Stream of Consciousness: Thoughts on Drake Setting Twitch Records

Stream of Consciousness: Thoughts on Drake Setting Twitch Records

tldr; Drake plays video games, Twitch goes bananas. I observe and reflect.

It’s hard not to be in the know of what happened last night. Just by being in the circle of video games, I know that Drake helped streamer Ninja set a record with concurrent Twitch views last night, while playing Fortnite with him. Travis Scott and JuJu Smith-Schuster later joined in.

Real talk. I know Drake is a rapper from Canada. I know what he looks like. He is in a meme. I can’t name one Drake song though. I also know Travis Scott is a rapper, what he looks like, and he is/was in a relationship with one of the Kardashians. I can’t name one Travis Scott song. I do know enough about JuJu Smith-Schuster, being I care more about football than rap music. And the Steelers are always on prime time TV.

The same can be said about Fortnite. I know it is a MOBA made by Epic Games, and is commonly #1 or #2 on viewed Twitch games since release. I see it as a more cartoony style of PUBG. That is literally all I know about that game.

Obviously, celebrities are the key to garnering views

This is the main takeaway, and one that will likely lead to much imitation down the line. The funny thing is that in Ninja’s stream, you don’t even see Drake on camera. Sure, it is documented that he did do the Fornite co-op, but Drake not even having to appear on camera broke records. Expect streamers to do everything in their capacity to try to get celebrities in co-ops.

Not just any celebrities

The thing is that you can’t just be a slouch. Like a side character from a 90s show, or drummer for Sugar Ray (not sure why I use this example, lul). Drake and Travis Scott are really famous in music, and JuJu Smith-Schuster gets more press than most NFL players. These co-op guests are at the top of their respective fields, and even appeal to the audience. This automatically makes younger celebrities a bigger hit with the Twitch crowd. Unless you’re someone like Bill Nye.

Good press countering bad press

Fortnite has been in the news lately in a bad light, in the war against video games, due to recent school shootings. I’m not getting into this too much, but this is harkening to 20 years ago with Columbine, as Harris and Klebold were really into Quake or Doom. I don’t expect this war against video games to persist for too long, as nothing was really done 20 years ago about it all.

But this Drake helping hit Twitch records was good news for esports. It showed the industry is stronger than ever as a medium of entertainment consumption. And whatever negative perception of this game, basically regarded an anathema in main stream media, was completely ignored for one night. The power of celebrity can do that, and probably holds more sway than anything.

It must’ve sucked for other streamers

I dabbled streaming Duel Links for a month, until an update on Steam made nothing record on OBS. It is a bit of work to be a good streamer. Some people stream just to dabble in it, but some people are more serious about streaming, keeping regular schedules. These people are either trying to build something, make a name for themselves, make money, etc.

Let’s say last night, you had some great stuff in-mind, and brought your A-game. Your viewer count is lower than normal though. Drake showing up in a co-op probably sapped views from every other streamer, pulling away people with the exception for stream-lurk/sub-diehards. It was an uncontrollable negative externality on any individual streamer.

Good thing for Twitch

I’m sure a lot of people who tuned in were not regular Twitch viewers. Drake posts a link on Twitter, a much more known social media outlet, into Twitch, something only traditionally gamers frequented. I’m not going to pretend to know the numbers of clickthroughs for this, but let’s say it was a lot of views. Twitch has expanded outside of gaming in the past year, and are seemingly doing anything for views. This was a great thing, mostly for exposing Twitch to audiences who don’t regularly use it.


I’ll end by talking about MOBAs. It seems the top 3 viewed games on Twitch at any given time are MOBAs. This is probably happening by now, but game companies should push out MOBAs that improve on what current ones don’t have. It makes sense to make more of a game that is enjoyed much by the viewerbase.



Hearthstone Arena 10.4 with Numbers!

Hearthstone Arena 10.4 with Numbers!

Out with the old, in with the new. Hearthstone has officially done away with rarity-based drafting, and has gone on to a power level-based one. I had some questions in my previous look at the draft, but it is great to finally dive in myself, to see what it looks like.


  • I drafted a Paladin Arena in Arena 10.4. I took a screencap of every pick.
  • I went to to find stats for Paladin Arenas. Stats are for the last 14 days, delayed half an hour, of this writing.
  • I copied the “Deck winrate” and “Played winrate” for each card.
  • I used Google Sheets to calculate the standard deviation for each pick. This was done for “Deck winrate” and “Played winrate.”



Deck winrate table

  • The cards were labeled A, B, C, from left to right.
  • SDdwr represents the standard deviation between the 2 or 3 picks.
  • A new Arena card, Hand of Salvation, was excluded from the data, in picks 27 and 28.
Pick Adwr Bdwr Cdwr SDdwr
1 55.60% 58.20% 57.10% 1.31%
2 58.30% 58.40% 56.40% 1.13%
3 61.90% 60.50% 62.20% 0.91%
4 57.70% 59.20% 57.10% 1.08%
5 60.20% 60.30% 60.50% 0.15%
6 60.70% 61.60% 61.80% 0.59%
7 60.80% 61.80% 61.90% 0.61%
8 59.40% 59.60% 60.10% 0.36%
9 60.20% 59.60% 60.70% 0.55%
10 61.60% 60.70% 62.20% 0.75%
11 45.90% 57.60% 58.80% 7.13%
12 60.50% 58.50% 58.60% 1.13%
13 58.60% 60.00% 59.50% 0.71%
14 58.90% 59.10% 59.20% 0.15%
15 60.80% 60.20% 60.30% 0.32%
16 57.00% 46.80% 56.60% 5.78%
17 57.70% 58.80% 55.50% 1.68%
18 59.30% 58.40% 60.00% 0.80%
19 59.60% 58.90% 61.10% 1.12%
20 58.10% 65.30% 56.70% 4.61%
21 59.40% 60.10% 59.90% 0.36%
22 50.80% 46.80% 54.60% 3.90%
23 61.90% 61.60% 60.80% 0.57%
24 60.90% 61.00% 60.70% 0.15%
25 60.70% 59.90% 60.70% 0.46%
26 59.30% 59.80% 58.60% 0.60%
27 . 59.60% 57.40% 1.56%
28 57.40% . 59.30% 1.34%
29 59.70% 60.70% 58.60% 1.05%
30 60.60% 60.10% 62.20% 1.10%

Played winrate table

  • The cards were labeled A, B, C, from left to right.
  • SDdwr represents the standard deviation between the 2 or 3 picks.
  • A new Arena card, Hand of Salvation, was excluded from the data, in picks 27 and 28.
Pick Apwr Bpwr Cpwr SDpwr
1 46.70% 48.40% 43.80% 2.33%
2 51.00% 46.50% 48.60% 2.25%
3 65.10% 56.70% 61.00% 4.20%
4 46.10% 46.50% 46.40% 0.21%
5 52.70% 54.70% 51.40% 1.66%
6 56.60% 57.10% 55.70% 0.71%
7 59.60% 55.70% 65.10% 4.72%
8 48.20% 50.90% 51.90% 1.91%
9 57.70% 50.90% 53.00% 3.48%
10 57.10% 56.60% 61.00% 2.41%
11 27.40% 47.30% 49.60% 12.21%
12 54.80% 51.60% 49.00% 2.91%
13 50.20% 53.20% 50.50% 1.65%
14 50.60% 49.10% 55.60% 3.40%
15 53.10% 53.40% 54.90% 0.96%
16 45.00% 28.50% 41.40% 8.68%
17 46.10% 50.70% 45.50% 2.84%
18 49.20% 47.40% 52.80% 2.75%
19 51.30% 54.60% 52.70% 1.66%
20 51.60% 43.70% 54.40% 5.55%
21 48.20% 51.90% 55.20% 3.50%
22 33.90% 28.50% 47.80% 9.96%
23 65.10% 57.10% 59.60% 4.09%
24 54.40% 55.70% 56.60% 1.11%
25 59.20% 54.30% 54.30% 2.83%
26 50.50% 50.90% 49.00% 1.00%
27 . 47.10% 43.60% 2.47%
28 43.60% . 49.20% 3.96%
29 50.90% 53.00% 49.00% 2.00%
30 57.90% 53.00% 53.70% 2.65%


  • This is just one draft, so small sample size. But it does provide 30 data points!
  • I drafted Paladin, a class with pretty good winrates. I suspect a bad Arena class would give more discrimination in winrates between cards.
  • The data was pooled from the last 14 days, which includes all of Wildfest. It is unclear if static Wildfest data was used to group cards, or if it will be a dynamic system with the winrates.
  • Standard deviation isn’t the best measure for 2 or 3 picks, but I was too lazy to figure out the best test.


  • The standard deviation of deck winrate was very low, with most values falling within 0-2%. Pick 11 (Street Trickster / Ebon Dragonsmith / Midnight Drake) had the highest deviation, and Pick 16 (Frostwolf Grunt / Eye for an Eye / Lesser Pearl Spellstone) second. These were the only standard deviations above 5%.
  • The standard deviation of played winrate was higher. Pick 11 had the highest rate again, and Pick 22 (Small-Time Recruits / Eye for an Eye / Sabretooth Stalker) had the second. Along with Pick 16, these were the three abnormally higher rates.


  • The power levels seemed to be grouped by deck winrate. While Hsreplay data is not perfect, it takes a sample from actual data. Overall, the deck winrates were fairly consistent among 3 picks.
  • There appears to be a distinction from premium cards from good cards. Cards like Vinecleaver, Spikeridged Steed, and Rallying Blade were grouped together often.
  • The premium pool was rather small for this draft. I drafted 2 Vinecleavers, 1 Rallying Bad, and 1 Spikeridged Steed. This could give Arena a more constructed feel, and definitely making playing around cards more important.
  • There were 2 legendary picks in this draft. I definitely saw a premium legendary group, and an average-ish legendary group.
  • The bad cards have greater variance in winrates. As shown in Pick 11, it appeared Street Trickster didn’t really belong with the other two picks. Possibly, there isn’t a really bad pool, and the really bad cards are grouped with just bad cards.
  • More data is needed. We don’t know how many card pools there are from this one draft, or what are acceptable differences in card winrates to be in the same pool. Obviously, more cards in other classes helps as well. Do your own analysis!

Implications of Hearthstone Arena 10.4

Implications of Hearthstone Arena 10.4

In a blindsiding fashion, perhaps the biggest change to the Arena was announced not long ago, due for update 10.4 of the game. This change will do away with the rarity draft system, which has been in place since the beginning, and move towards a power level draft system. For the first time, one can potentially pick from 3 cards of different rarities (common, rare, epic) for a draft slot. Also added in the post are the upcoming new Arena-exclusives, for a limited time, but the removal of the rarity draft system warrants much more discussion on it’s own.

Let’s run through some potential implications this has on Arena.

Different drafting

By taking away the power level quantity to compare three cards with, the focus becomes other useful factors. Most Arena picks have a clear-cut “good card,” which results in an auto-pick. If the new system works as defined in the video, every pick should warrant some critical thinking, in regards to deck construction. The demoed example of Fireball vs Leyline Manipulator vs Primordial Drake isn’t easy. Fireball is usually automatic for being one of the best Mage spells. Primordial Drake is probably the best neutral epic minion though. Leyline Manipulator is an Elemental Yeti with plus side for RNG decks. Tough choice.

The test would be prioritizing mana curve, synergies, win condition, minion/spell balance, etc. In a way, this accomplishes what the dreaded Synergy Pick era never did in testing synergies. Synergies are explosive and powerful when they hit, but are they worth all the other factors. Skillful drafting is definitely reinforced, when you take away sheer power level.

In a way, this drafting system works better for individuals who drafted without using a tierlist. When you have used tierlists for years, you become inured to cards having a certain value attached to that. With the new system, this becomes much more fuzzy, and there is more wiggle room.

target dummy.png
Things only 2015 Arena players will get


Definition of power level

This brings the question, how does Blizzard define power level? As the entire draft system will now be based on power level, it would be important to find out what that quantity is. My first guess is it would be the “deck winrate” value, which usually demarcates good cards from average cards from bad cards. “Played winrate” is more of a swingy value I don’t like so much, but is an option as well. Blizz probably has more internal stats that will make the basis of what power level is.

Power level buckets

Once power level is defined (whenever that is), how many buckets are there? As this is the key basis for how the drafting system works, it would be important to know. We definitely will have at least three, with good, average, and bad. The Lightforge Tierlist has 7 buckets for their valuations. I have a feeling there won’t be that many in the Arena update, but there should be enough to make the draft pick between three “same power level cards” seem similar. I personally believe there may be 5 power level buckets, which goes neatly in a 1-5 scale, and divides into 30.

Tiers of The Lightforge Tierlist

Valuation of neutral cards in classes

Thanks to class cards, hero power, and class identity, neutral cards have different values in different classes. An example is Violet Illusionist in Rogue, which is just a 3-mana 4/3 in another class. Taunts for Warlock are also great. The valuation of neutral cards should take class into consideration.

More epics

Epics are no longer bound by generally low appearance rates and should show up more in the new Arena. They usually have a reputation of being either really good or really bad. It is possible we may never see the really bad epics anymore, unless they were drafted for a specific deck synergy.

epic cards.PNG
Some Warrior Epics = Literally unplayable

Normalization of usage

If you look at class cards drafted by class, some cards appear in just short of 50% of decks. By grouping cards through power level, it is likely that the highest cards will fall in usage, and the lower cards will rise in usage. There should be a normalization of some sorts, just because cards of equal power level are pitted against one another, and you are bound to 30 picks.

shaman class.PNG
Crushing Hand vs Fire Elemental shouldn’t be this far apart.

Variability of power level

The biggest question regarding this change is how different each draft would be from another, in terms of power level offerings. And this probably will be determinant on how many power level buckets there are. If there are fewer buckets, like 3, you can easily say there will be 10 bad picks, 10 average picks, and 10 good picks. If you have 7 power level buckets, things get more complicated.

In the current rarity draft system, you typically don’t have crazy variability in terms of card rarity. Yes, you do see 3-4 legendary decks sometimes, but usually you will have 0 or 1.¬†Power level is much more important than rarity though. Many legendary picks aren’t even all that good. Premium cards are often common cards.

A bad scenario would be a highly variable system, as the number of premium picks afforded in a draft determines your fate. Hopefully this distribution of power level per draft is fairly consistent.