I spent all my time this week continuing my Limited prep for Ravnica Allegiance. With the first ever Mythic Championships on the horizon, I’m eager to give myself the best chance at a good performance possible. I’ve recommitted myself to Magic in a big way thanks to the Magic Pro League. I went into my preparation wanting to find out whether MTG Arena could be good enough practice for a major paper event.

Drafting on MTG Arena is a completely different experience than the Drafts I’ve done both in person and on Magic Online. Here’s a list of the differences I’ve noticed, and my conclusion as to whether the differences make the testing experience better, worse, or similar enough for the Mythic Championships or other paper events on MTG Arena.

Best-of-One versus Best-of-Three

While I’m aware that I can play best-of-three on MTG Arena, I strongly prefer playing on the ladder and best-of-one. There’s a goal to work toward, and the experience is just much smoother. Yes, there are edges to be gained with sideboarding and all of that, but I’d prefer to play more one-game matches, leading to more new experiences with the cards. The more decks I play against, the more I notice both how good and bad specific cards and interactions are in the format. The Draft experience is partially different because you may take a sideboard card here and there over a replacement level main-deck card in best-of-three, but I think the positives outweigh the negatives. Certain cards do actually gain value in best-of-one though. Kaya’s Wrath, for instance, can usually catch an opponent off-guard for a single game, before they can adjust their strategy in the next two.

Best-of-one does have the hand fixing algorithm that chooses the “better” of two hands to give you. For this reason, you can cheat on lands. This means deck building is a bit different, but nothing too drastic. Your decks will typically want one less land, but you’re not going to play 15 in a deck that normally wants 18.

For tournament preparation, I think the gains made in Bo1 for Limited outweigh the losses. Faster games with smoother draws while playing against a variety of different decks is more helpful for testing than it is a drawback. I can’t tell you how many times I’ve tested for Pro Tours and whenever someone mulligans below six we just go back to six cards unless the deck is one that’s supposed to be able to win with low resources or is mulliganing aggressively. These games are anomalies that don’t help determine who’s favored in a matchup or replicate a typical game. I think you’ll get enough information playing best-of-one, and the differences are obvious enough (bring in Plummet against flyers).

Drafting with Bots vs. Drafting with Humans

In theory, the bots are drafting like the average Magic Online competitor, using Magic Online pick orders. This raises the question whether this makes the average Draft better, more difficult, or more accurate for Mythic Championship testing than drafting with unknown humans in a Magic Online competitive Draft.

My answer is a very easy no. First of all, you should expect players at Mythic Championships to be better than the average player. Now let’s also assume that the bots don’t replicate the average player, and that they’re designed to replicate the players with the highest win percentages. Well, that’s a step in the right direction for the Mythic Championships, but at that point it’s not necessarily replicating the experiences of the average player, meaning the experience on MTG Arena would be more difficult than local game store Drafts. Either way, this isn’t going to translate well to the paper experience overall. Someone has to lose here unless they design bots with different difficulty settings.

But one thing is for sure—these bots are programmed to Draft in a specific way. As far as we know, the bots don’t draft based on their mana curve, or whether they need a removal spell or creature. The botsdon’t have to play games, they just have to attempt to mimic a drafting experience. They want to extract the correct amount of value out of the pack, not adjust for the correct amount of value to their own deck. This is great in theory, but if the bots aren’t calibrated correctly, it can create problems.

It also leads to a largely predictable Draft experience. Certain cards aren’t evaluated correctly, meaning I get insane value on specific cards, and know to avoid those cards if I don’t want duplicates. For instance, Grasping Thrull is a top-tier common. It’s gold, which is another issue for the bots because they don’t seem too keen on splashing. I often wheel Grasping Thrull on MTG Arena. In real life, maybe once in a blue moon, when no one else is Orzhov and no one wants to splash it, I’ll pick it up on the wheel in a real Draft. On MTG Arena it happens almost every Draft.

Also, while they updated the bots to pick Gates higher than they were, the pick orders of the Gates-matter cards seem unchanged. Gatebreaker Ram is going 4th or 5th pick pack 1, and since I know this I will almost always leave myself open to drafting a Gate deck for as long as possible. I’ve gotten Gate Colossus as late as pick 7 pack 1. Gates are going at closer to appropriate times, which means I’ve moved them up in my own pick order for MTG Arena, but since I get payoffs regularly its almost stupid of me to not anticipate having a Gate Colossus.

At the top levels, this means Draft decks will often all look the same. Players who are playing a lot are recognizing cards that are undervalued, and are starting to learn the trends of what will wheel and when. This becomes learned behavior and you start drafting with a specific style or end result in mind because you know where the value is. I’m often a Simic-based 4- or 5-color Gate deck, splashing whatever rares I happen to open or get passed. Usually I have 8-11 Gates. Bots, when given a set of rules to draft by, simply become predictable.

Humans have so many variables in their behavior that you can’t rely on anything. When one person changes something, the butterfly effect takes place and you then have to adapt your strategy.

While I like the ability to tank on picks and think about them critically, or just crank it up a notch and burn through a Draft in no time, drafting with bots creates an entirely different environment than drafting with humans. It’s difficult to make them similar enough. While that’s okay, and even something I prefer for casual play, if my goal is to prepare for a paper Draft event, I’d rather draft with humans because that’s what I’ll be doing in the actual event. This pushes me toward using Magic Online to prepare for the Mythic Championships.

No Self-Correction

We often hear the term “Draft is self-correcting.” If a color is weak, or in this case, a guild, players will avoid it, and then the players who end up in those colors or guilds will get a higher percentage of that color or guild, making their decks stronger because of it. With bot drafting, we don’t get that effect. I have very little incentive to draft a weaker guild if I know the bots aren’t treating those weaker colors as if they are weaker. This leads to issues like playing against Dimir or Boros every round. Why should I take Golgari cards? The bots will just do it! And of course, we never play against the bots. It would be the equivalent of preparing for a Modern event by only playing Dredge mirror matches or Dredge against Burn, ignoring all other decks.

This leads to huge shifts in how decks look, often playing sideboard cards main deck to counter the best and most popular strategies because there’s very little counterbalance to punish them for those decisions. This means playing a card like Mephitic Vapors main deck is often the correct decision because such a large percentage of the opponents are playing Boros that the upside is worth the potential downside of having a mediocre card against other guilds in your deck.

Without this balance it’s difficult to prepare for a paper tournament. “I don’t know what a good Selesnya deck looks like because I’ve only played it two or three times in my fifty Drafts because I just kept getting Hypothesizzles pick six.” This is the kind of issue that can come up, and as humans who want to maximize our own value, we will rarely sacrifice our own win percentage for a chance to try something new and likely fail. Maybe I will a few times, but it’s pretty hard to pass on a good deal. Without self-correction, the experience is different enough that this again pushes me in the direction of using Magic Online to prepare for Limited events.

The Average Decks on Magic Online vs. MTG Arena

I started my testing on Magic Online. MTG Arena didn’t have Bo1 Drafts yet, and I heard a lot about the Gate decks that were dominating on MTG Arena when Draft first came out. For this reason I decided that I should stay on Magic Online until the bot update for Bo1. I was a week behind since I spent a week at the Magic Pro League meeting in Seattle at WotC, and I was losing more than I was used to, to say the least. My decks all looked mediocre, and I couldn’t quite put my finger on what I was doing wrong.

I then played at MagicFest New Jersey, and had a reasonable 11-4 finish, going 4-2 in Draft. I felt that I had learned the format a bit, and immediately transitioned to MTG Arena again. Bo1 and ranked play got me hooked. I started to notice that my Draft decks were on average far better on MTG Arena. I was never struggling for playables, I was getting cards like Sauroform Hybrid and Aeromunculus much later than I was used to. I started to take them lower and lower, knowing that I’d be getting more of them anyway. My decks were substantially better. I was winning a lot more as well. My worst decks on Arena looked like my best decks on Magic Online. Maybe my sample is too small, or maybe I’m just drafting better now, but I’m fairly sure the average deck on MTG Arena is better than when you draft with actual humans. This is likely because the bots don’t learn as fast as humans, leaving value on the table, and the humans drafting with them are picking up that value. Since the bots don’t learn as they go, they’re always behind humans. Humans adapt, learn, and change their behavior rapidly. Bots need to be programmed to do so in this instance.

If everyone’s deck is better by the same amount, this doesn’t mean your win percentage will be higher playing against other humans—your decks will just often look much different. An average Simic deck looks slightly worse on Magic Online than MTG Arena because humans are more likely to hate draft, speculate with picks, and switch colors altogether.

While MTG Arena is preferable for actual gameplay—playing with better decks and not with train wrecks is more fun—I’d prefer my practice for an event replicate it as closely as possible.

MTG Arena is certainly a better tool for playing Magic for fun and even testing Standard, but the sacrifices in Booster Draft are too great to warrant using it over MTGO if my goal is to gain as much information as I can to prepare for the Mythic Championships. I will likely do all of my Standard preparation on MTG Arena, but I’ll double-down on drafting on Magic Online to make sure I’m as prepared for Mythic Championship Cleveland as I can be. My plan going into this event was to do all my testing on MTG Arena, and after thinking about it and experiencing both, comparing the differences, and coming to these conclusions, my mind has changed.