Over vs Underdrafting

11 posts / 0 new
Last post
I was reading an interesting article on Ars Arcanum today, which suggests that the strategy of 'Just Pick the Best Guild/Archetype/Color Combination' might be correct

puremtgo.com/articles/ars-arcanum-over-a...

The conclusion that the article draws is that under and over drafting of a particular archetype/guild/color combination doesn't have a large effect on the win rate of that archetype/guild/color combination. If this is correct, then it's mainly about figuring out which one is strongest, and agressively drafting it. I remember in AVR that I was often forced out of GU, only to lose to someone who stuck with GU in the finals. In RtR I found the same thing with Selensya, and in Gatecrash I find it with Orzhov. If I win the first two matches with a non-Orzhov guild, I inevitably face a tough matchup against a deck running Orzhov in the finals.

If this is correct, it doesn't seem like the other archetypes/guilds/color combinations have much chance at getting better, because the draft level of a particular archetype doesn't have a strong influence on win percentage. The person writing the article came to the conclusion that any changes in win rate based on popularity was attributable to the meta-game (frequency of matchups), not the average draft level of a particular color combination.
After last night, I can really get behind this. I drafted a great simic deck, couldn't really ask for better, 19 creatures curving to 5, loads of evolve, 2 pit fights, and the 2/4 mythic guy that buffs your creatures based on his power. The deck lost the two matches it was in (and got a bye), not really from mana flood / screw but from not drawing perfectly. Really it was quite hapless against my opponents with openings like crocanura - sage denizen and then topdecking "expire"ment one (that's how I think of him now) and cloudfin raptor. The creatures have the potential to outclass everything, that's how the deck wins, but I got unlucky. Probably if I had the chance to play it in, say, 6 more matches it might do better. But because of what I experienced, I don't really want to draft simic again, no matter how open it might be, because the deck is just relying on another tier of luck beyond the "dont draw too many or too few lands" tier of luck. I'm just not interested in winning from luck at this point - might as well just play "war". If I can get some playable cards passed to me then I will force orzhov or boros or dimir.
After last night, I can really get behind this. I drafted a great simic deck, couldn't really ask for better, 19 creatures curving to 5, loads of evolve, 2 pit fights, and the 2/4 mythic guy that buffs your creatures based on his power. The deck lost the two matches it was in (and got a bye), not really from mana flood / screw but from not drawing perfectly. Really it was quite hapless against my opponents with openings like crocanura - sage denizen and then topdecking "expire"ment one (that's how I think of him now) and cloudfin raptor. The creatures have the potential to outclass everything, that's how the deck wins, but I got unlucky. Probably if I had the chance to play it in, say, 6 more matches it might do better. But because of what I experienced, I don't really want to draft simic again, no matter how open it might be, because the deck is just relying on another tier of luck beyond the "dont draw too many or too few lands" tier of luck. I'm just not interested in winning from luck at this point - might as well just play "war". If I can get some playable cards passed to me then I will force orzhov or boros or dimir.



Your deck sounds fine.  Don't get ROTy.  Sometimes you just run bad, even with a great deck.  That's why if you looked at Constructed DEs, one day the same player with the same deck could 4-0, and the next time 0-2 (matchups also plays a part).  You could have drafted an awesome Orzhov deck and had the same happen.
I tend to agree with you except I have to believe there is more to this game than just luck. I have to believe that making good draft, deckbuilding, and play decisions will result in more wins. Otherwise I need to quit now.

Fact is, in my experience, orzhov doesn't care about curving out as much as the other guilds. So that second tier of luck is less of a factor for them. I think that is part of the explanation of why they're the strongest.
What I dislike about threads like these is that they use statistics and explications incorrectly, basically to justify red-herring equivalent arguments to generate a single strategy that "solves" limited. There are certainly times when you have a decent deck and end up final matching the "nuts" deck of a set, but there are also times you play a 5-color deck and it manages to wiggle around the top tier decks and crush opponents.

The game inherently has randomality in its structure. You simply cannot ignore this in an analysis, because it undermines any statistic-based conclusion.

Worse, it forces a singular perspective. If you conclude "Guild X is the best guild because its win percentages are Y," you're also forcefully concluding "Guild Z, etc. is therefore worse/unplayable compared to Guild X." In other words, you're weighing percentages without the wealth of data that factors into those games/matches, such as the number of Orzhov decks in the pod, how many people were drafting red, etc.

To sum, I agree Orzhov is the "best" guild if you define "best" as the most consistent. But that does not mean Gruul decks are therefore "worse" in any shape or form. This is great for speculation/mtg excercise in futility. It is not indicative of real-world drafting.

I found Carmen Sandiego before you were born unless you're Zlehtnoba.

Forcing what you think is the best deck is just a different way of hedging against the variance in packs and colors being open. You'll be in Orzhov when it's right more often, but also when it's wrong (and it is a lot of the time, if not quite 80%).

I want to emphasize what Matt and I discussed in the comments of this same article; data can help us shape an impression of some general insights, but a lot of the things you can draw from data describes a tiny slice in the space and time of a format which is in constant transformation. The key to using data is to use it right (not merely to cherry-pick the parts of it that supports an idea you had before seeing the data). I can recommend Nate Silver's book "The Signal and the Noise" if one is really interested in how to use data (and how not to).
Preparing for the M14 Prerelease - New article up! IN THE TANK - my very own blog for rambling about Magic!
Yeah, I don't think it is all luck, or even mostly luck, but I think good decks run bad sometimes and average decks can run really well.  I've drafted some decks that I felt were very powerful and they ended up not doing so well, and some that I thought were not very good at all and they win.  So, while luck is a factor, it is often mitigated by things like deckbuilding, playskill, and so on, but there were be times when a well-built deck just does not do very well because you don't draw what you need, you have to mulligan into oblivion, or whatever.
After last night, I can really get behind this. I drafted a great simic deck, couldn't really ask for better, 19 creatures curving to 5, loads of evolve, 2 pit fights, and the 2/4 mythic guy that buffs your creatures based on his power. The deck lost the two matches it was in (and got a bye), not really from mana flood / screw but from not drawing perfectly. Really it was quite hapless against my opponents with openings like crocanura - sage denizen and then topdecking "expire"ment one (that's how I think of him now) and cloudfin raptor. The creatures have the potential to outclass everything, that's how the deck wins, but I got unlucky. Probably if I had the chance to play it in, say, 6 more matches it might do better. But because of what I experienced, I don't really want to draft simic again, no matter how open it might be, because the deck is just relying on another tier of luck beyond the "dont draw too many or too few lands" tier of luck. I'm just not interested in winning from luck at this point - might as well just play "war". If I can get some playable cards passed to me then I will force orzhov or boros or dimir.



It's always painful when you draw into cards that are lower cost than your starting hand. I had a Boros deck that kept happening with. It had enough 1 drops I should have had a resonable expectation of getting one in my starting hand, yet I kept drawing hands that started at 2, and then top-decking 1 drops, that I couldn't play becuase the rest of my hand was curved out. Then, when I could finally play the 1 drop it was no-longer relevant.

I am starting to buy into the argument that Orzhov is best because it is the most forgiving to the various non-optimal draws you might get. If you play Orzhov with heavy extortion just about the only luck based handicap that can get you is mana short. If you draw too much mana it means more extortion kickers (provided you don't seriously flood and have nothing to play at all) If you have a slow start you can use extortion later on to get a solid cushion. Dimir has some homerun potential in getting a unblockable, ciphered creature, but it's hard to beat the consistancy of Orzhov. Best removal in the format also means you're less likely to be victimized by your opponent's good luck, drawing into their strongest cards.
I think the main flaw (not really a mistake, just an unknowable) in this guys stats is that he is forced to assume that either:

a) all players are equally skilled
or
b) the decks drafted are the same across skill levels.


In particular - Boros/aggro is is the easiest deck to play, and when I take my not-yet-nine year old to a FNM draft I encourage him to stay in Boros or Gruul.
Dimir on the other hand, requires a fair amount of finesse and it's very easy (speaking from experience) to end up in a half mill half tempo deck that does both poorly.  But since the meta knows this, it stays away.  

Basically I think you can sum up the MTGO results with a simple statment like:  White is the best color in gatecrash, good players will drift to Orzhov and mediocre ones to Boros.

But it's extremely difficult to prove this unless you actually had access to the win rates of individual players by deck type/guild. 
rstnme: "Everything looks good when your opponent passes 4 turns in a row."
I'm not sure this is really getting at overdrafting. 

1.  Its hard to justify 25% popularity as being "overdrafted".  That averages out to 2 in every 8 man pod drafting said guild.  Given that roughly 1/3 of the cards in GTC are red or white, this means that there should be plenty to go around for the "average"(Which may be skewed- more on this in a bit).  So, depending on where the Boros players are situated there may not be overdrafting occurring at all, even though they are the most popular.

2.  I'm curious to see a breakdown here.  To really get at if overdrafting is actually problematic, one should look at win rate in those pods where there are more than the statistical average players in said guild.  Does 3-4 boros players in a pod severely diminish the win rate?  I would imagine so. 

3.  Along the lines of both of those, one needs the data to see if the the statistical average is skewed.  This is to get at if the statistical average is really relevant, or if it is governed by math only.  I.E. are there a number of pods with 1 boros/Orzhov player and a number with 3-4 which averages out to an unlikely scenario?  Further, do the pods with only 1 or 2 players do significantly better than the ones with 4 players, to the point where this skews the win rate in a certain way?  Without this data, it is absolutely impossible to say much.

Frankly, I'd much rather have scatter plots of some sort showing Win Rate vs. number of players drafting that guild in a pod.  Among other such data. 

Basically these statistics aren't telling us much without seeing the raw data behind them.  With what's given, its basically following the same mindset that correlation equals causation, without actually giving enough data to even show that there is much correlation at all. 

There's a number of problems is what I'm getting at.  There are many factors that can skew this sort of data, and it doesn't particularly prove much on overdrafting at all.
So why is Boros doing better at the GP level?
bulletd Guidelines: 5.0: I will take this card no matter what. Creature 1 or playable 1 or hate 1.Archangel of Thune 4.5: Bomb and splashable. Creature 1-2, playable 1-2, removal 1. Jace, Memory Adept 4.0: Excellent first pick first pack, will sway me into same colors. Creatures 1-4, removal 1. Haunted Plate Mail 3.5: Excellent first pack pick two, will confirm colors or possibly sway into second color. Doom Blade 3.0: Good in-color addition, or splashable removal/creature. Creatures 3-9, removal 1-3. wall of Frost 2.5: Solid pick in-color; creatures 5-12, removal 3-5. Dark Favor 2.0: Creatures 10-16; removal 6-7. Elvish Mystic 1.5: My 23rd or 22nd card, depending on removal. Act of Treason 1.0: 23rd card if I don't maindeck an additional land. Lay of the Land 0.5: This card will sometimes be sideboarded in. Brave the Elements 0.0: I will shred this card for counters. Darksteel Forge