Introducing… The Season Summary Pages! (Or: What Have We Done With The Deck Database?)

Yesterday, I introduced you to the amazing new StarCityGames.com deck database, showing you all of the cool new features… or at least all the cool new features you could find by looking at a single deck. Today, we go over the Season Summaries, which give you the metagame at a glance.

Yesterday, I introduced you to the amazing new StarCityGames.com deck database, showing you all of the cool new features…

…or at least all the cool new features you could find by looking at a single deck.

Fortunately, we’ve provided much higher-level overviews for you. After all, when you’re looking to break a PTQ in half, you’re rarely concerned with a single deck that one guy played.* No, usually you’re concerned with a deck archetype — sure, Tooth and Nail is gonna be huge, but what’s the optimal way to build it? And more importantly, how are other people likely to build it? (After all, if you spend all of your time mirror-matching your teched-out version you might tweak it so much that it loses to the original build.)

And, of course, there’s the metagame. What decks are you likely to face at a given event? How much time should you allot to testing against Ponza? Or Mono-Blue Control?

Fortunately, we have all that ready for you. Do me a favor and click on this link. It opens in a separate window, so go ahead. Take your time looking around. Click a few links.

I can wait.

Back? Oh, good. Let me explain what you’ve just seen. Basically, what we can now do is assign a given deck to a “Season.” Now, the word “Season” is a bit of a misnomer — it can be an actual season, like this spring’s series of Extended PTQs, or it can be a single event, like Pro Tour: Philadelphia.

But the end result is the same: We can look at all the decks assigned to this season and compare them against each other to give you a high-level overview of the metagame.

Let’s take last weekend’s Pro Tour: Philadelphia summary. If you look at the page, you’ll see that we’ve broken it down by deck type to show you how many of each deck showed up, what percentage of the metagame each deck type occupied, the average finish for each deck type, and the total number of number-one finishes each deck type had. (The last item isn’t not that important for a Pro Tour, where there’s only one top winner — but for a massive event like States? You’ll be grateful it’s there, trust me.)

Furthermore, there are two links you can click on to get further information on a given deck:

Archetype Analysis
This gives you a breakdown of what cards were commonly played in each deck type. For example, looking at the top half of the page will show you that all Gifts Ungiven Control decks played at least one copy of Gifts Ungiven, Hana Kami, Kodama’s Reach, Sakura-Tribe Elder, Soulless Revival, and Tendo Ice Bridge.

47 out of 48 of Gifts Ungiven decks (or 97.92% of them) played at least one copy of Horobi’s Whisper, Sensei’s Divining Top, and Shizo, Death’s Storehouse.

..and so on. You can pretty much view at a glance what players consider to be the “essential” cards in a deck, right down to the oddball one-offs like the one deck that threw in Rootrunner and the four decks that played with the otherwise-omnipresent Umezawa’s Jitte.

“But that’s not enough, Ferrett!” you cry. “Sure, 100% of the decks played Hana Kami, but I know darned well that most of them didn’t play four copies! How am I supposed to know the optimal number of Hana Kami?”

Well, you can scroll down the page to the Advanced Results, which break it down by number and card name. It is here that you will learn that the number of Hana Kami is slightly in dispute; only 6.25% of all decks packed the full complement of four, while 58.33% of all Gifts Ungiven Control decks packed but a single flower.

But which is better? Well, if you look at the results, the average finish for decks with 4 Hana Kami is 121.67 — out of 308 players, mind you — whereas the average finish for decks with 1 Hana Kami is 161.11. So it’s obviously better to go with four…

….except, of course, that the winner of the whole tournament packed only 1 Hana Kami, as did the seventh-place winner, as did the 13th place finisher. The best finish for the 4-copy Hana is 12th place.

So how many should you run? It is here that you learn a valuable lesson, grasshopper: sometimes looking through a boatload of statistics only makes things more complicated. You’re gonna have to decide for yourself whether top placements or average finish is more important.

“Screw that, Ferrett!” you cry. “I want a chart! Give me a chart!” And lo, just below the Advanced settings is a big honkin’ chart. Go nuts.

But if you’re looking for analysis, I find the next section more helpful….

Statistically-Average Deck
If there are five or more decks at a given Season, the deck database can attempt to create an “average” deck. It does this in a several-step process.

1) It looks at the deck archetype and figures out the number of lands in the average deck type. For example, a Gifts Ungiven Control deck, as played at Philadelphia, had an average of 24 lands.

2) It then runs a query that asks, “Hey, can you give me a list of all the lands that were played in this deck type at this tournament, as well as the average number played in each deck? Oh, and could you sort it for me by the number of decks that ran this card? Thanks.”

In practical terms, we can see that 100% of the Gifts Ungiven decks had Forests in them. The average Gifts Ungiven deck had 9.6458 Forests, so we round that up to ten and add it to our “lands” pile.

Next, we see that 100% of all Gifts Ungiven decks had Islands in them. The average number was 1.10 — so again, we round that up to 2, and add it to the pile.* We repeat this process, going down and finding less-popular cards that weren’t played in all instances of this deck type, until we hit 24 lands total.

Any excess lands get discarded. If the average deck runs 3 Tendo Ice Bridge and we only need one more land to hit our average, only one Ice Bridge goes in.

3) We then repeat the process with the spells, adding on to the deck until we hit sixty cards total. Gifts Ungiven has an unusually-high number of agreed-upon cards, as we showed earlier, but eventually we start finding cards from decks that aren’t typical — for example, even though only 27.08% of all decks played Hero’s Demise, it would go in if we still had slots to fill.

As with the lands, any cards over sixty get trimmed. Hey, we’re not Jamie Wakefield.

4) We then repeat this process to create the sideboard. Since sideboards are usually much more volatile, I wouldn’t advise paying too much attention to this section — but I put it in anyway. Hey, you never know.

Things this process does not do:

Check for Color: That’d be way too much work. Still, it does mean that occasionally you’ll get a blooper in there from a significant subsection of decks; for example, though about 40% of all Gifts Ungiven decks packed Final Judgment, which was enough to bump it into the maindeck. It’s possible – though it didn’t happen in this case – that there were enough Final Judgments to make the main deck, but not enough to convince our deck database to add a couple of plains into the mana base.

Still, it happens comparatively rarely (and usually indicates the need for a separate category — for example, “Gifts Ungiven Control splashing White”), so you don’t have to worry about it that much.

Check For Sideboard Legality: If a card is very popular, it might appear in both near the bottom of the maindeck and at the top of the sideboard, giving you 5 Nezumi Graverobbers total. I told you not to pay attention to the sideboard.

These two errors may make it seem like the Average Deck isn’t that good — but remember, this is a relatively small sample with a large subtheme to it. If you’ll look at the States 2004 season, it gave some pretty decent average decks, and it shouldn’t be too hard to account for its quirks when building a gauntlet deck for an upcoming format.

And even if you don’t use it to build decks, it shows you in no uncertain terms what people believe to be the heart of the deck. I find this way more useful than the deck analysis, m’self.

In addition to this, you can also find the color breakdown of the event (it’s at the bottom of the season summary page), and there is also a list of all cards played at this event, ranked by the number of decks that card appeared in. Because the card list is a fairly intensive query, it is cached, meaning it’s only updated every four hours. That means the price list is usually accurate, but click through before you buy.

As usual, if you have any suggestions to make this better, let me know. Quick fixes can be done immediately, whereas the big changes might take awhile (and some might be flat-out impossible), but I definitely want to get your feedback.

Signing off,
The Ferrett
[email protected]
The Here Edits This Site Here Guy

* – Unless that guy won the Pro Tour with it, of course. But you know what I mean.

** – The chart is a matter of much debate around the SCG team. Ted loves the chart; I find it all but unreadable, and pretty much a waste of time. Still, there are amateur statisticians who won’t be happy until I give you a shot of Jeek-like goodness, so there it is for your amusement.

I should also note that the chart differs in some ways from Jeek’s, but Ted said it was fine. If you dislike it, talk to Ted, and then tell me what you’d like to see changed there; since I find it unusable, it’s difficult for me to tell whether I got it right.

This footnote is now irrelevant thanks to a bug. So never mind. In case you’re curious, as I posted this I discovered that hey, it was counting the number of cards left in the maindeck incorrectly, thus allowing more cards in than it should have, and so I was always rounding up. No need to do that.

But while we’re at it, I should note that the “Search for a card” function in the deck database will not look for Islands, Swamps, Forests, Mountains, or Plains, since there are seven zillion of them. That’s not a bug, though the failure to provide an error message saying that is an oversight. I’ll get around to it.