Torrent link from their discord:
https://bluesoul.me/r/slaythespire/SlayTheData.7z.torrent
OP on the discord:
Couple scripts for parsing the data:
Torrent link from their discord:
https://bluesoul.me/r/slaythespire/SlayTheData.7z.torrent
OP on the discord:
Couple scripts for parsing the data:
Why not release it as an sqlite database instead?
It looks like itās just raw backups of their⦠mongodb. Ouch.
Itās 27GB compressed.
7-zip reports 345GB uncompressed.
So the torrent doesnāt have all the data they released, and also the data set isnāt every Spire run.
The torrent has 2 batches of data: October 2018 through July 2019, and July 2020 through September 2020. The mega crit employee posted direct download links in their discord for October and November '20.
The game underwent substantial changes over that time frame (notably adding Watcher), so Iāll probably mostly look at the 2020 dump. Thereās 18 million runs in there alone, so plenty to go around.
I wish the runs had a player id or something, so you could find all of one playerās runs, or correlate improvement over time, etc. Oh well.
This person plotted winrate vs. Ascension for each character:
Interesting that the impression of Watcher being high-degree-of-difficulty-but-OP is borne out by the data. Winrate is higher at A20 than A0!
That might indicate that itās more specifically watched unlocks which are OP hence why they struggle at a0.
But having played a bit of watcher I would readily believe they are above the power curve.
Some preliminary stats:
I am operating only on the data from July through November 2020. There are more runs from 2018 and 2019 that I am not including, because of space (2018-19 is 290GB of json), and the game has changed since then. Notably, Watcher was added, but also some balance changes.
The raw json from the 2020 runs takes up 90GB on disk. I dumped them into an sqlite db, which takes up 77GB.
There are 18,209,495 runs in the dataset. It takes 6 minutes to run a count()
query on the DB, on my SSD.
Of those, 1,667,229 are victories, for an overall winrate of 9%.
Number of runs by Ascension level:
Ascension | No. of runs |
---|---|
0 | 6,323,588 |
1 | 1,646,934 |
2 | 930,744 |
3 | 770,032 |
4 | 700,716 |
5 | 576,515 |
6 | 474,732 |
7 | 425,891 |
8 | 380,263 |
9 | 372,430 |
10 | 409,098 |
11 | 374,928 |
12 | 344,341 |
13 | 306,371 |
14 | 327,470 |
15 | 365,797 |
16 | 311,086 |
17 | 377,304 |
18 | 436,051 |
19 | 491,141 |
20 | 1,864,034 |
Slay the Spire came out on Android yesterday.
So, Iām late to the game but whatever. Playing on PS4. So far I have beaten the first 3 acts with the Ironclad. Itās super fulfilling to FINALLY smash Dunkin Donuts (I forget their actual name, but yāall know who I mean). I basically centered my strategy on causing as much vulnerability and weakness as possible and raised my strength dramatically. Lots of free anger attacks meant that I had multiple hands of two or three 0-cost cards that were doing like 70+ damage. Feed being in my hand meant that along the way my health was at like 110-ish, but I still went through a fairy in a bottle and a lizard tail in the build up of my strength.
My wife has gotten really into it and is focusing on the Silent. She hasnāt been able to beat the first three acts yet, although she can reliably get to the third act boss. Her poison strategy seems similar to mine with strength, but she keeps getting the time keeper. I think she got dunkin donuts once, and the weird bird-crow once, and those were super frustrating. Not having as much health, and not having the right randomly generated relics seems to be the issue mainly? Idk. Iām sure itāll happen eventually.
The defect is interesting, although I canāt quite get past act two. For some reason Iām having trouble plotting out how to best use his powers and make a good deck. It seems like a knightās move type thing, and unfortunately I have yet to hit on that perfect combination. Looking forward to trying out the watcher. Iām really liking the hints of story that creep in and the world that is being hinted at. I would be interested in that board game, once pandemic ends. Iām assuming it is kind of like the Bloodborne card game, if anyone played that. Itās cooperative, kind-of, but everyone is selfishly motivated and there is a winner amongst the players (plus itās an Eric Lang game, so itās pretty solid).
Anyway, I am continuing to soldier on, despite building frustration at the defect/my inability to use them appropriately. Anyone else have personal favorite characters/strategies that they like?
I play Slay the Spire regularly still as an idle pastime. Right now Iām working on getting past ascension 10 with each character.
Defect is by far the most fun for me. If you can build a machine that churns out orbs and/or focus, and have a way to defend yourself long enough to spin it up, it is a riot. But, one mistake or a slightly sub-optimal card distribution and it goes south fast. Like, from full health and a wicked combo to dead against a regular enemy in act two or three.
I play Defect extremely aggressively in act one, focusing solely on how Iāll survive act 3 and beat the bosses. That means I tend to either die in act 1, or make it all the way to the end, with little in-between.
Yeah, that seems to be my issue; itās very swing-y but fun to play with! I think my frustration is mainly getting into act 2 and just being thieved to death immediately and sad ha ha.
Idk if ascension 10+ is in the cards for me, but Iām definitely having fun right now figuring out the patterns and the cards that work and seeing where they break down against certain enemies. I feel like itās a brainfeel that I havenāt had for a while with a videogame. Maybe red alert or warcraft 3-ish era game? I mean, obviously I know itās a different type of game, it just gives a similar brainfeel when analyzing the strategy that isnāt similar to other videogames for me.
The first strategy tips I needed to learn for Slay the Spire were to remember all the basic fundamentals of CCG and deck building games. Remember, Slay the Spire was made by people from the Android: Netrunner community.
This means, donāt take a card every time. You can, and should, skip. Destroy those junk cards from your deck. Destroying crap cards is usually more powerful than adding strong cards. Get those synergies rolling.
The second thing is just learning and memorizing the different enemies, how they behave, and where they appear. This just comes with playing a lot. Also, you can see the final boss at the start of act 3. If you got unlucky and have a shiv deck against time boi, or have a power deck against the cult, you have all of act 3 to work on another game plan.
This site is very helpful - https://spirelogs.com/
And just watch other people play on Twitch or YouTube to learn from them.
Oh absolutely; with the ironclad, I just zeroed in on maybe 3 or 4 specific cards and if it wasnāt one of those, I just skipped. Itās very tempting to take more, but Iād rather have more copies of effective cards that cycle quickly than an āevery eventualityā deck haha.
Yeah, the defect is definitely making me take my time and analyze WAY more than the ironclad or silent did. Iām getting further and further more often than not, just much slower than with the other two.
Iāll check out that site! Thanks!
Heal the Hallway, the sequel in which you administer first aid after some jerk came through, beat everyone up, and tipped over the building.
I learned about this video by Forgotten Arbiter, the āis every seed winnableā guy. A 7000+ score Watcher run.
It has all the RNG manipulation you would expect, plus 2 main ways it differs from a normal ascension 20 heart kill, to really jack the score up:
The abuse goes like this:
It ends up with 3300 cards in the deck.
He calculates the theoretical maximum score is 7893, using Ironclad. Winged boots can get you up to 16 elite fights, and Ironclad has more cards available, so more collector bonuses.
One thing the āwinnable seedā doesnāt consider is āwinnable run.ā
The latter is a subset of the former, though the Venn diagram is probably very close to the circle.
E.g., a forced elite battle might be seed-winnable technically (against, say, Lagavulin with The Silent), but your draw from the early limited deck is maximally inopportune in that battle.
Some seeds are only āwinnableā with RNG manipulation and foreknowledge.
An even less winnable seed, with mathy proofs:
Theyāre doing CUDA now