https://www.vicioussyndicate.com/drr/faq-data-reaper-report/
This is probably as good as you can get. I recommend looking at the websites of other statistic sites yourself.
Cheers, but if you scroll upwards, I already mentioned this link and labeled it as not even close enough.
I know, I’m just saying that it’s the best you’re going to get (unless you get a job there)
@Chicharito, I’m not sure what else you want to know. We know how they collect data, we know how they try to avoid self selection bias, we know how they categorize the decks, we know that pretty much everything else are descriptive statistics. What is the specific information that you want to have access to?
@Phillybear, actually I’ve engaged with them and HSReplay on Twitter way back when I did the video about this stuff. They are pretty accessible.
I hate bringing this up, but have you people not written a thesis?
It’s collected via programs players use like Decktracker which track and record the games players play, what cards are used, who won, etc etc. This information is fed directly into the databases of sites like HS replay from these programs. Their databases have program systems in place that analyse the information and break it down into statistical components.
The processing isn’t that complex. For example (and these are just made up numbers) 10,000 mech hunter games played. mech hunter won 5,700, therefore mech hunter has 57% win rate. Ursatron was played 16,000 times, ursatron has an 80% played rate. Hunter won 7,000 when Flark was played, hunter has a 70% win rate when Flark played. etc etc etc. Obviously these examples are slightly dumbed down but that’s generally how it works, it’s not really that complicated a process, as the information is fed to the site via the apps computer programs analyse can collate the information into their relevant fields.
I’m not saying I’ve got a better alternative, but you do see the numerous shortcomings in this method, right?
I’m a mathematician, I fully understand potential pitfalls of statistical analysis but, go on, humour me 
Quite arrogant there, I don’t know how you’re not embarrassed saying “I’m a mathematician” every other post. You should let your expertise do the talking. Statistics and math are two different things by the way.
Wow, any need for the snarky remark? And why should I be embarrassed about being a mathematician? I’m proud of the fact I successfully completed my masters degree thank you very much. I don’t shoot you down for your achievements, don’t shoot others down for theirs.
Now if your done been a d*ck while I’ve just been trying to be helpful, explain to me what you mean by there being numerous shortcomings so I can understand what your concerns are like I was trying to do in my last response
Alright, sorry, I may have misinterpreted this.
Came across like ‘go on and try your best, it’ll be wrong anyway’.
And thanks for your explanation, but there are so many variables that aren’t being taken into consideration. It’d take months of testing in controlled situations to get any kind of reliable information.
The language barrier is holding me back here, I didn’t study statistics (though it played a major role), so I only know the terminology in Dutch. Trying to translate would just be confusing.
1 Like
Sorry, it was supposed to just be an ‘off the cuff’ jokey remark, that’s why I put the smiley face 
I have quite a dry, sarcastic sense of humour which doesn’t translate very well in this medium 
Apologies if I came across that way it wasn’t my intention.
I fully understand that concern and you are right. These kind of statistical analyses are not meant to be hard gospel truth, rules to live by kind of thing. They are simply a tool meant to give you some insight into the cause and effects of things through probabilities. But, like anything involving chance, just because…for example……a warrior has a 70% chance to win if they play Dr. Boom, that doesn’t mean you WILL win 70% of your games if you play him.
They’re there for insight, nothing more. It is just as stupid to rely solely on statistics as it is to dismiss them as useless. They are only part of the bigger picture. Deck/opponent/card knowledge and experience is just as important.
But they are still good indications of things. Like all these people who complain about conjurer’s calling being OP and mage being too strong because of it. But we can see FROM that statistical analysis that conjurer mage only has an approximate win rate of 50% so by definition it CAN’T be OP, but there is a lot more going on under the surface than just that win rate or just the lucky high rolls that people complain about. For example, someone might complain they ALWAYS lose to this deck…but we don’t know what they are playing or how they are playing it. They could be playing priest which plays into mages strengths and so gets regularly beaten by it.
But as I say, statistics are a tool…not a rule
1 Like
Are they a good indication though? It feels like these percentages are just a bunch of data thrown together without any control whatsoever.
It’s not really that straight forward. You have to go a bit more in depth because general statistics, on the surface, are pretty ambiguous.
Lets go back to mage as an example. As I said, HSReplay has it’s overall winrate at about 50%. So it looks ok…not great not terrible. It BATTERS priest, paladin, most druid decks. It’s fairly strong against warrior too. But it gets destroyed by murloc shaman, zoolock, token druid…every hunter archetype.
So now you have to go look at the meta…and we can see from the tier power list that the strongest decks in the meta are mech and secret hunter, control warrior, token druid and murloc shaman.
So from that you can glean that mage is a reasonably good deck, it’s really strong in control situations…but the majority of the strongest decks in the meta aren’t control decks, they are decks that counter mage…so if you try to ladder with mage you can assume that you’re probably going to have a bad time in most games.
It’s not the statistics themselves that are the important thing, it’s how you use them in combination with other knowledge to come to a decisions that’s important.
I’m only really scratching the surface. I’m good with statistics, but to be totally honest, I play hearthstone for fun so I don’t go as in depth with them as I could because…well I just can’t be bothered to be honest. I just want to play the game and have fun, I’ll dip into them a bit to guide me in the right direction, but I don’t delve into it too far or it would take a lot of the fun out of the game for me.
You can guarantee that the pro’s…if their not doing it themselves…have a team that analyse the statistics a lot more in depth and use them to build their strategies. It’s happening the world over now. I’m a big football (soccer) fan and the level of statistic analysis that the top football teams go into now is INSANE, and I’m not just talking about games…they do full statistical analysis on potential players signings…….merchandise…everything.
But it’s how you use them to get to the answer you want that’s important. It’s can actually be easy to misuse statistics if you don’t use them properly. You have to have more knowledge about what you’re using them for. You could look at the meta list and see that Murloc Shaman is really strong. But in you’re reagion there could be a lot of people who are playing control warrior, which counters Murloc Shaman so, as strong as Murloc Shaman is, you’re going to struggle because a lot of your matches will be against your counter.
Interesting sidenote. During my degree I did a module called ‘Lying with numbers’ and it was all about how companies and government use statistics to mislead people by using selective surveys and things like that. A really simplistic explanation would be…you do a survey to find out if dogs are evil or not…but you only ask cats…and so your survey suggest that dogs are evil, which isn’t necessarily true Just an example of how statistics can be misused
VS and MetaStats
Tempostorm has lost it’s accuracy about meta and power lvls, still good with mulligan guides
This is what I’m saying though, I can’t glean anything, because I don’t know under which conditions these percentages developed.
And I know about the misuse, see my Trump example, which is kind of the whole point of this thread.
I began to wander, did the tempostorm and VS stats correspond before TempoStorm “has lost it’s accuracy” as I’ve seen most of those who have used them say
Well mages OVERALL winrate is worked out by taking EVERY mage game that Decktracker etc. has recorded since the start of the meta. Doesn’t matter if it’s in ladder mode or casual mode. Then, as I’m sure you know you divide the amount of wins by the amount of games played and this gives you your win %.
But then this is further broken down in the match up tabs where they specifically take every mage vs mech hunter game they’ve recorded and work out how often mage beats mech hunter. Here we see that mech hunter wins most of those games so we can assume that mech hunter is a counter to mage. Doesn’t mean mage can’t beat it, but it’s more likely hunter will win.
The conditions that the information is gathered is nothing more complex than loads of people who play hearthstone on pc download programs like decktracker, amongst others. And all this program does is record the games you play, what class and cards you and your opponents played and who won. It then feeds all this information back to a site like HSReplay who then breakdown and compile the information into a statistical format. As you can see by looking at something like HSReplay it’s surprising just how much statistical information you can get by just simply recording what class/cards were played and who won, which is mostly what decktracker does.
Yes, this was all clear. But to give you an example, this method does not check for potential misplays. Now I know it’s almost impossible to implement this outside of a controlled setting. But it illustrates why I’m so sceptic towards these numbers.
Also, since it’s my topic and therefore I can derail it as much as I want, which football club do you support?
No. Tempostorm stats had exactly 1 data point for each class: the player that was best at it. This was confirmed by Reynard in the most pathetic demonstration of “I don’t know what I’m talking about” some years ago. It just went downhill from there.
This led to things like a graphic showing that Warrior had a 60% WR versus Mage. And then a graphics showing that Mage had a 60% WR versus Warrior. Should be easy to spot this is impossible, unless their data collection was severely biased, which it was, towards the pros that collected it.
I’ve always defended TempoStorm qualitatively. Their insights are great. Their data though… just ignore it.
2 Likes