I’ve noticed some top ~10 players ignoring those percentages too. I was wondering if firestone ignores the “VS opponent” stats that something like hsguru shows.
But maybe it’s just mathematically impossible to calculate the chances because it might depend way too much even on one 1 single extra card mulliganned.
I don’t know how they calculate it, that’s the thing, it’s kinda sus
Whenever I build the deck (which is 80-90% similar to generic sludgelock decks), for the first session I play there are no mulligan stats, because noone has ever played the exact same deck.
Next session, however, suddenly they have the data, from 800 or over 1k games, and I’m like…how??
I think they calculate the usage of certain combinations of cards that often go together, at best, or usage of individual cards, at worst, when they show those data.
Obviously, that’s a problem, because if I have Yogg and Zilliax as my only “failsafe” late game cards in otherwise aggro deck, I don’t wanna see them before they’re playable for me (turn 9, or turn 11 against druid xD)
There’s one more problem with their mulligan data - they’re not quick to update it. Maybe it’s quick enough for someone in Diamond, but in Legend already, they’re too late, let alone in higher legend. They are STILL showing me how to mulligan against Painlock, when Painlock is 1/50 warlock games I play.
So I’m basically left to my own devices when playing against Insanity.
In every other matchup, it works like a charm. It even improved my mulligan a bit
Hsguru has a help page that explains it’s just “Kept Winrate - Deck Winrate”(for the example of “Kept in the mulligan Impact”). That means it just counts how many times the archetype won when they kept a specific card divided by the total games; e.g. if the archetype won 60 times after a card was kept in the mulligan of 100 games then the card has 60% “Kept Winrate”; THEN they subtract from that the total “Deck Winrate” of the archetype (which is just “number of wins over total games”) to get the final “Kept Impact”.
The thing is they try to refine it by only counting something like “VS Opponent”(I have no clue if Firestone does it but hsguru has the option); but I suspect it’s STILL not enough; at the very least hsguru doesn’t filter with a “VS ARCHETYPE”(only VS the entire Class) and it probably depends on what other single cards you have in the mulligan etc or the sample size is just too small because of the complexity.
Druid now has access to too much ramp for this card to be fine at 6 mana (and new heights allowing the druid to ramp past 10 makes doomkin much stronger as well).
It’s effectively being played as a 4 drop these days, and the card is definitely propelling the class into complete blowout turns that you can’t respond to because it’s doing 6 mana worth of stuff (removing mana from you and adding one to the opponent) while also having a 3/4 body attached, and hitting the board far sooner than you’d reasonably be able to pressure it in anything outside of hyper aggro or handbuff paladin.
As far as the mulligan win rate being low, that’s likely because Druid has much, MUUUCH more important cards to have in their opening hand (the cheaper ramp) and they’d rather draw into Doomkin later.
Yeah, keeping those kinds of cards in the mulligan is almost always a bad idea. The only real exception is if the games are virtually guaranteed to go to the point where they are playable regardless of what you do, and that card is required to break through a strategy you know is coming.
At some point the odds of you winning before its playable drop low enough that the odds of it still being in your deck at the time it’s needed starts to overpower it. That’s… really rare for an aggressive deck though.
Nah. It’s just that the calculation is extremely primitive; they mainly calculate single cards and ignore totally what other cards are in the mulligan in the same game; that’s what hsguru does according to their help page and I’m pretty sure it’s the only thing the others can do to keep it calculation-possible.
That means that card COULD be in the mulligan if you have a great curve with other cards in the Mulligan; sometimes the “Kept”(not the “Mulligan”) impact is very high on those stats sites because players kept something only when it had a curve already of other cards; you just have to do it live.
I don’t think in general its a bad stat, but I do think its’ inflated with Doomkin, a lot of players tend to just, concede the moment its’ played even if they still have a decent chance of winning?
Yeah, I’m generally a proponent of choosing your mulligan more dynamically than boiling it down to a singular impact number. There are a lot of things to consider when choosing those cards.
Yeah, that’s how it all started, with Unkilliax Warrior
And then I noticed that Yogg is a great failsafe against any deck if the game extends, because it’s flexible - it can steal, it can clear and it can provide some value
Just like in good old times when you ran one late game card to salvage what could be salvaged 
Now this is where things start to become a bit more complicated…
I’ve tested HS guru mulligan data, and they are terrible for me, most likely because top 500 or top 1k samples are miniscule, non-existant.
I’ll have to trust you about how they calculate it on HS guru, but I think they do it differently on Firestone - I don’t think they bother recognizing an archetype - they show you stats vs class.
If 90% of all warlock decks on the ladder are Painlock, the stats will be great for against Painlock and terrible against other warlock decks, and vice-versa.
Currently, it’s stuck in that Painlock era, so it sucks against Insanity
Similarily, against DK, class which has 6 viable decks at the moment (maybe even more, lol), it’s chaotic. You’re literally gambling. I don’t even bother watching the stats, I just pick what I feel is the best based on experience.
I have NO idea how they managed to bring about such a meta. We have 8 competitive classes with an average of 3-4 decks each. I don’t think I’ve ever played in such a meta in my 10 years here. NO mulligan stats in the world can work well in this meta.
Unless they utilize machine learning in real-time.
Make sure you at least filter at hsguru by “VS Class”(with the right opponent class). I mentioned that as a dead giveaway of a bigger limitation: they don’t even have a “VS Archetype” filter (e.g. Paladins are extremely different opponents on handbuff compared to flood). And since those calculations are primitive enough to never consider the OTHER cards in the sampled mulligans at the same time: they have no chance to ever be amazing; I doubt any site does it and if they did: it would be through a series of complicated filtering that would probably be low predictability anyway; I think the best one can do is to understand all those “Impact” numbers are very primitive and do very careful educated guesses like “well if the Kept is very good for this but the Mulligan mediocre it means the card was amazing in a Mulligan but only if it had another curve of other cards in the mulligan already” and so on and so forth but at the end of the day it’s complicated enough that it’s probably simpler to just …EYEBALL IT.
reno and concierge druid dont run doomkin only dragon druid does
It’s definitely in highlander druid.
Concierge is the only one i’ve seen drop it, but that one is trying to end the game by the time it has 6 or 7 mana, so it’s understandable why they drop it.