MTG Salvation

Quote from KingChewie

- Do you consider an update of the first post of the thread or any answer to a specific thread gives it "a new life" for your scripts?

I currently scan the first page for all dates in the format mm-dd-yyyy and I just take the last one. Not a 100% correct, but it'll do for now.

Quote from ApocryphaEffect

I'd just like to point out that this isn't a metagame breakdown.

Would you care to elaborate on that?

I am guessing a lot folks will be like me and say, "I want to take the most popular cards and build an awesome deck."

My next response was, "I want make an awesome deck without playing any of these cards."

Quote from Espuma

Can you perhaps look at the trends in colors? I mean looking at all the cards of a color (so just those 5 total numbers), and see the relative ranking between them. If that changes a lot over time, maybe make a nice graph?

Here is 8 months of data. The results are not that interesting.

Quote from robtheman

"I want make an awesome deck without playing any of these cards."

Exactly!

As a new commander player I have to give you a huge thanks for helping me discover what cards I need to be both aware of and need to be looking at getting

Quote from Espuma

What scale is that y-axis, by the way?

x-axis = months (going from may 2012 to january 2013, skipping august 2012)
y-axis = number of cards of that color in the top 200

Other request of stuff I could look into are always welcome.

Would it be possible to display the most popular cards by type within color combinations, out to 25-50 places? IE the most popular creatures/enchantments/artifacts/lands in mono-green/Grixis/Golgari? Or would there be a ton of extra scripting/computing time involved?

No, but some of those lists have fewer than 30 decklist entries. Dividing them any further would make those list even less statistically meaningful. And as I do not have enough room to include those lists in the first post, I will not do so. But I am willing to take requests...

Do you have any on demand requests? I'd be happy to post them here.

Request:

Your analysis starts to break down color appearances, but i would like to request a different summary:

what % of all the decks your script picks up have each color in them? I.e (making up numbers here); 23% (235) of EDH decks have R, 67% (576) have B.

Essentially what is the most popular color in EDH?

The reason for this request is because i would be curious to see the various trends as a function of color popularity. Example: Demonic Tutor appears to the be the most common colored spell, but is B the most common color? (they are not necessarily mutually inclusive)

Quote from Shmo40

what % of all the decks your script picks up have each color in them?

Already did that once, the total number of decks I count is 1058:

Note that I count the colors for each commander (I did not count the actual color distribution in each deck, e.g. Oona counts as UB although there exist mono black variants of that deck)

If I weight the colors for each deck (a grixis deck would only count 1/3 towards U, B and R while a mono red deck would count as 1 towards R) I get the following results:

G 21.6%
U 21.4%
B 21.3%
W 18.2%
R 16.7%
X 0.6%

If I don' weight the colors (which I think I should, so these results might not be that interesting) I get the following results in number of decks:

G 487
B 474
U 465
W 445
R 382
X 7

I left out the percentages for this result as they would total over 100%.

Black winning over Blue here might mean it is more of a support color? I don't actually know... the results are close anyway

Quote from scoeri

Already did that once, the total number of decks I count is 1058:
[Snip]

Nice! (you should add this to the OP )

you did this the same way i would have. I actually find the 2nd group of data points interesting because when you align it with the top 200 cards and how they rank in priority it doesn't necessarily align with top colors.

for example, Demonic Tutor may be the most common colored spell but it's not the most common color.

I'm actually surprised to see White at #4 because i would have thought the amount of removal it offers would have made it a more common splash color (i use the word splash to refer to multi-colored generals that have access to the color). and when add that to the fact that Swords to Plowshares is the 4th most common colored card but 4th common color. I at least find that interesting.

This kind of information confirms what people do when they have access to these different colors. Black tutor, white removal, green ramp/recursion (in reference to Eternal Witness). but i guess this isn't news to anyone. ha!

Quote from Espuma

Could you redo the graphs you did earlier in this thread? Not only are the links broken, but the data should have changed (a little bit), I think.

Do you mean the ones where I measure how many cards it takes from the top list to cover a certain percentage of the decks in the database?

I will be updating the first post next week. I will generate those graphs for you at that time.

Updated!

This months trends:

Quote from Espuma

I think +5000 for Sylvan Primordial is an error...

edit: yes, those are the graphs I mean. Thanks for updating!

That is not an error. Gatecrash cards were not ranked in the previous run of the script Sylvan Primordial is the only Gatecrash card that immediately jumped into the top 500, taking the 356th place in the ranking. That means it went from not being ranked (same as being in last place, or 5684th) to 356th. So yes, it went up 5328 places in the ranking.

Here is the graph you requested:



There are 941 decks I looked at.

Top 10% of the decks (or 94 decklists) plays 2534 cards
Top 90% of the decks (or 847 decklists) plays 5727 cards

Winner of the "I play ALL the staples" prize is Epochalyptik with his Damia list. It only takes 967 cards from the top list to cover his whole deck.

2nd place goes to IceBear's Karador deck. It takes exactly the top 1000 cards to cover his whole deck.

I LOVE all this data!

Always interesting to see the big risers and "fallers". I know why Izzet Chronarch is getting cut everywhere but I'm surprised about Ambition's Cost being sold here. It seems like a fantastic, splashable black card.

Also, how bout that big Cataclysm rise after I believe we had it brought up in a few [SCD]s?

Updated!

Crypt Ghast and Sylvan Primordial seem to be the hot new cards from Gatecrash. Can't wait to see what sort of impact Dragon's Maze will have.

Don't know if anybody still cares but... Updated!

After almost a year of being unbanned, Koko is almost in the top 200 (ranked 224th)

Our favorite dragon has finally caught up with Primeval Titan who is still ranked 231th despite being banned for so long.

Unsurprisingly, Modern Masters was also a good set for commander, ranking 2nd best product for commander.

As always, the trends for this month (actually 2 months):

This thread is amazing. Thank you very much.

Hey, what kind of script is this?
I ask because It looks more complicated than < 1hr of coding.
Thanks, and nice work!

Quote from Spongy Pengwin

Hey, what kind of script is this?
I ask because It looks more complicated than < 1hr of coding.
Thanks, and nice work!

The initial script I wrote just downloaded all the decklists from mtgsalvation and counted all the cards with the same name. That was pretty straightforward and only took me about 30min to write up.

However, I have been using this script for over a year and a half now. It is written in Ruby. Now, I download all the decklists and use a rules database that I construct from crawling gatherer to get all the meta-data for the cards. No idea how much time I spent writing it all up. Has to be somewhere between 15-30h of work. Unfortunately, as I added a lot of incremental changes to the script, it is a complete mess now, but it works

I am always looking for more ideas for statistics I could add to the results. Let me know if you know of something interesting I could look at.

Quote from scoeri

The initial script I wrote just downloaded all the decklists from mtgsalvation and counted all the cards with the same name. That was pretty straightforward and only took me about 30min to write up.

However, I have been using this script for over a year and a half now. It is written in Ruby. Now, I download all the decklists and use a rules database that I construct from crawling gatherer to get all the meta-data for the cards. No idea how much time I spent writing it all up. Has to be somewhere between 15-30h of work. Unfortunately, as I added a lot of incremental changes to the script, it is a complete mess now, but it works

I am always looking for more ideas for statistics I could add to the results. Let me know if you know of something interesting I could look at.

Good to know, thanks. The way I would have done it off the top of my head would have used Python and probably taken way longer to code and download decklists. But I don't know anything about Ruby so there's that.

I'll let you know if I have any ideas for statistics.

First, I'd like to say that the stats you provide are fascinating, useful, and I appreciate the effort that went into their production.

Here's one thing that I was trying to do using your work, but I figured it may be easier for you to do this with the data set you compile. When you list the top 200 cards, the numbers are raw counting values, so it would seem that colorless cards would be more highly represented, since they can be included in all decks.

What if, instead of just counting the occurrence of the card, the card's opportunity provided weight to the average inclusion score. That is, since there are 27 W decks, and Sun Titan appears 21 times in those decks, Sun Titan would get 21/27 averaged with 18/34 (from Azorius decks) and 15/37 (from Selesnya decks) and so on for the rest of the decks that include W.

Second example: 31/34 WU decks play Glacial Fortress. Average with that 15/36 for Esper, 17/30 for Bant, 9/15 for Raka, and ?/37 for 5-color.

The end result would be a score from zero to one (or multiply by 10,000 and round to present the score as a whole number for ease of presentation) which represents the inclusion percentage for Sun Titan in all decks in which Sun Titan is playable. If all the fractions are added together (instead of averaged), I would expect the list to mirror the top 200 list. Weighting the opportunity of card inclusion as I described above would yield a different list of top cards. I'm not sure how useful, enlightening, or important such a list would be, but it's a thought I had while looking at the breakdown presented.

I do expect this analysis to answer the question: How often is this card included in decks when possible? And how does that inclusion rate compare against other cards (that have other color requirements).

Thoughts? Questions?

Quote from Noble Larken

I do expect this analysis to answer the question: How often is this card included in decks when possible? And how does that inclusion rate compare against other cards (that have other color requirements).

A very interesting statistic indeed. And completely changes the top 200. While the original top 200 showed what cards can be expected in an average commander game, this top 200 shows how important certain cards are in their respective color combination. Bant Charm is an amazing example of this. Such a versatile card and almost an auto-include in all decks that can play it. Unfortunately, it is only ranked 349th in the original ranking because so few decks can play it. In this list it shows up 9th because 31 out of 67 decks play it.

Here are the other reslults:

The number before each card represents the percentage of decks that include this card when possible

Updated!

Kokusho finally in the top 200!

Wow, Boros Charm +140. Of the sample size, that's just about every single deck that would statistically be in the RW colors.

As every RWx deck should. It has 2.5 amazing modes for 2 mana.