The Barrow-Downs Discussion Forum - View Single Post - Tol-in-Gaurhoth XIV: Nyárë Nauroron Númenoreva

Aiwendil · 12-08-2005, 07:09 PM

Mathematical Methods of Wolf Detection

Okay, being a fan (as I said earlier) of quantitative methods, I figured I'd attempt a numerical analysis of people's voting records. I know this may sound like a dubious proposition, and I certainly don't claim that it is infallible. But it can't hurt to try.

The idea is that you gain suspicion points for having voted for a known innocent and lose them for having voted for a known wolf. Now, it seems to me that voting for a wolf is a stronger sign of innocence than voting for an innocent is of guilt. After all, the innocents outnumber the wolves - so a random vote is more likely to be for an innocent than for a wolf. One could try to calculate a ratio of guilt indicators vs. innocence indicators by considering the ratio of wolves to innocents, but that would be tricky and probably ultimately futile. So I'll estimate: I'll assign one point for voting for a known innocent and subtract two for voting for a known wolf. And I'll start with a base score of 5, so that some of you more rustic folk don't have to deal with negative numbers. I'm going to include those that have been slain, as their scores may be some indicator of how accurate this method is. Using this simple model, the results are:

Aiwendil: 4
Boromir: 4
Dancing Spawn of Ungoliant: 5
Fordim Hedgethistle: 4
Formendacil: 4
Gurthang: 6
Holbytlass: 6
Jack: 5
Kath: 5
Lhunardawen: 5
Mormegil: 7
Rune: 6
The Saucepan Man: 6
WaynetheGoblin: 5
Wilwarin: 6

Okay, maybe that doesn't tell us all that much. It is worth noting, though, that, going by this system, Mormegil stands out as somewhat suspicious.

Of course, the best way to refine this system is to use it over a greater number of votes - we still only two to work with. But we might try to refine it in other ways. For example:

1. A wolf is more likely to have voted for a known wolf at a time when it did not seem likely that known wolf would be lynched. So for such votes, let's only subtract 1 instead of 2.

2. A known innocent is more likely to receive the vote of another innocent if that known innocent has exhibited suspicious behaviour. Here, I am primarily thinking of Gurthang, who (I think) confused us all. So add only half a point for voting for Gurthang.

3. A wolf is unlikely to voice seriou suspicion of an innocent who is widely regarded as innocent. So subtract half a point from anyone who publicly suspected me, Fordim, Boromir, or The Saucepan Man yesterday.

Now we get:

Aiwendil: 3
Boromir: 4
Dancing Spawn of Ungoliant: 5
Fordim Hedgethistle: 3
Formendacil: 4.5
Gurthang: 5.5
Holbytlass: 5.5
Jack: 5
Kath: 5
Lhunardawen: 5
Mormegil: 6.5
Rune: 5.5
The Saucepan Man: 6
WaynetheGoblin: 5
Wilwarin: 5.5

This admittedly haphazard method, then, has me looking at Mormegil and The Saucepan Man right now - two people who seem very reasonable and very helpful, but who would make dangerous wolves.

I suspect that my method will come under some criticism and I'm prepared for that. But I'm also interested in any constructive criticism anyone might have - what other variables could be considered?

Thanks to our harbourmaster, by the way, for compiling a list of votes each day.

12-08-2005, 07:09 PM	#159
Aiwendil Late Istar Join Date: Mar 2001 Posts: 2,224	Mathematical Methods of Wolf Detection Okay, being a fan (as I said earlier) of quantitative methods, I figured I'd attempt a numerical analysis of people's voting records. I know this may sound like a dubious proposition, and I certainly don't claim that it is infallible. But it can't hurt to try. The idea is that you gain suspicion points for having voted for a known innocent and lose them for having voted for a known wolf. Now, it seems to me that voting for a wolf is a stronger sign of innocence than voting for an innocent is of guilt. After all, the innocents outnumber the wolves - so a random vote is more likely to be for an innocent than for a wolf. One could try to calculate a ratio of guilt indicators vs. innocence indicators by considering the ratio of wolves to innocents, but that would be tricky and probably ultimately futile. So I'll estimate: I'll assign one point for voting for a known innocent and subtract two for voting for a known wolf. And I'll start with a base score of 5, so that some of you more rustic folk don't have to deal with negative numbers. I'm going to include those that have been slain, as their scores may be some indicator of how accurate this method is. Using this simple model, the results are: Aiwendil: 4 Boromir: 4 Dancing Spawn of Ungoliant: 5 Fordim Hedgethistle: 4 Formendacil: 4 Gurthang: 6 Holbytlass: 6 Jack: 5 Kath: 5 Lhunardawen: 5 Mormegil: 7 Rune: 6 The Saucepan Man: 6 WaynetheGoblin: 5 Wilwarin: 6 Okay, maybe that doesn't tell us all that much. It is worth noting, though, that, going by this system, Mormegil stands out as somewhat suspicious. Of course, the best way to refine this system is to use it over a greater number of votes - we still only two to work with. But we might try to refine it in other ways. For example: 1. A wolf is more likely to have voted for a known wolf at a time when it did not seem likely that known wolf would be lynched. So for such votes, let's only subtract 1 instead of 2. 2. A known innocent is more likely to receive the vote of another innocent if that known innocent has exhibited suspicious behaviour. Here, I am primarily thinking of Gurthang, who (I think) confused us all. So add only half a point for voting for Gurthang. 3. A wolf is unlikely to voice seriou suspicion of an innocent who is widely regarded as innocent. So subtract half a point from anyone who publicly suspected me, Fordim, Boromir, or The Saucepan Man yesterday. Now we get: Aiwendil: 3 Boromir: 4 Dancing Spawn of Ungoliant: 5 Fordim Hedgethistle: 3 Formendacil: 4.5 Gurthang: 5.5 Holbytlass: 5.5 Jack: 5 Kath: 5 Lhunardawen: 5 Mormegil: 6.5 Rune: 5.5 The Saucepan Man: 6 WaynetheGoblin: 5 Wilwarin: 5.5 This admittedly haphazard method, then, has me looking at Mormegil and The Saucepan Man right now - two people who seem very reasonable and very helpful, but who would make dangerous wolves. I suspect that my method will come under some criticism and I'm prepared for that. But I'm also interested in any constructive criticism anyone might have - what other variables could be considered? Thanks to our harbourmaster, by the way, for compiling a list of votes each day.