damn this kingsbane rogue must have been really happy to finally queue into a good match-up for him because he emoted sorry every single turn

The Taverns of Time!

Time-Tinker Toki has begun tinkering with the Tavern! She’s set up arcane apparatuses all over the place to figure out how our Hearthstone home always manages to be at the right place at the right time.

Luckily, future Toki arrived just in time to warn us that present Toki’s use of experimental mana crystals (sample WP-477) will cause a resonance cascade resulting in dimension-twisting side-effects!

Tonight, on a very special episode of Hearthstone . . .

Dibby is pregnant with INANTP's child. But does TEG know about the affair? Later, Dover tries to explain to 3clipse why tide pods are not for eating.

+12

BeastehTHAT WOULD NOTKILL DRACULARegistered Userregular

protip: if you're a miracle rogue and you have played a bunch of fal'dorei striders, don't play sprint into my doomsayer because you'll probably proc 4 spooders and shame concede

Pushed up to Rank 2, 5 stars with @Beastwizard Big Beast Druid deck. Rank 1 was denied by Shudderwock Shaman, aka the most unfun deck to play against since Freeze Mage.

Oh well, maybe I'll be able to hit legend. We'll see.

Legends of Runeterra: MNCdover #moc
Switch ID: MNC Dover SW-1154-3107-1051 Steam ID Twitch Page

By my estimate, if you do every Taverns of Time daily you're given and don't let any lapse and are able to reroll a few 60 goldust quests into higher value ones, then this entire event is worth one legendary of your choice in dust (~1600). Not bad.

Took me 485 games, my longest legend yet, and I hit 1-5 over a dozen times. I blame the new matchmaking system for making me play against basically only legend players instead of other Rank 1/2 people.

No stats because I played 10 different decks over two and a half weeks so there's nothing to really glean from it.

By my estimate, if you do every Taverns of Time daily you're given and don't let any lapse and are able to reroll a few 60 goldust quests into higher value ones, then this entire event is worth one legendary if your choice in dust (~1600). Not bad.

Has Toki redeemed herself after her Witchwood monster hunt ?

Took me 485 games, my longest legend yet, and I hit 1-5 over a dozen times. I blame the new matchmaking system for making me play against basically only legend players instead of other Rank 1/2 people.

Aren't rank 1/2 players basically just legend players who haven't played enough games in the season?

Took me 485 games, my longest legend yet, and I hit 1-5 over a dozen times. I blame the new matchmaking system for making me play against basically only legend players instead of other Rank 1/2 people.

Aren't rank 1/2 players basically just legend players who haven't played enough games in the season?

I imagine that demographic has shifted with the rank floors. Those players are hitting legend earlier because they have to play way fewer games.

They made some (undocumented) changes to the matchmaking formula with the last patch and it's greatly widened the floor that legend players can queue into. High legend players hate it because their mmr gets gutted if they get unlucky and lose and low legend/rank 1-2 players hate it because we got stomped by high legend players a bunch.

Rank 2, 4 stars! 14-6 overall! Let's see if we can't get to rank 1 with this Big Druid deck @Beastwizard posted....

...MAN FUCK YOU GAME! Just casual drop 4 stars to 4 Odd Paladins in a row. Each game they had 3-4 Level Up Recruits on curve. Meanwhile Spreading Plague and Ramp are taking a breather.

Talk about taking the wind out of the climb to Legend. Ugh...

MNC Dover on

Legends of Runeterra: MNCdover #moc
Switch ID: MNC Dover SW-1154-3107-1051 Steam ID Twitch Page

I think I figured out how to calculate that card pack problem. A permutation tree isn't feasible because it would have 36^n permutations, which at only 3 packs is already close to 50000 permutations.

If I treat them as permutation strings instead then I can replace each card with a variable and then check if the last term in the string is repeated an x number of times (in this case 2 or more) beforehand. The probability of each of those strings is also identical, so instead of needing 36n calculations I'll only need n calculations, which is light work that I could do in excel.

If I test an nth level string for repetitions I think I can also just straight build a distribution and find some measures of center that way. I'm not sure matlab can handle a 200 dimension matrix though.

I think I figured out how to calculate that card pack problem. A permutation tree isn't feasible because it would have 36^n permutations, which at only 3 packs is already close to 50000 permutations.

If I treat them as permutation strings instead then I can replace each card with a variable and then check if the last term in the string is repeated an x number of times (in this case 2 or more) beforehand. The probability of each of those strings is also identical, so instead of needing 36n calculations I'll only need n calculations, which is light work that I could do in excel.

If I test an nth level string for repetitions I think I can also just straight build a distribution and find some measures of center that way. I'm not sure matlab can handle a 200 dimension matrix though.

I think I figured out how to calculate that card pack problem. A permutation tree isn't feasible because it would have 36^n permutations, which at only 3 packs is already close to 50000 permutations.

If I treat them as permutation strings instead then I can replace each card with a variable and then check if the last term in the string is repeated an x number of times (in this case 2 or more) beforehand. The probability of each of those strings is also identical, so instead of needing 36n calculations I'll only need n calculations, which is light work that I could do in excel.

If I test an nth level string for repetitions I think I can also just straight build a distribution and find some measures of center that way. I'm not sure matlab can handle a 200 dimension matrix though.

WTF are you talking about? What did I tell you about doing high-level math on the drugs!

Legends of Runeterra: MNCdover #moc
Switch ID: MNC Dover SW-1154-3107-1051 Steam ID Twitch Page

I just had Toki straight up win me an unwinnable game. Mirrorish vs big spell mage, I'm 5 cards ahead on fatigue and played my jaina 5-6 turns later than he did and had to burn most of my good removal to prevent the elemental snowball, ie I'm dead long term

Except Toki gives me Elise. The Golden monkey gives me benedictus and now I'm ahead on fatigue, I have his deck so I can still do silly jaina things and a bunch of random legendaries

I haven't learned how to do that yet but I've seen it referenced before. It's something to do with simulating randomness right?

Yeah, the Monte Carlo method is a way of evaluating integrals or expectation values by generating (pseudo)random events. It's incredibly useful for sampling higher dimensional parameter spaces such as that in the pack opening problem you described, and can be applied to a wide variety of problems.

I haven't learned how to do that yet but I've seen it referenced before. It's something to do with simulating randomness right?

Yeah, the Monte Carlo method is a way of evaluating integrals or expectation values by generating (pseudo)random events. It's incredibly useful for sampling higher dimensional parameter spaces such as that in the pack opening problem you described, and can be applied to a wide variety of problems.

You write the code for randomly determining the outcome of one event (or event chain), and then simulate a huge number of events. The statistics of these samples approximate the true distribution.

Super useful when you have more CPU power than statistics knowledge

I haven't learned how to do that yet but I've seen it referenced before. It's something to do with simulating randomness right?

Yeah, the Monte Carlo method is a way of evaluating integrals or expectation values by generating (pseudo)random events. It's incredibly useful for sampling higher dimensional parameter spaces such as that in the pack opening problem you described, and can be applied to a wide variety of problems.

You write the code for randomly determining the outcome of one event (or event chain), and then simulate a huge number of events. The statistics of these samples approximate the true distribution.

Super useful when you have more CPU power than statistics knowledge

Sometimes even with large statistical knowledge, when you have a large number of parameters it's just faster and more useful to build out a Monte Carlo simulation rather than calculate by hand. The simulation can also provide insight into other questions without requiring a complete redesign of the parameters your testing.

The other nice benefit is that errors are usually easier to spot. If you know the distribution should be Gaussian, anything else is an obvious coding mistake. Whereas when calculating by hand, you can wind up with an answer that you have no idea if it's right or not unless you calculate all of the permutation space and add up all the probabilities.That can take hours.

Here's where I ended up. Sample size of 1000 for each, I would have liked to go higher to get under 5% error but Excel was slowing down a bit and even crashed once.

So under the assumption that you get exactly one rare per pack, most people will collect 80% of the rare pool around 75-80 packs.

On a mathematical note, this is an undeniably effective method but I dislike how 'hidden' most of the work is. Not directly working with an equation makes it more difficult to envision how altering variables impacts the final behavior. Like if the rare pool was 40 cards instead of 36 I'd have to resimulate all my samples instead of just seeing the effect of that variable on the equation.

Here's where I ended up. Sample size of 1000 for each, I would have liked to go higher to get under 5% error but Excel was slowing down a bit and even crashed once.

So under the assumption that you get exactly one rare per pack, most people will collect 80% of the rare pool around 75-80 packs.

On a mathematical note, this is an undeniably effective method but I dislike how 'hidden' most of the work is. Not directly working with an equation makes it more difficult to envision how altering variables impacts the final behavior. Like if the rare pool was 40 cards instead of 36 I'd have to resimulate all my samples instead of just seeing the effect of that variable on the equation.

Have you used math lab or any statistical libraries in python? That might help a bit on the efficiency side. I usually don't do these in excel and use other tools that are specially built to handle these types of calculations.

So I caved and bought 5 packs of the current deal
I got Lillian Voss and Splintergraft
Hmm
But I am back at over 4k in dust I just find it sad I have not bought classic cards just did the brawl for them and yet I keep getting repeats

It is maddening how quickly every classic pack becomes 40 dust despite missing huuuuge portions of the set, including commons.

To this day, hundreds of packs later, I still have gotten 0 Upgrades or Feral Spirits.

Edit: One of my most recent classic packs actually finally gave me a second of one of the occasionally useful rares. I think it was a Violet Teacher?

Now you'll only get those in the next few packs.

It didn't happen a few months ago the last time I bitched about it!

## Posts

fortyonThe lie of the year.

copSilly Dibby, that's a Hunter, not a Rogue! Clearly all the Rogues got their "sorry" emote back via Toki.

Switch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

that's the first thing i could find on short notice

okay dadgodI'LL BE IN MY ROOM

*SLAMS DOOR*

Battle.net Tag: Dibby#1582

GET A HAIRCUT AND GET A REAL JOB!

Switch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

Hearthstone . . .Breakside / GOWUntappd: TheJudge_PDX

Dibby is pregnant with INANTP's child. But does TEG know about the affair? Later, Dover tries to explain to 3clipse why tide pods are not for eating.

Oh well, maybe I'll be able to hit legend. We'll see.

Switch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

fortyonThe lie of the year.

Took me 485 games, my longest legend yet, and I hit 1-5 over a dozen times. I blame the new matchmaking system for making me play against basically only legend players instead of other Rank 1/2 people.

No stats because I played 10 different decks over two and a half weeks so there's nothing to really glean from it.

Has Toki redeemed herself after her Witchwood monster hunt ?

The lie of the year.

I imagine that demographic has shifted with the rank floors. Those players are hitting legend earlier because they have to play way fewer games.

https://www.reddit.com/r/CompetitiveHS/comments/8sfgqo/matchmaking_issues/

I didn't know about that. That's dumb.

The lie of the year.

Just casual drop 4 stars to 4 Odd Paladins in a row. Each game they had 3-4 Level Up Recruits on curve. Meanwhile Spreading Plague and Ramp are taking a breather....MAN FUCK YOU GAME!Talk about taking the wind out of the climb to Legend. Ugh...

MNC DoveronSwitch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

If I treat them as permutation strings instead then I can replace each card with a variable and then check if the last term in the string is repeated an x number of times (in this case 2 or more) beforehand. The probability of each of those strings is also identical, so instead of needing 36n calculations I'll only need n calculations, which is light work that I could do in excel.

If I test an nth level string for repetitions I think I can also just straight build a distribution and find some measures of center that way. I'm not sure matlab can handle a 200 dimension matrix though.

Or you could monte carlo it.

Sic transit gloria mundi.WTF are you talking about? What did I tell you about doing high-level math on

the drugs!Switch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

bwanieonThat's what I thought this said at first

Held on to Shaw until I could pull Call of the Wild. Picked up the fire-the-first-spell-twice treasure with Toki. And then I flipped to Tess.

"Cast CoW twice" as a hero power three turns out of four can flip a 70 to 15 health deficit around in a big hurry.

Breakside / GOWUntappd: TheJudge_PDX

Naturally the chain is broken by a Shudderwock Shaman. Sigh.

Switch ID: MNC Dover SW-1154-3107-1051

Steam ID

Twitch Page

Except Toki gives me Elise. The Golden monkey gives me benedictus and now I'm ahead on fatigue, I have his deck so I can still do silly jaina things

anda bunch of random legendariesThe lie of the year.

Yeah, the Monte Carlo method is a way of evaluating integrals or expectation values by generating (pseudo)random events. It's incredibly useful for sampling higher dimensional parameter spaces such as that in the pack opening problem you described, and can be applied to a wide variety of problems.

You write the code for randomly determining the outcome of one event (or event chain), and then simulate a huge number of events. The statistics of these samples approximate the true distribution.

Super useful when you have more CPU power than statistics knowledge

Sic transit gloria mundi.Sometimes even with large statistical knowledge, when you have a large number of parameters it's just faster and more useful to build out a Monte Carlo simulation rather than calculate by hand. The simulation can also provide insight into other questions without requiring a complete redesign of the parameters your testing.

The other nice benefit is that errors are usually easier to spot. If you know the distribution should be Gaussian, anything else is an obvious coding mistake. Whereas when calculating by hand, you can wind up with an answer that you have no idea if it's right or not unless you calculate all of the permutation space and add up all the probabilities.That can take hours.

So under the assumption that you get exactly one rare per pack, most people will collect 80% of the rare pool around 75-80 packs.

On a mathematical note, this is an undeniably effective method but I dislike how 'hidden' most of the work is. Not directly working with an equation makes it more difficult to envision how altering variables impacts the final behavior. Like if the rare pool was 40 cards instead of 36 I'd have to resimulate all my samples instead of just seeing the effect of that variable on the equation.

Have you used math lab or any statistical libraries in python? That might help a bit on the efficiency side. I usually don't do these in excel and use other tools that are specially built to handle these types of calculations.

My brawl pack this week had an

Upgrade!The lie of the year.

It's nice to just be able to fuck around on this boss.

There was a new Wild report yesterday. Pretty much the same comments I made a couple weeks ago.