The Hearthstone Arena is in a weird place now. Untouched for practically 3 years of the game’s existence, the format is now seeing changes here and there. One of the more recent changes, introduced around the time Knights of the Frozen Throne was released, are the synergy picks. Your first 2 picks out of 30 in the Arena all come from a much smaller pool of cards, the synergy pool.
Most people who play Hearthstone don’t care about the new synergies, or don’t know it exists. People seemingly only care about Ranked and even Tavern Brawl, more than Arena. In the minority of Arena players that do care, some people like the new synergy, and others don’t.
It won’t take a gun to my head to have me tell you that I don’t like the new synergy system. I don’t like being forced to pick a mediocre or subpar card, for the chance of a fringe synergy. My game is about tempo, and synergy is just extra. But that didn’t stop me from thinking about these picks, given how pervasive they are in the Arena now. How good are these synergy picks, and across classes?
- I tried to determine how good each synergy pick is, in respect to the class-rarity pool. I used HSReplay.net to find where a card ranked in “Deck winrate,” out of all cards for that class, in that rarity. Deck winrate varies greatly among classes, so I decided to look ranks within class. Rank analysis is used in nonparametric study design, and is not outlandish.
- I pulled the source of synergy picks from Heartharena. I’m not sure if this list is complete, but from a visual check, it seems correct.
- I made 3 tables for Synergy Commons, Synergy Rares, and Synergy Epics.
- Each card received a percentage. This percentage stands for percentile in the class-rarity pool. e.g. The #10 deck winrate card out of 100 will be 10%.
- I added a color scale on the spreadsheet for visualization. Green is good, yellow is meh, and red is bad.
Misinterpreting or having misleading data is worse than having no data, so let’s make some assumptions before the results:
- HSReplay data is not representative of everyone – People who track all their games, and have decktracking lists are probably better than the average Hearthstone player. While this doesn’t include very good players who play on their phones, I would assume more casual players don’t use the technology that exists for computers.
- Deck winrate is not perfect – Far from a perfect metric, but the best we have to evaluate card performance. Bad cards in 12-win drafts and good cards in 0-win drafts get muddled in the metric. This goes into the fact that high-performing decks typically have better cards. But individual player skill, outside of the draft isn’t really taken into account.
- Intervals between ranks are not uniform – Cards next to each other on the rank aren’t separated equally. Typically, the worst card is really bad, and a few percent worse than the penultimate. You may even have cards with the same exact winrate, but placed arbitrarily on the rank order.
- Data is dynamic – HSReplay data constantly refreshes, and the free version looks at the last 14 days. To avoid changing data in the dataset for analysis, I did not refresh the page I looked at, to make sure I had a static snapshot for all data used.
- Statistical power is equal – Power is basically the bigger a sample size, the more reliable the results. In the ranks, I included every card not considered sparse. So the rank of a card played 100,000 times is on the same level as a card played 2,000 times. The card played 2,000 times probably needs more reps to see it’s “true deck winrate.”
Also some limitations of this analysis/visualization:
- No legendaries – Legendary picks make up a good portion of the synergy picks actually. But when looking at the data, there wasn’t enough power to have a definitive denominator for each class. For example, The Voraxx, a synergy legendary, was only picked 360 times in 2 weeks by Hunters. Everyone knows it is horrible by now, and avoids it. With that, I scrapped evaluation of legendaries. Know that Medivh, Kazakus and the DK Heroes are all very good.
- No sparse data – HSReplay automatically filters out cards that haven’t been played much. This is done primarily to weed out old drafts that include Wild or banned cards. However, picks perceived as bad by the public (and subsequently undrafted) sometimes don’t show up. An example would be Am’gam Rager, which is very seldom drafted by Mages, or Blood of the Ancient One for Druids. While it feels bad to leave out currently draftable cards from the ranks, their lack of statistical power had to be considered. Ultimately, only a few cards were left out for each class, which won’t significantly affect the rankings for the synergy picks .
Synergy pick commons
- Rockpool Hunter shows up consistently as a high-ranking deck winrate card, primarily because it serves as 2-drop stability.
- Primalfin Lookout is the worst neutral common synergy pick, likely because very few Murlocs were found after picking it.
- Netherspite Historian is actually in the top third of Priest commons, given Dragon synergy.
- Warlock got some great expansion cards, but the solid synergy pick commons help too. Same for Druid.
Synergy pick rares
- Tol’vir Stoneshaper is everywhere, as the 4-mana 3/5 is not backbreaking, being off 1 point. Much better than picking Gadgetzan Auctioneer or Coldlight Seer.
- Devilsaur Egg performs much better in Warlock, possibly due to Unwilling Sacrifice.
- Book Wyrm confirms Priest Dragon synergy.
- Warrior has relatively better ranks in the synergy rares, which could mean their overall card pool is worse compared to the synergy picks.
Synergy pick epics
- There are a few relatively safe epic synergy picks in Murloc Warleader, Blazecaller, and Southsea Captain.
- Blubber Baron is as bad as advertised. Don’t try to make that “synergy” work.
- The Paladin epic class synergy picks are mostly horrendous.
- Corpsetaker is predictably best in Paladin, thanks to Divine Shield, but is just the 50th percentile.
- Warrior has relatively better ranks again in the epic synergy picks. Horrible card pool overall confirmed.