195 replies to this topic

[istaril]

I'm not sure how much work this might take, but I'd like too see agenda percentages of making the cut for each house. For example, differentiate Martell KotR from Bara KotR, and let me see that x% of lanni decks to make the cut were no agenda, y% pbtt, z% house of dreams, etc.

To answer that... it's a fair bit of work, but also fairly useful data. I honestly wasn't sure we'd reach a sufficient number of datapoints to make that level of resolution meaningful, but as we're currently sitting at 548 datapoints (124 for the cut) 3 weeks into the season, I could see us reaching that threshold level. For instance, that's only 15 Greyjoy decks so such stats wouldn't be very useful for them. If people keep submitting and we reach a level where I feel that data is worth the effort, I'll put it in this version. If not, I'll built it in to the version for the next FAQ.

Also, updated the metadata to include data from the 7 tournaments submitted over the weekend.


To address some other feedback I've received: the idea of "nesting" sheets so that all tournaments from one area (say, a city that runs regular tournaments) get their own meta/data is *fantastic*. I just have no idea how to do it at the moment without creating new documents.

[Jensen22]

That's the beauty of it though. Once the data is collected you can do whatever you like with it. If people keep up the data entry then this list will prove more and more useful.

It was getting everyone to post the data that was the hard part. I am glad it is going over so well...

[trooper6]

May I ask for some stats clarification since I'm a humanities person and you know...stats aren't my strong suit?

So 18.5% of all the decks so far have been Lannister.
Yet, 24.6% of all the decks that make the cut have been Lannister.
It is this data that leads to Lannister having the highest qualification ratio, right? What this says is that Lannisters make the cut more than they should, right?
But then, that last chart says that only 10% of the winners have been Lannister.

So...that what do those three sets of percentages mean for Lannisters?
The Targaryens seem consistent in all three charts: @20% play, @20% make the cut, @20% win.
The Baratheons seem are @17% play and 17% make the cut...but then a whopping 35% win.

I can make up ideas about what I think these percentages mean...but I'd rather a person who understands the statistics explain an analysis of these numbers to me. Because I'd hate to come to poor conclusions because there is some secret stat thing I don't know.

So...stat people? How do you analyze these stats and what conclusions are you drawing?

[NorthMaester]

May I ask for some stats clarification since I'm a humanities person and you know...stats aren't my strong suit?

So 18.5% of all the decks so far have been Lannister.
Yet, 24.6% of all the decks that make the cut have been Lannister.
It is this data that leads to Lannister having the highest qualification ratio, right? What this says is that Lannisters make the cut more than they should, right?
But then, that last chart says that only 10% of the winners have been Lannister.

So...that what do those three sets of percentages mean for Lannisters?
The Targaryens seem consistent in all three charts: @20% play, @20% make the cut, @20% win.
The Baratheons seem are @17% play and 17% make the cut...but then a whopping 35% win.

I can make up ideas about what I think these percentages mean...but I'd rather a person who understands the statistics explain an analysis of these numbers to me. Because I'd hate to come to poor conclusions because there is some secret stat thing I don't know.

So...stat people? How do you analyze these stats and what conclusions are you drawing?

That's the beauty of statistics; you can pretty much use them the way you want to.

First off: representation. We see that Targaryen is quite popular, and that not that many people believe Greyjoy is the best way to go.
Then we look at the cut. The popular Targs are quite even here, 19 % vs. 20 % overall. This means that they are performing as they "should", considering their overall representation. Also Martells and Greyjoys are as they "should" be, considering nothing but the overall representation. The diversity comes from the houses Stark and Lannister. Very few Stark players make the cut, while a lot of the Lannister players do so.

I believe we can trace this to two things: what players believe is the best house, and what actually is the stronger house. A decent amount of the player base seem to think House Stark is a good choice (this may also be due to personal preferences), but not that many of them make the cut, probably meaning that the house is weaker (or the players choosing that house are) than players think. The opposite goes for Lannister.

When you look at the winners ratios, keep in mind the amount of data points here is significantly smaller. Meaning the basis to draw conclusions is weaker. But still a couple of things is interesting, as you pointed out: House Baratheon has an enormous percentage of winners! What does that mean? There may of course be several reasons. The players playing House Baratheon are better, more capable of winning the games also after the cut; Baratheon has good match-ups against typical oppositions in the cut; or simply that they have some decks within the house that are simply stronger than most, but that not all Baratheon players play said deck.

Another interesting thing about the winners stats is that Lannister bumps down to 10 % from their high 24,6 % in the cut. That may be because they have a lot of decks that are good, but maybe they are not quite the strongest of them all.

Note again that I think the most important stat is percentage making the cut, and the qualification ratios between the houses.

These are all my takes on the house stats, and not any kind of absolute truth. Hope it helped you a little bit at least.

[istaril]

Trooper, you seem to have the gist of it. Lanni's qualification ratio can be seen in the third Figure over on the top row, and is in fact about 1.3.

NorthMaester's explanation and interpretation is excellent. I, too, tend to disregard the "win" information because of the small data size (548 datapoints overall, 124 for the cut, 20 for winners), although we can hope it'll grow large enough to make meaningful inferences. I'd assume the high representation of Baratheon in the win is due to one of the winning decks (Bara KotHH) being a very demanding deck, hard to play in the hands of the less-experienced. Most people won't play it, and those that do, most won't make the cut - but those good enough to pilot it to the cut do well in the cut, primarily because it does quite well against both KotHH mirrors and against Lannister, two very common matches in the cut.

We have to be careful answering questions of "significance" in looking at any of these data, because I've added no means to determine that. For instance, at a glance I'm almost certain the Lannister overrepresentation *is* significant, and that the overrepresentation of, say, Noble Cause or Brotherhood without Banners is not. However, I have not yet put in place a statistical test that will impartially asign confidence intervals/thresholds of significance etc.

[JCWamma]

All I think we can conclusively say from the data so far is that only a true genius can pilot a Maester deck to victory.

[trooper6]

We have to be careful answering questions of "significance" in looking at any of these data, because I've added no means to determine that. For instance, at a glance I'm almost certain the Lannister overrepresentation *is* significant, and that the overrepresentation of, say, Noble Cause or Brotherhood without Banners is not. However, I have not yet put in place a statistical test that will impartially asign confidence intervals/thresholds of significance etc.

If Lannister overrepresentation in the cut is significant, wouldn't their then big drop off when it comes to the winner circle also be significant?

NorthMaester, thank you for your analysis! It was really interesting. I like how you didn't attribute the states to only if the deck is overpowered or not...but also to the possibility that the players who choose Deck X or Y and make the cut might be better or worse, or that they might have better or worse match ups. Etc.

Also, would it be possible to find out percentages of Agenda used by house: overall, in the cut, in the winners?
I mean, it would be interesting to know if all the Bara winners were noble cause, or if the majority of Lanni cut makers were no agenda, etc.

[Bomb]

Is it me, or does the House Stark portion of the pie chart have the same colored text as the background? I can't read anything from that part of the pie chart.

[istaril]

Also, would it be possible to find out percentages of Agenda used by house: overall, in the cut, in the winners?
I mean, it would be interesting to know if all the Bara winners were noble cause, or if the majority of Lanni cut makers were no agenda, etc.

Yes - this information is available (you can go through individual tournaments and see on each sheet), and there's a way for me to pull it out and report it, but it's a bit of extra work on my end for something that isn't quite yet worthwhile (For instance, telling people that 100% of Greyjoy winners were running black sails isn't very meaningful). Once the data gets more robust, I'll see about getting the House/Agenda pairings for each house overall and in cut (and winner).

I think it'll be a great addition, but because of the considerable extra work involved in pulling the paired info from the sheets rather than just treating House and agenda independently, it may not happen before the next FAQ causes me to re-release the sheet.

[boothby]

Just added info for the Glen Burnie, MD tourney today. Let me know if I did anything wrong

[istaril]

Looks great!

Updated the Metadata to include the three tournaments reported yesterday.

[istaril]

Important announcement: Google was telling me the spreadsheet was getting too large and complex. We have too much data! What a great problem to have to deal with.

Fortunately, the new "google spreadsheets" launched a few weeks ago will allow us to grow the sheet, so I've migrated over to the new link. Please update your book marks (Original post is also updated)

https://docs.google....#gid=1031032158

[DJJ]

I presume this is the way Google creates charts rather than something that's configurable but in my head the key for the pie charts should list labels in percentage order rather than alphabetically. As there are two oranges on the agenda I initially looked at the large percentage that were orange, looked over at the key to see that Alliance was orange and had a momentary - "WTF?". Realised that couldn't possibly be right and then found out the slightly darker orange meant No Agenda...

[istaril]

This I can do fairly easily, assuming the spreadsheets don't dislike vlookup as much as the old version did. I'll try to have that up tonight.

[istaril]

Ok, I've done two things

1) Fixed the input sheets, which were given Data Validation errors since I migrated over to the new sheets.

2) Made it so on the metadata sheet, Agenda and Restricted pie charts are sorted by the overall presentation (largest to smallest). Note that that means the "cut" pie chart is sorted according to overall representation, which means it's not necessarily largest to smallest for the cut. The "winner" ones have not yet been updated accordingly.

The reason for this is that if I sort each one individually, the colours are assigned arbitrarily (or I can put them all in by hand, but no thanks). In sorting them both according to the overall representation, I can keep the colour pattern constant between the two graphs.

[istaril]

The metadata has been updated to take into account the Louisville, Moonboy, Manchester(CT), Providence, Gatineau and Padis and Sydney Store Championships.

Greyjoy, to my considerable surprise, is convincingly on the rise.

[mdc273]

Wow. I'm shocked by the lack of handles and last names, though I guess that doesn't matter for the metadata.

I only got like 3 tournaments where I could actually add the data to my rankings sheet. Surprise! They were all in the northeast (and wherever dobbler played, handles ftw!)

[agktmte]

Dobbler played in tulsa, so I made lots of edits to ensure handles (actually I had names and handles but someone removed the names).

[Bomb]

They didn't want to use last names for privacy purposes.

[JohnyNFullEffect]

Just added Michigan SC on the Input 1 sheet!