Choose Your Champion: 2019 Hearthstone World Championship
Hello and well met! Today I would like to share with you all my statistical analysis of deck archetypes for the HCT World Championship in Hearthstone. Last championship in the fall, I used this method to choose Bloodyface, who ended up taking second place. For the winter championship, I picked Bunnyhoppor, who also got second place. I figured I would give this another try and improve on what I did last time. I know it is a bit late and many people have already chosen champions, so I apologize for not posting sooner. The stats took a really long time because I had to do extra calculation by hand this time. The stats for archetypes on HSreplay, even at legend, do such a poor job representing actual match ups because the decks brought for worlds are so different from the typical ladder deck. In order to work around this, I had to find decks as close as possible to those at worlds, and use the stats for them at the highest rank there was data for.
After first seeing the deck lists, I used HSreplay to create a custom scenario in the matchups tab of the meta overview where each deck is as prevalent as it will be in the World Championship. I then sorted the decks based on expected win rate using legend data from the last seven days. This suffers from the issue mentioned above, but I did it anyway as a starting point for predicting which decks would do well. Here were my findings:
While nearly every player brought a warrior and a rogue to the World Championship, the data in the table to the left justifies this decision. These decks are clearly powerful, even in a field were they were certainly anticipated. Additionally, conjurer mage also appears to be very powerful. Only about half of the field brought this deck, but it seems like a great answer to warrior. It actually is favored vs zoo and token druid too if the deck is teched for those match ups, as many players chose to do. Many players designed lineups around banning rogue, and rogue is one of the only effective counters to teched conjurer mage, so mage has a great match up spread in ban rogue lineups.
However, 2 of the most popular decks in the current ladder meta, zoo and token druid, are not in a position to succeed at the World Championship. It seems that most competitors anticipated that opponents would bring these decks, and many brought decks like control shaman to counter them.
So these are the base win rates of each deck in a simulated ladder where they each play all of the other opponent decks once. Obviously, this is different from the tournament, where bans can get rid of the most polarizing decks. While I cannot possibly know the inner thoughts of the 16 players in World Championship, I did try to identify the most problematic archetypes for players to face, then remove those decks from the field and determine the average win rate of the lineups without the decks that will be banned. Since this method doesn't ban any player decks as a counter ban, win rates of all line ups will be inflated above 50%
Based on the chart above, LFYueying and Bloodyface look like the strongest contenders against this field based on lineups alone. Viper and Bunnyhoppor also look set to do well, and Roger, Killinallday, and Bloodtrail also have lineups that should give them a solid chance. The rest of the field would have a hard time if the format was a giant swiss round. Luckily for them, the groups system could protect them from some bad lineups, but I suspect even that can't save Jing, A83650, or Tyler from their terrible odds according to this data. However, it is worth pointing out that HSreplay combines Miracle, Burgle Tempo, Lackey Rogue, and Myracle decks all in one deck, Tempo Rogue. I had no choice but to use the combined data for match up calculations here. Obviously these have differences in match ups that aren't accounted for, particularly Miracle Rogue. Miracle Rogue seems to be worse than the other types of tempo rogues at least, so Viper and Bunnyhoppor may be experiencing significant win rate inflation in the chart above.
While all players must be equipped with lineups capable of contesting the entire field in order to have a chance at winning worlds, it is more important for players have strong lineups against the rest of their groups. After all, only half of players will survive the group stage, and it is possible to play the same person twice in groups. So lets try to predict who will make it to the top 8:
Group Analysis:
Wow! How do you know that Bunnyhoppor will beat Bloodtrail 50.75% of the time? Well, I don't. But I do know that if Bunnyhoppor and Bloodtrail have their best decks against each other banned away, an average legend player piloting one of Bunnyhoppor's other 3 decks will beat another average legend player piloting one of Bloodtrail's other 3 decks 50.40% of the time based on HSreplay stats. Over the course of a match, this difference in chance for each game means that Bunnyhoppor has a 50.75% chance of taking a match victory. Obviously, the highest tier of play may have slightly different match ups, but I hope that the trends of the match ups would still be similar.
To determine the average match win chance, I had to account for the fact that each player's first match is predetermined. Therefore, they can play the same person twice and the chance of beating the player they play first is more important than the chance of beating other players. The importance of each win rate can broken down by thinking about it like this:
-The first match is against one person
-The second match is against one of two people, who are close to equally likely to win
-The third match can be against any of the other three people in the group, but it is only played half of the time
I then figured out the odds of each possible match scenario happening using odds of possible outcomes, and summed up the individual probabilities in each scenario multiplied by the chance of the scenario happening using excel to find the total probability a player made it out of groups.
Side note: In order to more accurately represent rogue win rates in lineups with the class, archetypes at worlds were sub classified into decks with Myra's Unstable Element, a small burgle package, heavy lackey synergy, and Miracle. When calculating win rates, win rates from the sub-archetype used by the player were used instead of win rates for all of tempo rogue. This required using data from a smaller sample size, but I believe it to be more accurate as different variations of rogue do differently in different match ups.
That's enough math for now. Let's talk about what the data says! According to the group A predictions, Bloodtrail is the most likely to move out of the group, while Bunnyhoppor is also likely to do so. Bloodtrail brought some of the most common decks in the tournament, and his lineup isn't particularly strong against the field as a whole. However, it is effective against the ban rogue lineups in his group because it punishes the choice of bomb warrior in them. His lineup is very aggressive, and doesn't mind leaving up a bomb warrior because the warrior doesn't have the defensive control tools to stop him like control warrior does. Additionally, it is strong against the token druid and zoos in his bracket. Bunnyhoppor, however, opted to not bring a warrior at all. He brought a beatdown lineup that does ok against warriors but is very strong against board flood strategies like zoo and token druid, since his mage is teched against them. Because both of these players brought anti-token strategies and everyone in group A has zoo, they are likely to be the ones moving on, though first and second seed will likely be determined by the first game they play. XiaoT has an interesting lineup that would indicate a ban rogue strategy, but his techs show that he is banning warrior. The choices of witchwood grizzly in hunter, no rafaam in zoo, and no elysiana in warrior force him to ban warrriors, since his decks lack the toolkits to take warrior into the late game.
Here in group B is one of the players identified earlier as having one of the strongest lineups against the World Championship field: LFYueying. Yueying's strategy is to tech for warrior and bring the 4 best decks that can contest it, and it turns out to be a good plan. The other members of group B opted for anti-token strategies, and Yueying has neither zoo nor token druid, giving him a substantial edge. Additionally, there are two shamans in group B, which is a great match up for most of Yueying's decks (Rogue is the only one not favored, and it is still close). In order to represent the championship matches as well as possible, shaman data was taken from decklist win rates, not archetype win rates, because the control and shudderwock archetypes on HSreplay are riddled with bad ideas that would cause win rate deflation. The issue with this, however, is that for some match ups L-5 data had to be used instead of legend data only, and many match ups involving shaman had low sample size. This could cause lineups with shaman to appear worse than they really are.
Killinallday is the only player at worlds to try bringing Miracle Priest, and it wasn't the worst idea. The deck actually does very well against token decks, and is capable of highrolling other match ups. However, it does very poorly against midrange hunter and conjurer mage, and in order to leave group B it would have to defeat a lineup with both of those. Muzzy had a similar idea with control shaman instead of priest, but shaman doesn't do much better against the hunters and mages used by Yueying and Languagehacker. Languagehacker's lack of any warrior is certainly a detriment in this field, but not having a zoo or token druid may be enough to take him to top 8, as his entire group focused on beating those.
As shown in the chart for group C, it's pretty much impossible for Hunterace to get knocked out in groups. Hunterace is known for consistently playing really well, and I expect him to have pretty much flawless play this week. Additionally, Hunterace got a really lucky group in which all members except him have Zoo, and he went for the best possible anti-token lineup. All of his decks are very favored against token druid and zoo, and he gets to play Ike first, who has both of them. Hunterace's lineup struggles vs players with Mage, Hunter, and Warrior, but warrior is the only one of those present in his group and it can be banned away.
Ike actually had a good idea with the token druid he is bringing. His build totally gives up the warrior match up for much better match ups vs rogue and zoo. The healing token build actually beats zoo 15% more often than the regular version, and it is the only thing keeping him from having a sub-40% win rate vs Justsaiyan and A83650. It's just unfortunate that he has to face Hunterace's anti-token strategy first, because his lineup isn't that bad against the field as a whole. But a second forest's aid will be deeply missed against Hunterace's control decks.
Justsaiyan and A83650 brought very similar lineups, except saiyan has a control shaman while A83650 has a midrange shaman deck. The midrange shaman looks terrible from the stats available for it, because the deck is much worse than what Justsaiyan and Hunterace brought against rogues. Maybe it can beat down some combo decks, but Roger is the only one in this tournament to have any of those and he is in a different group.
While many people on twitter and reddit have been referring to group C as the group of death, I believe that the true group of death is group D. Roger, Bloodyface, and Viper all have very strong lineups against the field as a whole, and I suspect that whoever escapes the group will be in a good position to win worlds, or at least make top 4.
Roger's lineup is going to have a lot of free wins this tournament. There are only a few ban rogue lineups that can effectively contest his counter warrior strategy. Viper and Bunnyhoppor have no warrior lineups that are very hard for Roger to defeat, while Yueying is also favored against Roger. Bloodyface and a few others have a chance at winning, but the rest of the field running shamans effectively has no chance of taking a match from Roger. The secret to beating Roger will be banning his Mech Hunter. The other decks will almost certainly beat Roger's pass until turn 4 decks, even the bomb warrior since it has no warpaths and only 1 brawl. Basically, the warrior has 3 tries to beat something from Roger, and it needs the warrior mirror match to have the best chance. Warrior only beats Roger's hunter 5% of the time so it has to be banned.
Roger actually got a terrible group since his lineup can actually lose to Bloodyface and is very unfavored vs Viper. Depending on how matches line up, Roger's tournament run could be decided by who wins between Bloodyface and Viper. If Viper moves to winners and faces Roger, Roger probably has to face Bloodyface for the spot in top 8, and that's a match he can lose. But if Bloodyface wins round 1, Roger's top 8 is very likely.
Viper has the same lineup as Bunnyhoppor, and has a better group since he is favored vs Roger. However, Roger and Bloodyface are the most likely to move beyond groups because they are both so favored against Tyler's passive decks. Tyler has to win games against an anti-control lineup with both Shudderwock Shaman and Control Warrior in order to make it to top 8, which is almost impossible. The only match Tyler seems to have a good shot in is vs Viper if they both lose round 1, but even then he would likely just be securing Roger's path to top 8, as it would be very difficult for Tyler to win losers finals.
Bloodyface seems like one of the top choices for choose your champion, as he has proven his skill in the last two seasonal playoffs by making finals and top 4. Not only does he have a lineup that evades most of the anti-token strategies, but it also has a good chance against the ban rogue strategies likely to be in top 8. If Bloodyface makes it out of groups, I wouldn't be surprised at all to find him in the grand finals.
Final Thoughts:
After careful analysis, I've decided to pick LFYueying for Choose Your Champion. He has one of the best lineups against the field, and has a very favorable group that makes advancing to top 8 likely (64%). I would choose Bloodyface if his group was a bit better, because I think he is a stronger player, but he is in a group with Roger and that could create difficulty for him. If you believe that quality of play trumps deck choices, Bloodyface would be my recommendation because I think he has the best chance to make finals for all 4 packs. On average, however, I believe that Yueying provides the most packs.
The other two good choices for Choose Your Champion are Hunterace and Roger. We all know Hunterace to be a fantastic player, and he is the most likely statistically to make it out of groups with a 70% chance. However, since he is in group C he has to face someone from group D in top 8. Bloodyface and Roger have lineups that hard counter his anti-token strategy, so he is very unlikely to advance beyond top 8. His best hope is to face Viper, which is not a good match up but should be fairly close, and Viper has a lower chance of advancing beyond groups. Hunterace can beat Bloodyface with good RNG, but there is not likely to be much Zoo in the top 8 for Hunterace to counter so it will certainly be an uphill battle for him after groups. Hunterace is the best decision for Choose Your Champion if you are most concerned with getting at least two packs, but he is unlikely to provide more than 2.
Finally, Roger has been receiving a lot of attention for his fantastic lineup. The lineup is very good, but it is also volatile. Viper and Bunnyhoppor are incredibly hard for Roger to defeat, and he is unfavored vs Yueying who will likely be top 8 and on the same side of the bracket if both players win groups. However, he gets free wins vs many players predicted to be in top 8 including Hunterace, Justsaiyan, and Languagehacker. Roger is the champion with the highest probability of providing exactly 3 packs, because he is guaranteed to face someone from group C in top 8, and Hunterace and Justsaiyan are likely to give him a free pass into top 4 because the lineups are so one sided. I'm personally hoping however that Roger gets eliminated in groups because I hate seeing cheaters win and get paid for it.
Ultimately, skill and RNG are big factors in Hearthstone as well, and nothing is guaranteed in terms of packs. I'm excited to see how my predictions turn out when matches start in a couple days.
-HeatShock