I am the resident expert on the TrueSkill system.

#101lderivedx(Topic Creator)Posted 4/1/2013 2:27:24 PM
From: That1Guyy | #100
My mistake then for not clarifying my assumption that only highly probable real world scenarios should be considered.


Suppose there exists a counterexample, regardless of whether or not it's "realistic." We know there 1) are counterexamples and 2) know nothing about the plausibility of all "realistic" scenarios.

If we can show there exist no counterexamples at all, then we know all "realistic" scenarios are plausible.
---
I bet you're one of those people who ruin my mainslaying by going -15 so stop saying that other people ruin gaming.-im r awful gt:i derive dx
#102lderivedx(Topic Creator)Posted 4/1/2013 3:11:38 PM(edited)
Also, people don't really start out at rank 1. They start out higher, but your displayed rank is a conservative estimate of what the system thinks your actual skill is.

From: That1Guyy | #078
It is mathematically impossible for the average to be over 25 without removing low ranked players from consideration.


http://atom.research.microsoft.com/trueskill/rankcalculator.aspx

Do a head to head using standard values. Assume your displayed rank is mu-3sigma, rounded up. Note this gives everyone rank 1 at the start. A and B alternate winning until sigma stops changing (0.8something). Without loss of generality, call whichever of the two with the highest mu at the end A. Then new player C constantly wins against A. If at any point a win by C would cause A's mu-3sigma to fall below 22, then switch to C constantly beating B (this shouldn't happen, but just in case). Eventually the sum of the three (mu-3sigma) will be greater than 75, meaning the average is greater than 25.
---
I was a Captain Grade 3 back in Halo 3 and every other incredibly good player out there who didnt hack was at least an officer-Lawfish66
#103FinaliizePosted 4/1/2013 2:33:40 PM
lol at this entire thread. these posts today surely have to be april fools?
#104That1GuyyPosted 4/1/2013 8:24:06 PM
lderivedx posted...
Suppose there exists a counterexample, regardless of whether or not it's "realistic." We know there 1) are counterexamples and 2) know nothing about the plausibility of all "realistic" scenarios.

If we can show there exist no counterexamples at all, then we know all "realistic" scenarios are plausible.


And if it isn't necessary to know if all realistic scenarios are plausible? You're right that under more scrutiny than a casual analysis, I'm wrong. But I still fail to see why I should care in the real world scenario of me being an Xbox or PC developer, why I should care about any scenario other than the most likely ones. Given a situation like you've presented, I literally could not care less about the feasibility of my ranking system. I would be focusing all of my efforts on understanding why my gameplay mechanics have caused the population to drop to three people.
---
Perfection. Where everyone fails.
If you expect nothing you can be happy for everything.
#105lderivedx(Topic Creator)Posted 4/1/2013 8:44:51 PM
Your posts confuse the hell out of me. They're so non sequitur. You claim X is impossible, get proven wrong, then start talking about the feasibility of the ranking system, something never mentioned up until your post.
---
I've been known to write some pretty good stuff from time to time. Just ask the mods. They probably know me like a brother by now. -MIxeD
#106That1GuyyPosted 4/1/2013 11:13:29 PM
There was a context to this conversation that I framed all of my posts around. Even my first post about the average trueskill of players. I falsely assumed that you were arguing within the same context as me. That context being the application of Trueskill to how it would be used in Halo.

Perhaps I worded my previous post wrong. I'm still talking about the same thing I was when this started. What I had intended to say was that given the situation you provided, I don't care if the average trueskill can be above 25, because, again, within the context I assumed, by the time Halo has playlist populations of such low values, the resulting statistics of the ranking system are of no interest and thus to be discarded. I'm only interested in how the system behaves when there are still thousands of players playing. And given thousands of games with a strict search system to prevent repeat matching of players, I fail to see how an average trueskill of over 25 would ever happen.

You thought I was talking about the general case, I thought you were talking about the case of Halo as it is right now. Now, if you think I'm wrong about the possibility of average trueskill for this specific case as well, by all means, please enlighten me.
---
Perfection. Where everyone fails.
If you expect nothing you can be happy for everything.
#107lderivedx(Topic Creator)Posted 4/2/2013 1:04:39 AM(edited)
Since you have done a horrendous job of describing the situation you're interested in, would you explain how

From: That1Guyy | #078
The theoretical average is still 25.


and further

It is mathematically impossible for the average to be over 25 without removing low ranked players from consideration.
---
#1 Deriver on H4GB
GT: i derive dx
#108RedheadedPantsPosted 4/2/2013 2:29:18 AM
Is True Skill the thing Halo 2 used?

That one was good.
---
I'm no Superman
GT - Thumbserelly NOT RedheadedPants
#109VauntXPosted 4/2/2013 3:28:01 AM
I don't think so.
#110lderivedx(Topic Creator)Posted 4/2/2013 5:28:39 AM
From: RedheadedPants | #108
Is True Skill the thing Halo 2 used?


No.
---
I bet you're one of those people who ruin my mainslaying by going -15 so stop saying that other people ruin gaming.-im r awful gt:i derive dx