Wall of text and a buncha numbers. I dub thee a filthy NERD!
2 Likes
Even if you literally just won 15 games in a row, 100% winrate, to get from d5 to Legend, that’s 90 minutes to make $1.50 in store credit. That’s top end. A far more reasonable expectation for the time it takes to climb to Legend — using a tier 1 netdeck, mind you — is a little less than 15 hours. That’s scarcely more than ten cents an hour, and a lot of climbs are with worse decks. Free to play grinders are acting like working for 10¢ per hour isn’t absolutely pathetic. Unless you’re broke you’re better off paying real money. More than 10 hours of game time or less than 5 minutes of IRL work time and y’all pretend that you’re progressing
I know we all like to think on these kinds of big brain gaming celebrities but that’s not what I’m doing here.
Let’s imagine that you somehow plotted the actual skill of the pilots in every Diamond 4-1 game. I assume it would look something like a normal distribution, but the exact shape of the distribution doesn’t matter, just add a dot for wherever the mean of that distribution is. Then you also do the same thing with top 1000 Legend games, with their own dot for its mean. What we’re looking at here are the effects of the distance between those two mean dots. This is NOT about the Hikarus and Magnuses of the world, it’s about the effects of the average of the top rank of a data set (as contrasted with the effects of average of a lower rank).
Ya got me officer
He may have.
Whatever the reason, I do see, and appreciate the work that went into both his data, chart, and explanation. Credit where it is due, and I believe it is due here. 
3 Likes
Neon literally thinks that evidence isn’t evidence if the person pointing them out is a meanie doo-doo head
Like I could be Darth Satan and if I say something true then it’s true. The point of the ad hominem fallacy is that it doesn’t matter who says it, truth is independent from speaker.
I mean, you’re right, when a kid makes a car out of a card board box, puts a pot on their head, and “drives” around the living room, tell them good job and that you appreciate their creativity and efforts because that’s pretty much the equivalent of the sophistry that started this thread.
1 Like
Neon, do you believe that “statistical noise” is the ONLY thing causing the matchup winrates (such as for Mech Rogue vs Enrage Warrior) to be different for D4-1 than they are for T1KL? Do you believe that there is ANY real world cause to explain part of that difference, or do you believe that, in reality, the matchup winrates are identical and the only reason they appear different is due to limited data?
↑ Is this just a red herring within the data or does it point to an actual difference?
Out of curiosity are the deck lists for ‘control’ priest static at both diamond 5-1 & top1k legend? From my understanding the deck lists are estimated based upon the few cards seen… some of that winrate could be since the popularity of rogue/warrior are much lower in diamond… the tech cards dont come out until 1k legend where the popularity explodes and you would marginally adjust the deck list for the matchup since its much more expected
2 Likes
By nature they can’t be. The opponent’s deck archetype needs to be determined based off the limited number of cards that are revealed, so archetype definitions are created in such a way that about 95% of the time a deck can be assigned an archetype based on the number of cards actually seen from them. (Approximately 5% of games are too short to classify the opponent archetype, most of these being due to early concede.)
Specific decklist data is always from the perspective of the recorder and never their opponent, so it’s mostly invalid due to selection bias. The entire /decks/ section of HSReplay is practically worthless.
Lmao, that’s half the fun of it rofl.
There’s some games i like literally lucked out or merched my way into a f2p with like 300-1500$ virtual items worth of bunk. OFC you can’t take it out but i get honest fun out of it lol.
I think maybe if you’re not winning, sure you need to pay to win. If you’re already comfortably winning? Why not play to screw around. Legend is just one of those things for fun,
You’re right, if you want to optimize all the fun out of the game, you could just log off, send blizzard a check to your bank account, buy a billion packs. And then get stuck in diamond 2 with them. 
By technicality, playing a video game itself is a waste of time, yet millions of people still come up to show how much they hate hearthstone all day lmao.
My point is fun and it’s like a self imposed limit. If i buy my way to win, i threw a wallet at someone to win, that’s not satisfying. But i mean all things are relative in life.
If i were to save up to splurge i’d probably save up 700$ for NGreedia’s 4070ti or maybe a 4070ti super when it comes out for all the ai and computing power or if i was some 1600$ whale, one of those 4090 gpus. Not really because it’d do anything to win hearthstone or i’d need 700 fps when my current optimized computer could already run 360 and the screen is 100 hz lol.
But just if im gonna play a game to be extracted 100$+ for money, all that or not, you’d still might be paying 100-200$ in cash for the 10k-20k dust at 100-105~ dust a pack per 1$.
If one hypothetical might pay 100-200$ to play a deck for 1 month to climb to legend for 1 dollar of packs +1 pack.
But i suppose nearly half the decks in hs have a expiry date, even the 10-20k decks can get really knocked out if you nerf a core common and lose 20k to get 40 dust.
But fair yeah, if you’re living your life on a clock it’d make sense to optimize for time or your personal needs. But everyone does what they want to. If you aren’t playing a game because it’s fun. But you’re playing it to accumulate a virtual currency you don’t even enjoy using, to grind a rank you don’t even find rewarding, what’s the point? To be a live cash extractor? 
I jest, opposing end or not, spending keeps the servers up or not. Blizzard ain’t running a charity and even the free/ freemium players getting cards to brew, chase, compete, and not just be a desolated wasteland are what keeps the “f2p” model alive anyways lol.
No, no, we can’t have any sort of logical or theoretical justification.
All you need to know is that our OP has made a measure of 100% skill and only skill and exactly perfectly skill based on their extensive expertise at … looks back at notes to see qualifications… driving an uber, which is mostly statistics, too, right?
But in all seriousness, yes, you are correct it is not and can’t be an apples to apples comparison with how the information is collected and aggregated, so trying to make these comparisons is bad form. The data is neither predictive nor sufficient for inferences of causality.
These reports are literally a snapshot of what was seen in the reported place over the reported timeframe. That’s it. Trying to parse more out of them is silly and naive.
1 Like
And yet, you can get a ton of good information about what is/is not working in the game to help you climb using these numbers.
The numbers also tend to be pretty stable from week to week, and trends can be seen through the entire climb upward that make it pretty clear that there are major skill components involved with win rate changes from diamond to upper legend (because the gains/drop off happen gradually).
It’s not all just noise you can draw zero conclusions from, even if the original “it’s X% skill!” Measurements are low confidence.
1 Like
Yup, most people completely underestimate the value of tech cards in Hearthstone and their importance to the game.
Because the game only has 30 cards typically, you have very good chances to draw a tech card before or when you need it, for example something like 60 percent chance by turn 5 if you hard Mulligan.
And tech cards can literally change some matchups from 20 80 to 40 60, which is a massive shift. When metas get more narrow, tech becomes that much more important as well.
1 Like
The VS snap shots? Yes, you can get good information about what is working from them.
But they only tell you what was working because they aren’t predictive. I have high confidence the people at VS would tell you exactly that.
I have never disagreed with any of this.
I have emphasized how the data set makes it very difficult to quantify skill factors, particularly when we abuse the data to make spurious inferences.
Bad stats are bad, though. Trying to overcomplicate something that is axiomatic is sophistry.
The underlying data set is nominal data, and one is pretty limited in that sense.
I mean, from the start of this whole thing, this right here has been the obvious answer to quantify differences.
In a data set that measures only opponent win or loss, the difference in sample variance for two subsets of the population (s2 computed using all decks in each bracket) would be a better indicator of skill than anything in the original analysis and be an acceptable use of the underlying data.
And there are only a few people reading this that will understand what I mean, why it would be the better indicator, and how to actually look at it.
1 Like
Still waiting for an answer.
Nah, when they were doing the podcasts, they were usually really good about noticing small trends in card/deck usage that predicted later shifts.
There’s plenty of predictive power if you know what to look at.
That’s not what predictive means in this context.
Making educated guesses from the data that anticipates ongoing trends is not statistical prediction.
For example, the data is unsuitable for any type of regression analysis because it’s nominal data (as opposed to ordinal, interval, or ratio data).
Edit: It’s also not predictive in the sense that the meta is fluid and changes in play of different decks can make or break other decks. It’s a backward looking snapshot of what was.
Nah, Neon is actually spot on here. At some point I got on the VS Discord and asked Ridiculous Hat grilling questions for a little more than half an hour. I can basically summarize the experience as the following loop…
Me: [inquisitive, moderately annoying question]
Ridiculous Hat: [vague, deeply unsatisfying answer] Also, the reports are not predictive. Did I mention that they’re not predictive? Because they’re not predictive.
Random on the chat: HE DOUBTS YOU, BAN HIM
Now, naturally, the first time this happened I said what you said to Neon about it having predictive power, and then someone in the chat was like NOT IN THIS CONTEXT. It was just annoying that I tried to move on with my questions and they were bizarrely stuck on “predictiveness”
Also, although I still contend that the process was unnecessarily painful, Ridiculous Hat allowed me to question him for an unreasonably long amount of time and did not ever ban me, so I’d like to thank him again for that and I eventually was satisfied by the answers
Please note that by spot on, I mean that neon is correct in terms of what they would say. Not necessarily in terms of its truth or its relevance.
1 Like
This has been a very interesting read. It’s always appreciated when Scrotie posts his analytical topics.
Really eye-opening. 
Thanks, my man!
1 Like
I’m not sure how to take the idea that you find complete garbage to be eye opening, because that’s what it is - garbage.
You don’t have trash cans where you live?
2 Likes
What a knee slapper! 
At any rate, I’ll read the comments to (hopefully) discern the truth. I always respect your comments, so it’s only right (119 comments (jeez)).
I must’ve read most of the comments here, as I immediately remember them.
Nevertheless, I do applaud Scrotie’s effort, right or wrong.
1 Like