Hearthpwn

Nope, this is not a whining thread...at least not yet. I think you're more likely to face the same class after you lost a game to that class and I do have some number to back it up. 

Here's how this started. I was playing Freeze Mage today, I lose to a CW (instant concede after I identify it as CW) and then I faced 4 CW in a row. This could be just a bad luck, but I wasn't facing any CW for past 14 games and Warrior only accounts for less than 8.3% my ranked encounter this month (why I was playing Freeze Mage in the first place). So, the chance I face I would face a 5 Warrior is a row is like 3.94e-6, which is like 1 in 2.5 million odd.

So, I was a little suspicious and decided to take a look of my 4 months game records (1213 games). I downloaded spreadsheets from Google Doc (where I keep most of my game records), saved them as csv, wrote a simple MATLAB program, and checked chance of facing the same class after losing to that class against that that class's overall encounter rate for that month.  

To my surprise, for all 36 comparisons (9 classes and 4 months), the chance of facing the same class after losing to that class is higher than overall encounter rate of that class.  Some of them are very close while others aren't, but nonetheless, for every 36 comparisons over 1213 games, above statement holds. 

Does anyone else keep the game the way I do? If so, can you guys run the same test to see what happens?

Or is Hearthstone match-making system is just designed this way?  Any insight would be very appreciated!

It's not just you, I've noticed a lot of Control Hunter on the ladder tonight. That would refute your point, however, as I've beaten all but one while playing Dragon Priest.

https://en.m.wikipedia.org/wiki/Confirmation_bias 

Quote from magiclon21 >>

https://en.m.wikipedia.org/wiki/Confirmation_bias 

I don't think confirmation bias applies here. After all, with imperfect information, what I have done should be the best way to test my hypothesis.  

Can you post your data in a tabulated form for peer review?  Everything that we could tell you is anecdotal.  My experience is that I have not noticed a higher chance of meeting the same class after losing.  Maybe there were just a lot of control warriors on when you played freeze mage.

Quote from badassnewbie >>

Can you post your data in a tabulated form for peer review?  Everything that we could tell you is anecdotal.  My experience is that I have not noticed a higher chance of meeting the same class after losing.  Maybe there were just a lot of control warriors on when you played freeze mage.

I can put 36 comparisons in a table, but these numbers by themselves tells a little. Of course, there are more dependencies like ranks, time frame, which I don't keep, and all sort of other stuff, but I don't have enough data for them. Can you come up with a better testing method? 

Quote from DoxBollox >>

You need to compute just two probabilities from that entire set:  

• p(next class = same | loss) 
• p(next class = same | win)

After that you compute the likelihood of that outcome under the assumption that the two are equal given X wins and (1213 - X) losses (happy to help if needed). I would be surprised if the difference is statistically significant ... would be interesting to know.

I actually did that as well.  P(same class | loss) = 9.7% and P(same class | win) = 6.8% but only 201 games are played facing the same class as  the last game (much smaller sample size) and I'm not sure which method has more statistical significance. 

Unfortunatly, to draw conclusions, you need a immensly larger sample size. Not just one guy. It would be great if others could help get data, but I won't because i'm a lazy jerk.

By large amounts of data, I mean as much data as Ben Brode claims he has.

 Imagine it were true that after losing to a particular class, you were more likely to face the same class in the next game.

That would HAVE to mean that after BEATING a particular class, you were also more likely to face that same class in the next game.

The only mechanism that would explain such a thing is if Blizzard programmed it that way. (Having more of that particular class on the ladder at the time would also explain the results for that particular test run - but it would NOT make it true that "after losing to a particular class, you were more likely to face the same class in the next game").

Anyway, in the absence of a varied, statistically significant and carefully obtained set of test data, I think we can safely discount any likelyhood of this hypothesis being correct.

Quote from ContentsMayVary >>

 Imagine it were true that after losing to a particular class, you were more likely to face the same class in the next game.

That would HAVE to mean that after BEATING a particular class, you were also more likely to face that same class in the next game.

The only mechanism that would explain such a thing is if Blizzard programmed it that way. (Having more of that particular class on the ladder at the time would also explain the results for that particular test run - but it would NOT make it true that "after losing to a particular class, you were more likely to face the same class in the next game").

Anyway, in the absence of a varied, statistically significant and carefully obtained set of test data, I think we can safely discount any likelyhood of this hypothesis being correct.

What if it was:

Winning versus a class: Less likely to face it
Losing versus a class: More likely to face it

Lost to mage as warrior: More likely to face a mage who lost against warrior

Obviously, the Blizzard coders would have to be sadistic sociopaths to do something like this, but.
Hey.
Activision-Blizzard.

Warriors is simply your local meta, which varying from rank to rank.

Quote from Frostraven >>

Quote from ContentsMayVary >>

 Imagine it were true that after losing to a particular class, you were more likely to face the same class in the next game.

That would HAVE to mean that after BEATING a particular class, you were also more likely to face that same class in the next game.

The only mechanism that would explain such a thing is if Blizzard programmed it that way. (Having more of that particular class on the ladder at the time would also explain the results for that particular test run - but it would NOT make it true that "after losing to a particular class, you were more likely to face the same class in the next game").

Anyway, in the absence of a varied, statistically significant and carefully obtained set of test data, I think we can safely discount any likelyhood of this hypothesis being correct.

What if it was:

Winning versus a class: Less likely to face it
Losing versus a class: More likely to face it

Lost to mage as warrior: More likely to face a mage who lost against warrior

Obviously, the Blizzard coders would have to be sadistic sociopaths to do something like this, but.
Hey.
Activision-Blizzard.

Think about it for a moment. You can't have

1.  Winning versus a class: Less likely to face it
2.  Losing versus a class: More likely to face it

because those are mutually exclusive requirements. If class A beats class B then point (1) means that it is less likely to face class B, but point (2) means that if class A beats class B (i.e. B lost to A) then B is more likely to face class A, which is the same as saying that A is more likely to face class B.

These two options are clearly and trivially mutually exclusive.

Quote from ContentsMayVary >>

Quote from Frostraven >>

Quote from ContentsMayVary >>

 Imagine it were true that after losing to a particular class, you were more likely to face the same class in the next game.

That would HAVE to mean that after BEATING a particular class, you were also more likely to face that same class in the next game.

The only mechanism that would explain such a thing is if Blizzard programmed it that way. (Having more of that particular class on the ladder at the time would also explain the results for that particular test run - but it would NOT make it true that "after losing to a particular class, you were more likely to face the same class in the next game").

Anyway, in the absence of a varied, statistically significant and carefully obtained set of test data, I think we can safely discount any likelyhood of this hypothesis being correct.

What if it was:

Winning versus a class: Less likely to face it
Losing versus a class: More likely to face it

Lost to mage as warrior: More likely to face a mage who lost against warrior

Obviously, the Blizzard coders would have to be sadistic sociopaths to do something like this, but.
Hey.
Activision-Blizzard.

Think about it for a moment. You can't have

1.  Winning versus a class: Less likely to face it
2.  Losing versus a class: More likely to face it

because those are mutually exclusive requirements. If class A beats class B then point (1) means that it is less likely to face class B, but point (2) means that if class A beats class B (i.e. B lost to A) then B is more likely to face class A, which is the same as saying that A is more likely to face class B.

These two options are clearly and trivially mutually exclusive.

No.

Player A: Lost versus Mage as Warrior
Odds of facing Mage: +0.5

Player B: Won versus Warrior as Mage
Odds of facing Warrior: -0.5

Player C: Lost versus Warrior as Mage
Odds of facing Warrior: +0.5

Player D: Won versus Mage as Warrior
Odds of facing Mage: -0.5

It works perfectly well, however -- it'd be a contrived and extremely cruel system that creates trauma and frustration.

Quote from DoxBollox >>

However, the data is sufficient to suggest that the Blizzard match-making system in general tries to prefer matching you against different classes from the one just played. => If it was fully random then you would expect to get the same class significantly more than 1 in 9 times (as not all classes are equally popular), whereas you seem to get them significantly less than 1 in 9.

I'm getting the same class more than 1 in 9. P(same class) = 16.6% > 1/9 = 11.1%. 

Quote from macbx_jfc >>

I think you're more likely to face the same class after you lost a game to that class

I dont know about that. I faced 3 CWs back to back the other day on ladder. It wasn't until i lost to the 3rd CW that i faced a different class 

Quote from ContentsMayVary >>

 Imagine it were true that after losing to a particular class, you were more likely to face the same class in the next game.

That would HAVE to mean that after BEATING a particular class, you were also more likely to face that same class in the next game.

The only mechanism that would explain such a thing is if Blizzard programmed it that way. (Having more of that particular class on the ladder at the time would also explain the results for that particular test run - but it would NOT make it true that "after losing to a particular class, you were more likely to face the same class in the next game").

Anyway, in the absence of a varied, statistically significant and carefully obtained set of test data, I think we can safely discount any likelyhood of this hypothesis being correct.

Good point. This kind of behavior can also be an unintended byproduct of match-making system. Also, I realized that I tend to queue quicker after losing (I tend to take a short break after a win streak).  I guess there are just too much room for bias to make any sound conclusion.

In match-ups i even experience other sadistic coincidences: I have hearthstone decktracker and have a ton of statistics.

I play big priest ranked wild. and i notice i get matched more and more with rogues. which are impossible to beat. After i win a few games i get matched with an enemy 2 times, because the system knows i lose to that deck? is that even possible? because i experienced that over 10 times in 2 days..

This has mayor influence in my game statistics but mainly on psychic level. I get demotivated so much, i even start to swear publicly, so i stop playing and i dont get a fun experience i am looking for.

I am used to having around 60% winrate with big priest but it now dropped to below 50%. and that is in just 50 games, i find this development concerning.

Any way, that is a bit of my observation. I have a bulk of stats to back it up, u can ask me for it. I will just play hearthstone less and less. for sure

Nice Necro...

Well there is merit to it. Very, very, very little merit.

When you win or lose a game, there is a possibility that you cue up again instead of closing the client. Your opponent could be doing the same. There is a very, very, very small chance you will face the same opponent again and a chance neither of you have changed decks.

Other than that it is confirmation bias. You are looking for a pattern and of course you will find one.

Put your foot in the ground and start writing ALL your games down. You are not allowed to look for any other pattern than “If I lost the last game, will I face the same class again?”

You have to do this at least 1.000 times.

Glad we're necroing 3 year old threads.