Results 1 to 14 of 14

Thread: Attempt at a metagame analysis

  1. #1
    Member

    Join Date

    Aug 2011
    Location

    Italy
    Posts

    780

    Attempt at a metagame analysis

    Hi all,

    In another thread I mentioned that I created some excel files in an attempt to analyse the metagame, so here they are. I welcome any constructive criticism, feedback and opinions. I started this for my own curiosity but decided to share it here so if you think that this could be modified, progressed in any way to become more useful for everyone please share your ideas.

    So let's start. Here's the data I gathered from thecouncil.es
    https://rapidshare.com/files/4107329001/thecouncil.xlsx

    Basically the first sheet is structured like a database for decks and their performances. There is a single entry for deck X for each month. Each deck have their type, subtype fields. These may be off so feel free to comment on those.

    The top4, top8. etc. fields are filled manually by scouring thecouncil.es events lists for each month. I've taken data from events only with 30 players and up. However there are a great deal of gray areas and overlapping architectures so I may have pigeon holed a few and made a few mistakes while categorizing some. I didn't copy-paste the deck names exactly but made my own adjustments since I believed there were some mislabeled decks. However I still may have made some mistakes while categorizing them and also I may have made some typing errors when filling in the numbers (as it was a very long manual process) so accept the fact that there might be a very small percentage of error in the data. It should be OK for the most part though. Feel free to go over all of those events and double check if you have the patience :)

    "In For DTB?" field is calculated automatically. If the deck top8ed during that month that deck's points are taken in for the DTB cutoff calculations. "Points" field is calculated automatically for the number of placements (I have a weight system: 6pts for 1-2, 4pts for top4, 3pts for top8, 1pt for top16). "DTB Cutoff" field is calculated manually for each month. I filter the decks for a single month. Sum up the points of all decks which are "in for DTB" and then calculate the 3%. Usually this gives me decent cutoff with 6-8 decks as DTBs for that month. "DTB?" field is automatically updated if any one deck's points are above the cutoff point. "Num of top 8s" field is also calculated automatically showing the number of top8 placements for that deck in that month. And finally "DTW" field is the deck to watch field which is calculated automatically. What I did here is if the deck is above 22 points but not a DTB it is marked as a DTW. This is something I found fitting according to the results I've seen so far but again feel free to make recommendations.

    The next sheet is the "Complete Analysis" where you can find a pivot table summarizing the whole information. The yearly averages don't mean much for deck types and subtypes such as Combo, Storm etc. but you can use the yearly averages to compare the performance of individual decks. For general deck categories I suggest enabling the "grand totals" from the pivot table options. Also if you intend to carry this further and fill in the first sheet in the following months, make sure to update the data selection of this table to include the new rows and refresh it. You can PM me for your questions.

    The next sheet is the "DTBs" section with a pivot table including only the decks which have been DTB in their lifetime. If a deck doesn't have points for a particular month it means that, that deck wasn't a DTB for that month. But here you can see how the decks come in or fall out from DTB status and how well they fared month after month There is also a barchart below the pivot table but it is getting really crowded as more data comes in so please feel free to make recommendations again.

    The final sheet is the "Top8 stats" which is quite similar to the previous sheet but instead it presents you the number of top8 placements of DTB decks (rather than their points). Again there is a graph below which is getting quite crowded.



    And the second file I have prepared for the SCG open results.
    https://rapidshare.com/files/3894160812/SCG.xlsx

    Since the SCG opens provided a more discrete set of results I decided to do a sliding window calculation to calculate the points(or ratings) of decks. I used a 6 week window. 1st and 2nd weeks get a multiplier of 3, 3rd and 4th weeks get a multiplier of 2 and 5th, 6th weeks get a multiplier of 1. For each of those week the decks get 3pts for top4, 2pts for top8 and 1pt for top16. This is system I made and you can find the scale illustrated in the "Weighing Scale" sheet. In short if a deck place 3rd this week it gets 9pts, if it placed 14th 5 weeks ago it gets 1pt etc.

    I believed this system to be somewhat informative because SCG open results have effects on the metagame and perception of players lasting several weeks. If Spiral Tide plays finals this week people will be looking at the deck for the following couple of weeks so even if Spiral Tide doesn't gets a placement for 2-3 weeks the system will give it diminishing points every week until it gets forgotten so to speak :) I'm not very confident about this system as it was purely experimental for my own satisfactions so please feel free to comment and make you will of it.

    Anyways the breakdown of the file is as follows. Top16 sheet includes the top16 info. week after week. If you intend to follow this up on your own you shouldn't be "inserting" new columns each week but there is a little trick so that the formulas won't get broken. If you are interested PM me and I'll try to explain how it should be done.

    "Last 6 Weeks" sheet includes the points (ratings) breakdown of each deck. This will reflect the results of the last 6 week window. It will get updated every week since the window will be sliding.

    "Long term summary" sheets includes my manual notes. I gave the decks scoring above 10 points a DTB status in this system and followed every deck which scored equal and above 10pts here. If you want to see the previous results you have to manually slide the weeks from the first sheet, note the results from the second sheet etc. Since I did it week after week I already have the results here so trust them. There is also a graph here which is also getting way too cluttered.

    So here it is. Ask, recommend, discuss...

    P.S. I just came in from a 3 hour flight on a Friday evening and posted this. I have a -2 for Vitality and -1 for all other stats right now :) So the grammar and typing might be quite off, I hope it is intelligible enough.

  2. #2

    Re: Attempt at a metagame analysis

    Don't have rapidshare account not going to create one, but sounds interesting. If only i could access. Any chance you can paste to Google Docs?

  3. #3

    Re: Attempt at a metagame analysis

    Quote Originally Posted by jhhdk View Post
    Don't have rapidshare account not going to create one, but sounds interesting. If only i could access.
    You don't need an account to download stuff from rapidshare. (Unless perhaps rapidshare treats some countries different than others, and in yours, there is no free downloading.)

  4. #4
    bruizar
    Guest

    Re: Attempt at a metagame analysis

    I'm fairly busy right now so I can't really add much at the moment, but once the dust settles I'll try to give my input.

  5. #5
    Member

    Join Date

    Aug 2011
    Location

    Italy
    Posts

    780

    Re: Attempt at a metagame analysis

    You should be able to download from rapishare with a free account. Copy-pasting in google docs probably wouldn't work because there are a lot of stuff like pivot tables, conditional coloring, name definitions for addresses etc . :) I'm still trying to figure out if I can attach it to my posts here but I just can't find the "Advanced view" thing while posting messages to give me file attachment options.

  6. #6

    Re: Attempt at a metagame analysis

    Quote Originally Posted by bilb_o View Post
    You should be able to download from rapishare with a free account. Copy-pasting in google docs probably wouldn't work because there are a lot of stuff like pivot tables, conditional coloring, name definitions for addresses etc . :) I'm still trying to figure out if I can attach it to my posts here but I just can't find the "Advanced view" thing while posting messages to give me file attachment options.
    No worries im just stupid and havent used rapid share b4.I was able to access files.

  7. #7
    They call me a slob, but I do my job...
    Cthuloo's Avatar
    Join Date

    Sep 2009
    Location

    Back to the city by the sea, blowin' in the wind, fighting with hordes of retired people
    Posts

    274

    Re: Attempt at a metagame analysis

    Very good work. I will have to look carefully through it to see if there's some considerations I want to make.
    Team Stimato Ezio: You're off the team!

    People demand freedom of speech as a compensation for the freedom of thought which they seldom use.
    -Kierkegaard

  8. #8
    Member

    Join Date

    Aug 2011
    Location

    Italy
    Posts

    780

    Re: Attempt at a metagame analysis

    I just noticed that the graph in DTBs sheet is actually the same with Top8 Stats graph. You just need to delete it and insert a new graph there.

  9. #9
    hai 2 u
    zulander's Avatar
    Join Date

    Apr 2007
    Location

    SoCal - Anaheim
    Posts

    1,688

    Re: Attempt at a metagame analysis

    Also there's a series of this on StarCity by the Hatfields. Here's the most recent one: http://www.starcitygames.com/magic/l...formation.html
    #mtgfinance follow on twitter: @mtgStaples


    Quote Originally Posted by OBFREELY
    You should all immediately fire emails at the DCI requesting the banning of Tarmogoyf and Golgari Grave-Troll.
    Quote Originally Posted by TheInfamousBearAssassin View Post
    I'm pretty sure I'm not compelled to address your non-argument based simply on the fact that you're obviously borderline retarded.
    Team Brown & Team Unicorn. Does that make me a Brown Unicorn?

  10. #10
    Member

    Join Date

    Aug 2011
    Location

    Italy
    Posts

    780

    Re: Attempt at a metagame analysis

    Yes I know but I don't know yet how to use that information to extract something useful (I'm not saying they are useless, just don't know how to incorporate it into something else other than what it is already now). To be honest I think I've burnt-out myself with the effort I put in these within a short amount of time and don't feel like putting any more thought into it for a while :) I'm open for comments and ideas though, I was hoping to get a discussion going with the community. It's fun to discuss such things you know :) I guess most people are concentrating on the weekend ahead so probably after the results of the GP comes in metagame discussions will be rekindled.

  11. #11
    bruizar
    Guest

    Re: Attempt at a metagame analysis

    Can you explain the difference between the variables "1st/2nd", "top 4" "top 8" "top 16" and "num of top8s"?

    I have a problem with "top 4" "top 8" "top 16" and "num of top8s". If I understand it correctly, you are basically making these variables categorical instead of continuous.
    Each 'player result' can only have the value of 0 and 1 if you're making it categorical, instead of just making the variable continuous by using the final standings in that particular order. Right now, you can't tell if the deck got first, second, third or fourth, even if you look at 1st/2nd. Its much better to just keep all the raw data from each player and than analyze it. Right now, I wouldn't be able to calculate a standard deviation because I don't know the exact values of each player. Top four could be third or fourth place, 1st/2nd could be first or second, top 8 could be... etcetera.

    Also, if a deck places 1st/2nd, is it also counted in the variable top4, top8 and top16, or not?

  12. #12
    Member

    Join Date

    Aug 2011
    Location

    Italy
    Posts

    780

    Re: Attempt at a metagame analysis

    "1st/2nd" field contains the number of 1st/2nd placements the deck made during that month. You are right, I didn't differentiate between the 1st and 2nd. I'm assigning the same points to both finalists. We don't know if they split or not :)

    "top 4" includes the number of placements of 3rd and 4th. "top 8" includes 5,6,7,8 and "top 16" includes 9-16.

    Num. of top8s is the sum of everything between 1-8. So I'm basically adding the "1st/2nd" + "top 4" + "top 8"

    I wish we spoke sooner so that I would actually have noted the exact placements of each deck. I'm pigeonholing them into categories now. The guys at thecouncil.es have the actual raw data, I don't know how their DB is structured so they may be able to query the exact placements of each deck.

  13. #13
    bruizar
    Guest

    Re: Attempt at a metagame analysis

    You can still do some analysis with categorical data (chi square and perhaps others I don't know about.), but its much better to leave them as continuous variables if that's what they really are.

    Also, the mean isn't the be-all-end-all statistic. The mean is weak to outliers.

    2+2+2+2+2+20=30 with a mean of 5. (result disguises outlier)
    5+5+5+5+5+5=30 with a mean of 5. (same mean but a very different sample)

    The median looks at the middle number and is not weak to outliers.

    Another example to point out how incomplete the mean is, is from the book Fooled by Randomness by Nassim Taleb, if I recall correctly.

    Two men walk on the highway, between two driving lanes. Each x describes a footstep and the amount describes the inches he deviates from the straight line he is walking on. If he deviates from the line more than 15 inches, he is hit by a car and dies. Consider the two scenarios.

    Scenario 1 (not drunk):
    4 6 6 6 6 6 8 -> mean is 6, no values above 15

    Scenario 2 (drunk):
    2 2 4 4 4 6 20 -> mean is 6, one outlier above 15 (20).

    In Scenario 1, the person is safe, in scenario 2, the person gets hit by a car. The mean disguises this life or death outlier. You can imagine how that would apply to magic tournament results too.

  14. #14
    Member

    Join Date

    Feb 2012
    Location

    England
    Posts

    17

    Re: Attempt at a metagame analysis

    I wrote an article on the general theory behind the British metagame, just waiting for the editor to let it go live on the website.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)