Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

If you know a statistician who likes 40k and is willing to be interviewed in a podcast, could you please send me a PM?
In particular I am looking for someone who is willing to discuss the stats from Torrent of Fire. Namely the following.Most Popular Armies ------------------------------------------------------------------------------ Army Allies Players Percent Win Percent ------------------------------------------------------------------------------ Eldar -none- 371 8.08% 59.84% Chaos Daemons -none- 351 7.64% 58.40% Necrons -none- 276 6.01% 52.17% Tau Empire -none- 253 5.51% 58.10% Chaos Daemons CSM 230 5.01% 63.48% Tau Empire Tau Empire 200 4.35% 63.00% Eldar Tau Empire 191 4.16% 71.20% Tyranids -none- 188 4.09% 48.94% Grey Knights -none- 172 3.74% 42.44% Space Marines -none- 165 3.59% 43.03% How relevant are these numbers exactly? Is a 71.20% win ratio as 'over the top' as it seems? Is a 58.40% winning ratio that much better? I know in voting a margin of 16.8% is considered huge -- but does that apply to 40k wins/losses. Is it unusual that the most popular armies do not correlate with the top winning armies?Top Winning Armies --------------------------------------------------------- Army Allies Players Win Percent --------------------------------------------------------- Eldar Tau Empire 191 71.20% Tau Empire Eldar 160 68.13% Eldar Dark Eldar 140 63.57% Chaos Daemons CSM 230 63.48% Tau Empire Tau Empire 200 63.00% Necrons Tau Empire 52 61.54% Eldar -none- 371 59.84% Tau Empire Space Marines 67 59.70% Grey Knights Tau Empire 44 59.09% Chaos Daemons -none- 351 58.40%Thanks for taking the time to review this.

In order to make any statistical analysis you have to have something to compare them with. At least if you want someone to say something about the relevance of the numbers.

Normally you look if the values are significant which only says how high the probability is that the values you are seeing were created by chance and therefor dont tell us anything. Thats really almost the only thing statistics are able to say in such a scenario.

Since there are always new books and models that influence the win-lose ratios its really hard to tell anything about them that isnt redundant.

You could assume that in a perfectly balanced scenario every army wins as much as it loses (considering all players with the same skill). If you compare them you could more than certainly determine that the eldar win loss ratio isnt created by chance but that still doesnt tell you which variable (opness, player skill, matchups etc) was how relevant to get to the values you are seeing.

To say something concrete. The propability that the 70+% wins of the eldar are just created by chance doesnt seem to be high (without having calculated the statistical significance). So they are most likely significant (maybe even highly significant) So there seems to be another variable (maybe opness) that took it to 20% more wins than expected.

Before I can comments on any of these "statistics", I need to know a little more about them.

Did you really have 371 Eldar players, or did you have 61 players each play 5 or 6 games with a handful dropping out?

"Is it unusual that the most popular armies do not correlate with the top winning armies?" Impossible to answer without matching data from other major tournaments, I'm afraid.

I'd want to see the rest of your numbers that included the "less popular" armies in attendance, like Space Wolves, Blood Angels, and so on.

Thanks for the feedback and thoughts.

 Mywik wrote:
In order to make any statistical analysis you have to have something to compare them with.

The only comparison we have are the ToF values for other armies. There are 81 different combinations of armies recorded in the ToF database, over 4,594 games.

 Mywik wrote:
Normally you look if the values are significant which only says how high the probability is that the values you are seeing were created by chance and therefor dont tell us anything. Thats really almost the only thing statistics are able to say in such a scenario.

I'm sorry, I don't understand. Are you saying that "Normally your seeing how high the probables distribution is created by chance and not some other factor". Therefore what is normally considered a statistical analysis is not really useful in this context.

 Mywik wrote:
You could assume that in a perfectly balanced scenario every army wins as much as it loses (considering all players with the same skill). If you compare them you could more than certainly determine that the eldar win loss ratio isnt created by chance but that still doesnt tell you which variable (opness, player skill, matchups etc) was how relevant to get to the values you are seeing.

I think that you can assume equal player skill. This is because over the course of a 6 round GT, players will tend to be matched up evenly in brackets. Nothing is perfect, but after the first two rounds your less likely to see players like Alex Finnell clobbering some local guy playing SoB.

 Mywik wrote:
To say something concrete. The propability that the 70+% wins of the eldar are just created by chance doesnt seem to be high (without having calculated the statistical significance). So they are most likely significant (maybe even highly significant) So there seems to be another variable (maybe opness) that took it to 20% more wins than expected.

What I find is key is that its not Eldar that are enjoying a 70+% win ratio. Its Eldar and Tau. Eldar without allies are at a 59.84% win ratio (#7 on the chart)

That seems to indicate to me that the Tau allies are what make the Eldar army better. I don't think its that they are more popular as Eldar/Tau is the 7th most popular army.

 kronk wrote:
I'd want to see the rest of your numbers that included the "less popular" armies in attendance, like Space Wolves, Blood Angels, and so on.

Have you not heard of the service Torrent of Fire? A lot of GTs are using it for their scoring system. It handles paring and allows players to see their next matchups on their smart phone. It also allows players to enter their scores on an ipad or smart phone.

This includes major GTs like NOVA, 11th company GT, Da Boys, etc..
This service also records all win/loss ratios for every army type. You can find this information freely off their site.
Its a pretty cool thing that the 40k community has now, and I think will prove a very valuable tool for the future of 40k competitive playing.

 labmouse42 wrote:

 Mywik wrote:
Normally you look if the values are significant which only says how high the probability is that the values you are seeing were created by chance and therefor dont tell us anything. Thats really almost the only thing statistics are able to say in such a scenario.

I'm sorry, I don't understand. Are you saying that "Normally your seeing how high the probables distribution is created by chance and not some other factor". Therefore what is normally considered a statistical analysis is not really useful in this context.

To clarify my sentence.

When you look at statistical data you normally build 3 hypothesis.

The null-hypothesis normally is something like "with tau allies the eldar are as powerful as without".
Hypothesis1 is "the eldar are more powerful with tau as their allies"
h2 would be "the eldar are less powerful with tau as their allies"

You now have your data that the hypothesis are tested against. What i assumed (without actually calculating since there is a lot missing to actually do that) is that the values ARE statistically significant which means that the null-hypothesis can be rejected. This means the higher win-loss ratio wouldnt be created by chance and is therefor created by other variables.
Problem i see with such an analysis is that two tau-eldar armies arent the same normally. Which makes determining the factors that lead to such a win-loss ratio harder. You can always speculate but you dont need statistics to do that

Learned statistics in german not english so maybe some technical terms arent translated correctly ... sorry for that.

Needs more data for useful analysis.

If you can get hold of an actual dataset, and not just aggregated results, I'll tell you everything you want to know.

As per above get me the raw data and I can shed some light to it, after my exams though, so not for a couple days.

http://app.torrentoffire.com/#/stats/armies

This is the site that holds all of the data. Inside you can view individual armies. How they faced vs specific opponents, and any allies involved. If you click all 3 check boxes you can get 1565 different sets of values (army/ally vs opponent/ally) That's as granular you can get.
I copied the results from the web page into my handy excel spread sheet then ran some numbers on it.

ToF does not record army composition. We don't know if the Tau/Eldar armies are all fire warriors or Riptides with buffmanders/farseers. I think we can infer that the latter would be more common than the prior based upon the meta today.

So, would any of you fine gentlemen be willing to be interviewed on Skype? If not, no worries! What you have said here I will compile into my next "Simhammer" segment of the 11th company podcast. Again -- I appreciate your insight.

 Mywik wrote:
The null-hypothesis normally is something like "with tau allies the eldar are as powerful as without".
Hypothesis1 is "the eldar are more powerful with tau as their allies"
h2 would be "the eldar are less powerful with tau as their allies"

This concept makes sense. Thank you for explaining to me.

I'm just getting auto-directed to a a site that tells me to join the revolution.

Apparently I must "prove that I am a warrior."

You may need to create an account. That should just take a minute to do.

They send a email notification talking about their events and stuff to you, but you can probably unsubscribe from it.
They also offer a premium service for a few bucks a month that gives you other datapoints -- but for this exercise I think that the free service will suffice.

Ok. Did it.

Data is still not good, though, as it doesn't allow me to see what I want to see and use it the way I want to use it.

Basically, all I can get out of it is what you have already found; how many percent of its games Eldar (or any other army) armies have won.

With the raw data I would be able to tell you how many games Eldar armies have won against non-Eldar armies and the significance of these results.

If you wanted to get even an inkling if this data is meaningful you would need the variance. If the number is 71+/-20% then it probably isn't meaningful even if you assume player skill, luck, and other untested factors were not important.

If the value is 71+/-1% then there is a decent chance it is meaningful even though your assumptions may not be valid.

It may not be a derivative relationship then but could in fact be reversed in cause in effect, ie CWE+Tau wins more not because they are CWE+Tau but because the player wins often (are good at playing) they choose CWE+Tau.

I am a biochemist/spectroscopist so I am not a statistician but what you are looking for is a causal relationship. The problem is we have no way of judging the validity of our assumptions. A measurement of variance would at least give use some sort of data on how much this data deviates from expectations. Can you get the spreed sheet of the individual W/L/D ratios? I could help you generate variance measurements.

Thud wrote:
Ok. Did it.

Data is still not good, though, as it doesn't allow me to see what I want to see and use it the way I want to use it.

Basically, all I can get out of it is what you have already found; how many percent of its games Eldar (or any other army) armies have won.

With the raw data I would be able to tell you how many games Eldar armies have won against non-Eldar armies and the significance of these results.

Chip from Torrent of Fire here. Here are the statistics on Eldar vs non-Eldar, current as of today:

Games: 803
Wins: 471
Losses: 312
Draws: 20

Note that this includes non-Eldar primary with Eldar allies. If we include Eldar as allies on the winning side, and remove Eldar as allies on the losing side, we have:

Combinations: 223
Games: 917
Wins: 552
Losses: 344

Hope that helps. Happy to answer any other questions about the data as well.

You will most likely need more data points to lower the standard deviations of the data. It's why using stats for die rolls is nice, but ultimately not a useful measurement.

Hey chip,

I don't suppose you would make the excel with the W/L/D of the individual lists available? That would allow determination of variance which could give some indication of if there is any correlation at all (even if not cause and effect).

Thanks for your efforts in running a good GT and the data.