible.

 

[mwin]

Doesn't RNG through the dice again after the attack started. For example, if you have 45% odds of winning, then doesn't it need to generate a random number with 45% odds of winning? Or is it like the random number is always random and then you multiply that number with the odds somehow to get final number?

If it is first one then there is no way to get RNG number until you start attacking the unit.

Lurker's Comment: I am pretty sure that you can figure out the state of the RNG, and hence figure out what the combat results will be before attacking. For multiplayer games, it seems that the entire game state is sent to your computer, so if you poke around in memory you could figure out what the sequence of numbers the RNG will spit out will be (as well as all the other information you aren't supposed to see such as where your opponents' units/cities are, what they are building/researching, etc.).

If you can figure out the state of the RNG, you could then sim out the whole battle and figure out exactly what would happen if you attacked with your uniits in different orders. I'm not sure that would be enough to get the kind of success that team SANCTA has seen, however. If you had a way to use up values of the RNG, you would have more flexibility to get results. I'm not sure there are many ways in the game to do this, especially repeatedly. You can try spreading a religion to a city that already has one, and you can put a unit on auto-explore, although I doubt that will actually use up a random number if the unit has already moved that turn.

So it looks like it might be possible to cheat in this fashion. Even if it is, it seems unlikely that someone who was able to do so would do it in such a way that they won every single battle with above 21% odds. That just seems too detectible.

 

[Sullla]

Folks, I'm well aware of the oddities that random chance can throw out. I actually have a degree in statistics, if you can believe it. It's entirely possible that all of the results we've seen are perfectly random. It's also very possible that they aren't as well. As someone watching all of the team forums, it was my responsibility to take up this issue when it was raised by the other teams. I believe in the accused being innocent until proven guilty, which is why this entire discussion has taken place here, rather than in the General forum where you'd have Kaz and Cav accusations turning it into a flamefest.

Obviously there's no proof that any cheating took place. But of course there's no proof that cheating didn't take place either, which is why we're in this predicament. Krill can point out that SANCTA lost a unit to a barb at crazy odds, but a naysayer could respond by arguing that mere barbs weren't seen as worth spending the time cheating. That route leads to paranoia - we could go around in circles like that forever. Best to spend our time in more productive fashion.

The overwhelmingly best defense of SANCTA is the fact that no one seems to know any way to cheat in a Pitboss game. I've been searching around, and I haven't found much. The one concern that I have found is the fear that in a Preserve Random Seed Pitboss game (like this one), it may be possible to download a savegame and have the random seed be preserved locally. Then one could theoretically test out a coinflip attack locally, find out the result, and only make the move online if your unit were the winner. Going further, one could keep trying out a long chain of unit attacks until one hit on the perfect random path to win all (or nearly all) of the combats. That could potentially explain some of the SANCTA combat streaks (e.g. retreat a chariot here to change the random seed so that the next combat wins at low odds).

I'm not saying that SANCTA did do this, merely that it could be possible for SANCTA to do this. Before moving forward with the inquiry, I'd like to get confirmation one way or another if this method of preserving the random seed is possible. If not, SANCTA is probably in the clear. If it is possible, things get a lot more dicey.

Again, for the record, I'm not raising these issues because I hate SANCTA, or I think Memphus is a dirty cheat. I'm doing so because the other teams have made complaints, and it's only fair that they have them settled one way or another.

 

[dutchfire]

My knowledge about the RNG (could be wrong):
Every call to the RNG (spreading religion, does a random event happen, who wins combat...) comes from the same RNG, so a call to either of those will change the result of all subsequent events.
A battle isn't just a single call to the RNG, every round in the battle is a separate call. That makes simming the battles from a known seed a lot harder, since you're unlikely to find long sequences of "good" RNG values.
There aren't many ways to call the RNG during a turn. Spreading a religion is one, but we haven't had a missionary all game, have we? Popping a hut (haven't done that much either) is another one.

I don't know if the RNG seed (or other information like the location of enemy units) is loaded to your computer (I'd suspect not, but I don't know how PitBoss works), even so, tracing the RNG seed from that information seems like it requires very hard core computer knowledge, and I don't believe anyone has ever done it.

Which brings us back to the main point: Is there any suggestion to how this cheating is supposed to have happened? I don't think any discussion on wheter or not this is/can be cheating is pointless if we don't know how the cheating is supposed to have happened.

 

[dutchfire]

Well, that was a nice crosspost with Sulla.

Sulla, could you try to recreate the process you're describing, to see if it's really possible?

edit: It's the 1990 post in the thread, my year of birth I don't know many threads that reach that number.

 

[Krill]

This is quite simply farcical. So, things get dicey simply because they may be a possible way to cheat, and we get lucky? I'm liking the prejudgment there.

 

[Ronnie1]

WOW...gone for a couple days, and it takes me 2 hours of reading to catch up!

I can say from the private pitboss that we were playing, Memphus does not ALWAYS get lucky. I used his strategy of continuing to attack, even at low odds, to completely wipe out one of his stacks that should not have been possible. I was on offense, that seems to be the thing that sticks in my mind! If you DON'T TRY, you CAN'T WIN!

Why didn't he "cheat" when he was playing against me? Cause I am starting to kick his A$$!

 

[Memphus]

I am so happy I shared that strategy with you too . Overall makes for a more exciting game. As much as it sucked to see those loses

 

[Ronnie1]

There is a certain amount of insanity involved....but as you explained it to me...you have to try, or you NEVER succeed!

 

[mostly-harmless]

I think I understand the method Sullla is outlining. However wouldn't that most likely result in a very weird order of attack? So you try or test the RNG until you found a match that gives you a win despite bad odds. Surely that wouldn't result in a sensible attack order like Memphus documented. And I believe the attack order is shown in the combat logs.
If Sullla outlined method would have been used, I would expect something crazy like:
cat first, then wounded HA, then WE, then a couple of mace, then a longbow, then another cat and so on and so forth. Now, that would have been very suspicious.

Also, I like to echo Krill, just because someone might find proof that the method is possible, does not bring Memphus an iota closer to be a proven cheat.

mh

 

[Sullla]

So... no one has actually gone so far as to test what I requested yesterday? I was hoping that someone would have a definite answer one way or another by now. I really don't have time for this, but I will try to address the problem myself.

Please hold the snarky comments. I've explained why it's important to take allegations of cheating seriously. Unfortunately, we can't simply go by Memphus' word on this.

 

[Krill]

EDIT: nm.

 

[Memphus]

Sulla I think I also undertand your request for a test but am unaware how to perform it ( not to mention I doubt I am the one who should do the test)

but if I understand it correctly then wouldn't we need cav to test it? Or am I missing something? Can the game even be open locally if it isn't your turn? (meaning only cav could test this right now?)

I am trying to stay calm in all of this but again why I'd the burden of proof on sancta (and me).

Isn't the legal system innocent until proven guilty beyond a reasonable doubt.
It seems like we are operating on a guilty until proven innocent and the burden is on us.

 

[sooooo]

I thought the mods were investigating? See this quote from Rik Meleet in the public thread:

Moderator Action: Please pause the game.
We'll let you know when to continue. Shouldn't take long.

We should probably see what they come up with. The game has been paused for a while now - what have they found?

Seems a bit strange to task SANCTA with the task of finding a way to cheat in order to provide evidence that SANCTA are cheating .

By the way Sullla, this sentence:

Folks, I'm well aware of the oddities that random chance can throw out. I actually have a degree in statistics, if you can believe it. It's entirely possible that all of the results we've seen are perfectly random. It's also very possible that they aren't as well.

directly contradicts your earlier stance of:

It's simply not possible for these to be random results. I'm sorry Memphus, but I don't believe you. No one can possibly be this lucky, so consistently, over such a long period of time.

So have you changed your mind over the statistics?

 

[Sullla]

Rik Meleet's post meant that I was investigating the issue, sooooo. Before I was asking if anyone on SANCTA had been able to determine whether or not the random seed was preserved for combat results when downloading savegames from an online Pitboss server. Since no one had an answer for that, I went and checked it out myself. Here's what I found.

I'm sorry to report that it is in fact possible to swing the combat results in your favor by means of the process we've been discussing. If "Preserve Random Seed" is turned on, then one can test coinflip battles ahead of time to see if they will be victorious or not. It's less easy to swing multiple battles in one's favor, but you can indeed test out various "strings" of combat results, and keep testing various permutations of dice rolls until finding one that is most heavily in your favor. Furthermore, this appears to be fairly well known in the Multiplayer community, to the point that many MP Ironman games are done with Preserve Random Seed off, just to avoid this situation from occuring. I heard stories that some people were using two computers at once, running a MP game on one and sending saves over to the second computer, just for the purpose of testing out combat results. Hard to believe, I know, but that's what I was told.

Unfortunately, this more or less blows up the argument of "it's impossible to cheat in a Pitboss game", and we're left with nothing but Memphus' word again. While I don't want to call Memphus a liar and a cheat, it's also true that there's a lot of circumstantial evidence here that arouses some suspicion in the combat results. Furthermore, I have to point out that Memphus has been playing hard and fast with many of the Demogame elements, such as continuing to mislead the other teams with research swapping even after the issue was raised as somewhat of a cheese tactic. If asked as to whether or not I thought Memphus had been using savegame results to predetermine combat results, I would have to answer by saying I don't know. That's not an accusation, but it's not a confirmation of innocence either. This is not something so inconceivable that I couldn't see it happening. I just don't know.

My recommendation after considering the situation is once again that SANCTA appoint someone else to play their turns for a little while, and let Memphus take more of an observational/lurker role. Just until the end of the current SANCTA vs. Kaz/Cav war, since that's the major bone of contention. Since this team seems to have a dozen or so members, that shouldn't be too difficult to coordinate. This isn't a punishment or an accusation of guilt for Memphus. Rather, I think it's the best way to cool down a volatile situation with some hot tempers on both sides of the issue. Since innocence or guilt is literally impossible to determine, a compromise solution seems best to me.