Match result calculation theories

[antiaircraft]

I've noticed that a lot of people seem to think that TNT rigs the results of the Altador Cup, but I've always had the impression that fabricating convincing results would be a lot harder than coding an algorithm to decide match outcomes. Until recently however, I haven't actually been able to confirm this, but I started studying statistics this year and it seems there are plenty of simple methods that could be applied to generate fair AC results without much trouble at all.

 

For one, TNT might be recording the number of valid plays (scores above the minimum threshold) submitted by members of each team for each game, and using that as the sample data to generate a confidence interval. By using this data as a sample (instead of a population), the effects of freeloaders can be largely discarded, and a 5% trim (approximate proportion, I can't make a better guess since I can't see the data) on the bottom part of the sample would probably take care of the rest that only submitted one or two scores.

 

The mean and standard deviation could be derived from the remaining data and used to generate a 90% (approximate percentage again) T confidence interval for each team's mean performance in each game. Then the confidence intervals could be matched up against each other: if they overlap, the result would be a draw, otherwise the higher interval would belong to the winning team. With this method, results could be determined for each game in each match.

 

This is the fairest statistical analysis I could think of off the top of my head - it's only slightly more complicated than doing a basic point estimate, and a lot less arbitrary. :yes: That said, my knowledge of statistics barely classifies as 'basic', so I could be doing completely the wrong thing. :P Are there any skilled statisticians on TDN who can comment on this? Or does anybody else have any theories as to how TNT calculates AC results?

[Secre]

Lol, I didn't understand a word of that, my maths is awful! I just assumed they have a program which keeps track of all scores above minimum requirements and so it was kind of highest overall score wins or something along those lines! Because even people only playing once or twice a day help towards the overall effort even if not by a lot.

 

Granted this is probably a very stupid comment, but I don't do maths so I over-simplify everything to allow my brain to comprehend it!

[antiaircraft]

Because even people only playing once or twice a day help towards the overall effort even if not by a lot.

Hmm... you actually have a good point there. The 5% bottom trim probably isn't the best idea, unless they try and normalise the data a bit with a standard 10% trim (5% on both ends).

[nikh]

I've noticed that a lot of people seem to think that TNT rigs the results of the Altador Cup, but I've always had the impression that fabricating convincing results would be a lot harder than coding an algorithm to decide match outcomes. Until recently however, I haven't actually been able to confirm this, but I started studying statistics this year and it seems there are plenty of simple methods that could be applied to generate fair AC results without much trouble at all.

 

For one, TNT might be recording the number of valid plays (scores above the minimum threshold) submitted by members of each team for each game, and using that as the sample data to generate a confidence interval. By using this data as a sample (instead of a population), the effects of freeloaders can be largely discarded, and a 5% trim (approximate proportion, I can't make a better guess since I can't see the data) on the bottom part of the sample would probably take care of the rest that only submitted one or two scores.

 

The mean and standard deviation could be derived from the remaining data and used to generate a 90% (approximate percentage again) T confidence interval for each team's mean performance in each game. Then the confidence intervals could be matched up against each other: if they overlap, the result would be a draw, otherwise the higher interval would belong to the winning team. With this method, results could be determined for each game in each match.

 

This is the fairest statistical analysis I could think of off the top of my head - it's only slightly more complicated than doing a basic point estimate, and a lot less arbitrary. :yes: That said, my knowledge of statistics barely classifies as 'basic', so I could be doing completely the wrong thing. :P Are there any skilled statisticians on TDN who can comment on this? Or does anybody else have any theories as to how TNT calculates AC results?

 

 

Totally mathematical! I think it's of how many players who join a team and their total team's win percantage! :woot: :laughingsmiley:

[tempdisplay]

I thought it was amount of scores divided by those who played. You know, mean, average. Didn't understand what you said though

[Tyler.]

I think my mind just imploded. Er, nice job finding out how it works

[Divya Bean]

Nice! It makes sense, however there are multiple ways of doing this. Simpler ways too. TNT might be doing what Blue of Maraqua said, find out the average, or TNT might be just adding up all of the scores and then ordering them from first to last! Its that easy!! I sound like an advertiser. :laughingsmiley:

[-Ryan]

Yes, I could easily see TNT coding it like that. Nice work, as usual. :)

 

 

[antiaircraft]

I thought it was amount of scores divided by those who played. You know, mean, average. Didn't understand what you said though

Nice! It makes sense, however there are multiple ways of doing this. Simpler ways too. TNT might be doing what Blue of Maraqua said, find out the average, or TNT might be just adding up all of the scores and then ordering them from first to last! Its that easy!! I sound like an advertiser. :laughingsmiley:

Well yes, they might be doing this, but as I mentioned in my original post doing a simple mean point estimate is probably a bit arbitrary, especially when it comes to determining draw margins. But really, I have no idea. xD

 

I think my mind just imploded. Er, nice job finding out how it works

This is just a wild guess I pulled out of my recently acquired statistics knowledge, I don't have any clue as to how close it is to TNT's algorithm. :P

[jumpingbeans]

I personally think that it's magic. All magic :yes:

To be honest, I don't have a clue.

But it's obviously magic. :yes:

[Benji]

Hm. That's a nice theory. Math foundation is solid, but I don't actually think that TNT does the whole average thing- simply because they explicitly say that all games help, and maybe I'm naive but I don't think they'd blatantly lie like that, it's probably some sort of addition, so the more you play the better your team does. Because that's what it's all about. Playing as much as you can- spending as much time on the site as you can (which is what free sites like this are all about. ;) )

[antiaircraft]

Hm. That's a nice theory. Math foundation is solid, but I don't actually think that TNT does the whole average thing- simply because they explicitly say that all games help, and maybe I'm naive but I don't think they'd blatantly lie like that, it's probably some sort of addition, so the more you play the better your team does. Because that's what it's all about. Playing as much as you can- spending as much time on the site as you can (which is what free sites like this are all about. ;) )

Hmm... why do I get the impression I'm talking to a mathematics professor? :P

 

I wonder if a similar statistical analysis could be applied to a summation of submitted scores? That would probably be more in the spirit of what TNT's suggested. :yes:

[Mireia]

I had a theory, but I can't express it here because my english is not good enough :laughingsmiley:

But let me try it... I thought TNT took all the scores submited and calculate the average score for every team. The highest one wins. That would be easier to calculate why do we have so many draws...

 

(If this is what you said, I'm sorry. I disn't understand it completely :crying: )

[antiaircraft]

I had a theory, but I can't express it here because my english is not good enough :laughingsmiley:

But let me try it... I thought TNT took all the scores submited and calculate the average score for every team. The highest one wins. That would be easier to calculate why do we have so many draws...

 

(If this is what you said, I'm sorry. I disn't understand it completely :crying: )

That's fine. I'm not very good at explaining stuff. :P

 

And what you described is essentially a basic point estimate of the mean, which I mentioned briefly in my initial post. :yes: