Balatro Calculator

[dark pier]

https://efhiii.github.io/balatro-calculator/

I've been working on a Balatro Calculator where you can go to the webpage, put in any hand + jokers, and it'll calculate the score.
It also optionally will look for the optimal order for the cards/jokers to maximize the score.

Please let me know (either here or via DM) about:

• questions about the calculator
• bugs you find with it / incorrect calculation
• changes / additions you'd like to see with it

[bright cloud]

I would love the addition of the plasma deck

[dark pier]

I haven't played with the Plasma deck yet, nor do I think I have unlocked it, but I'll look into adding a toggle for it.

I just looked into it; should be easy to add.

[bright cloud]

That would be nice since going into high score endless runs with plasma is so much easier

Also I think @junior cairn would be great here with his knowledge of the game, if anyone could spot something wrong it would be him

And while I know negative jokers don't have any effect on score calculations besides the base joker effect I think visually it would be nice if you could fully match the joker setup you have

Also some edge cases I noticed here that need to be added: first of all a money score for bull/bootstraps and planet cards for observatory voucher giving 1.5x per planet card

Also decaying and scaling jokers like popcorn/ice cream or square joker

And ceremonial dagger

Sorry for the incoming spam of messages but I'll try to spot all the jokers with changing values I can see

Well you know what I won't spam but instead just try and find all the jokers with values that can change

[restive jolt]

Misprint goes to 23 btw

u have it listed as 0-20

[hushed cove]

This is pretty neat

The optimizer doesn't seem to move cardd sharp if its not active (I also am not sure how to activate it)

I was mostly messing around but here

where it would be optmial to have the normal joker before card sharp

[restive jolt]

ye, litteraly just about to put the same thing

[dark pier]

I think it's live and working.

[dark pier]

The value UI controls the state for all the jokers on the right, so change the value to 1 by clicking + and then click the joker.

[junior cairn]

not an error, but...

[dark pier]

fixed, thanks.

[junior cairn]

wanted to try the WR strat on this calculator, but Baron is calculated based on played kings, not held in hand

calculator also appears to crash when the numbers get too big

when I try to place the second RPG 2 of hearts it bugs out

[hushed cove]

oh I see

also a joker search would be nice (even like a jokers.filter(j => j.name.toLowerCase().includes(search.toLowerCase())) would be nice)

[dark pier]

The bug was a bit different than that. Should be fixed. (wait a minute and refresh the page)

[junior cairn]

Fixed

[dark pier]

fixed

[versed roost]

Creat

Chipps

Retriger

that should be all

[dark pier]

Thanks, those should all be fixed now.

[dark pier]

Added.
It checks both the name and description of cards.

[narrow arch]

For the joker "Crafty Joker" nothing happens. When I open the console, I got :

[dark pier]

Aparently I typoed "Flush" as "FLush". That's it.
Should be fixed now.

[narrow arch]

When I reload the page, I obtain the same error.

[junior cairn]

It could be cached

[dark pier]

refresh again.
CTRL+SHIFT+R to hard-refresh (in chrome at least) but normal refresh should be enough.

[narrow arch]

Yes, that works. Thanks 🙂 ❤️

[dark pier]

You probably just refreshed before it was actually live; it takes a minute for Jekyll to update things.

[dense elk]

Two suggestions: Add text saying what the played hand is (ex: full house), and being able to edit jokers you already added by clicking on them (to add an edition or change the value).

It's really nice though!

[dark pier]

Yeah, I can do both of those.

[dark pier]

I added the name of played hand, I'll hold off on the "editing jokers when clicked" since it's a little more complicated.

[bright cloud]

also i think some of the scaling is wrong, a level 2 straight is 60x6 and not 50x7

[golden sluice]

Jokers don't have negative values anymore

Also unless it got changed, this is incorrect

#🕹・game-discussion message

Holo and foil trigger before joker effect (if it wasn't changed)

Polychrome after joker effect

Does not apply to playing cards

This is what I believe the complete order for stuff to trigger (but I'm not 100% sure about editions and seals order, does not usually matter tho)
#1202386969867595827 message

[dark pier]

Thanks. I was pretty sure I had some order-of-operations wrong since I did it just by memory, and was going to look into it this evening.
I applied the changes you said, let me know if you find any more that are wrong. I only just implemented the breakdown screen last night.

[bright cloud]

Also the planet scaling is wrong

I said it in the chat before but idk if it was fixed

[dark pier]

The scaling I used came from the source code, I guess I got it from the wrong spot?

[bright cloud]

I mean all I know is that I had a level 2 straight as 50x7 while it's 60x6

[golden sluice]

30x4 base
+30 +2 saturn

not sure what it is in calculator

[bright cloud]

I remember someone posted an image

[dark pier]

Okay, I just updated it to be in line with https://balatro.wiki/en/Planet-Cards

[bright cloud]

I'll try to make a list of things that need to be added since they can add score when I get back home

Like observatory voucher is the simplest example

Pair that up with perkeo and you can make use of that

[warped walrus]

i suggest that when Plasma Deck is enabled show the “balanced” values as well

[dark pier]

It shows it in the breakdown

[warped walrus]

ooh i see

[hybrid pewter]

Very impressive!

[dark pier]

added

[bright cloud]

very nice

another suggestion i have is being able to change the value to a specified value, not only being able to up it by one

and maybe being able to change per joker

[dark pier]

Yeah, those are both on my TODO list, lower priority since they are QoL rather than functionality.

[bright cloud]

yeah aight if its on the list ill move on

also idk if it really matters but like i thought about swashbuckler and liquidation, since it makes jokers buy price lower it also makes their sell value lower so its also a nerf

swashbuckler with value changing jokers is the basic category

like swashbuckler and egg or gift card

[dark pier]

Right now, Swashbuckler doesn't automatically calculate its mult, but it'd probably be nice if it did.

[bright cloud]

wait i guess you can just input the final value into the site

not really a problem but rather something players will have to keep track of

also misprint, it shows the range on the card but doesnt show the range of possible score like how lucky cards do

same with copying it

[dark pier]

Could you give an example?

It's possible you encountered a different bug which caused it to break (there would be an error in the console)
and refreshing would fix it.

[bright cloud]

ok now it works

yeah idk why it happened

[dark pier]

added (also for planets, though not hands because of the "lvl." part of the text. Should I just remove the "lvl." and make it text-editable? and hands)

[warped walrus]

you can make the “lvl.” part show up after input

like when inputting you only see the numeric value

[dark pier]

I think I know a way I could make that work.

[warped walrus]

it’s definitely possible but im not expert in coding field so

[dark pier]

there

[warped walrus]

nice

[static aurora]

i may have missed it, but if the function doesn't exist then support for negative jokers would be nice. matters for stencil and swash at least

[junior cairn]

Speaking of Swash you could also have a way to edit the sell value

[dark pier]

Both of those, you put the multiplier in for manually.
Stencil, it's easy to just do yourself. Swashbuckler, less so (since it depends on joker order), I'll add negative jokers if I automate his multiplier.

[dark pier]

What would be the order of operations here?

[junior cairn]

Uhh idk

If you have Steamodded then Debug Mode is a great way to set it up in-game https://discord.com/channels/1116389027176787968/1209857245316255744

[dark pier]

I do not

[static aurora]

makes sense

[dark pier]

How do you give enhancements or stamps (namely steel & red seal) to cards in the debug mode?

[junior cairn]

Spawn Chariot and Deja Vu from the collection and use them

@dark pier

[dark pier]

Okay, it goes:

• cards played (left to right)
• cards held (left to right)
  • steel (can repeat)
  • shoot the moon (can repeat)
  • red seal (repeats others; can not repeat)
  • mime (repeats others; can not repeat)
• jokers (left to right)

[junior cairn]

Shoot the moon applying before steel is kinda sad

[static aurora]

so, another tricky thing: shoot the moon executes before the xmult does visually, but in the code it happens second. there may be inconsistencies

im not sure if it makes score display incorrectly or what

(to verify, check the order h_mult and h_x_mult execute in eval_card and card_eval_status_text in common_events)

[dark pier]

Here?

    if context.cardarea == G.hand then
        local h_mult = card:get_chip_h_mult()
        if h_mult > 0 then 
            ret.h_mult = h_mult
        end

        local h_x_mult = card:get_chip_h_x_mult()
        if h_x_mult > 0 then 
            ret.x_mult = h_x_mult
        end

        local jokers = card:calculate_joker(context)
        if jokers then 
            ret.jokers = jokers
        end
    end

[static aurora]

yeah, mult seems to be added first? but it definitely displays afterward

i ended up checking yesterday after i had a steel queen/shoot the moon run

[dark pier]

I just went through frame by frame, and cross referenced with new code in my calculator, and it's definitely calculating it with shoot the moon first.

The final round score is consistant with that.

[golden sluice]

isn't this what I already sent with my message above 😭

I was 100% sure about that one - that's exactly how it happens for played cards, so I don't see why would it be different for held in hand stuff

the only thing I am not 100% sure are seals, but that shouldnt matter

[dark pier]

It was helpful to work through it and re-figure it out myself.

[bright cloud]

Another suggestion I have is to the actual hand in observatory

[dark pier]

As apposed to the planet names?
I felt like planet names made it more clear what the numbers are for.

[bright cloud]

You can have both

[dark pier]

How would you suggest having both?

[bright cloud]

Saturn (Straight) for example

[dark pier]

That looks clunky to me though

[bright cloud]

Seems fine to me

Oh it's saturn not venus

Another reason to have hand names

[dark pier]

I guess this is acceptable.

[dark pier]

Low-ace straights, four-card straights & flushes, and 4-card straight flushes weren't implemented correctly before, but now they should be:

[bright cloud]

I wonder if anyone here knows rust or any fast coding language to make the site into a fast application, that way maybe the optimization part could be done faster and not slow down your browser to a crawl when dealing with really high scores

It's just js isn't a fast enough language I think for all the calculations people might wanna do

[dark pier]

I could multi-thread it, which would be maybe a 10x speed boost (and would get around blocking the UI), but language isn't really going to matter much when the time complexity is O(n!).
Algorithmic improvements are far more important.

[dark pier]

As an example of an algorithmic optimization I'd like to implement, right now, when optimizing Joker order, it tries every possible ordering, but that could be minimized a lot given that the order of many jokers doesn't matter outside of some edge cases where another joker's position relative to it matters. In many cases, just putting x_mult jokers on the right is the only thing to optimize, and that can be optimized easily via sorting them like that.

[bright cloud]

Well good to know it isn't language related problems

[bright cloud]

another suggestion i have is to show the effect you have on a joker while hovering

and on the cards

[dark pier]

Baseball should be working correctly now.

[versed roost]

found a bug with the boots

i put in a 53 for it to simulate the +106 someone got with their hand, but looking at the breakdown it seems to be giving +108 for some reason

[dark pier]

I was adding 2 for some reason. Should be fixed now.

[versed roost]

👍

[somber arch]

I'm just popping in here to say thank you for this <3

The calculator is really useful <3

[dark pier]

Editing jokers when clicking on them is finally implemented.
This should be a significant QoL improvement.

This also means that clicking jokers no longer removes them. Instead, on hover, you get a red tab which, when clicked, removes the joker.

[dark pier]

The next things I'm planning to implement are:

• modifier tooltips
• boss blinds
  • The Flint
  • disabled joker modifier (Crimson Heart)
  • force card to be played (Cerulean Bell)
  • disallow hand types (The Mouth/They Eye or when building for Supernova)

[bright cloud]

a possible suggestion from the high score nerd in me: to add the score you would get even if you went past infinity. idk how hard that would be on the system to keep the score written down when already dealing with such higher numbers but if its not something that effects performance that would be nice

[junior cairn]

I think JS supports big integers natively so it might not be that hard to do

[bright cloud]

well lets hope

[dark pier]

It's on my todo list already, but I can't imagine it wouldn't effect performance.

[frank granite]

epic

[bright cloud]

cant imagine it wouldnt meaning it would?

[dark pier]

Yes

[bright cloud]

big rip

[dark pier]

It don't know by how much, and it could be a toggle/falback-only.

[bright cloud]

i mean a toggle would probably be good

theres no need for it for 99% of cases

only in extreme cases

[full phoenix]

@dark pier made a PR to fix a bug :)

[dark pier]

Thanks. Merged.

[full phoenix]

Just a silly little typo squashed
Will definitely try breaking it some more

[dark pier]

Please do try to break it more.

[bright cloud]

random thing i noticed: while i dont think it matters too much popcorn and ice cream overflow into negative numbers when the value number is above their actual lifespan

[dark pier]

Yeah. I know. It's not something I've bothered fixing just because it's such a minor oversight, but it'd be nice if it just bottomed out at 0 or something elegant like that.

[bright cloud]

yeah it aint too serious so makes sense

[bright cloud]

I think the balatro wiki isn't up to date since I just noticed flush 5 isn't scaling properly, level 9 is 480x40 not 430x38

@junior cairn can you just find the scaling from the game code in an easy way to paste?

[dark pier]

You can figure it out pretty easily with the debug mode

[bright cloud]

I'm not on my computer

So I'm asking him to help

[dark pier]

(weird choice of place to ask, but makes sense besides that)

[bright cloud]

Why would it be weird, the main use of a table like that is in a calculator site if anything

[dark pier]

I think all of these are correct (except maybe flush 5, flush house, or five of a kind):

I don't think I tested flush 5, flush house, or five of a kind, but I'm pretty sure I did for the rest

[junior cairn]

Not on my PC either atm

And won’t be for a bit

[bright cloud]

Rip

[swift barn]

can you add some sort of "reset" button? i usually only use it on bosses, so by the time i want to use it again, im already in completely different setup and clicking on everything is no ideal

also, how am i supposed to setup ramen? figured out, but im not sure if doing it by number of discards is better or no

walkie talkie?

[swift barn]

another suggestion - add some way to compare the jokers, i want to test whether scholar or fib would be better (assuming perfect hand), but rn there is no fast way to check it. either "disable" joker, or some way to weight pair of jokers
*fib doesnt work aswell

[dark pier]

I'm already in the process of adding two features right this second that would help with that:

1. disable joker (lmao, already done, just not published)
2. URL parameter of hand (so you could quickly just open a second tab pre-loaded to be the same)

[dark pier]

As for resetting, you can just refresh the page to reset everything.

[frank harness]

Does this add a draw chance for specific cards?

Like % chance to draw 9 of clubs, or % chance to draw any club or any 9

[dark pier]

No. This calculator only considers what you currently have in your hand; not what you have in your deck.

If you want the chance to draw a specific card/set of cards, you need to look at the cards remaining in your deck, how many cards would satisfy what you want, and how many cards you're going to draw.

I could probably make a different calculator which, given a deck-state and hand-state could tell you the odds of drawing various types of hands with various discards, but it'd be a very different kind of thing.

[frank harness]

I just am bit probability nerd, so i do that calculation in my head all day, but it does get tedious doing it every draw

[dark pier]

As a general rule of thumb; just count how many cards would satisfy what you want (if you're looking for a single card) and divide that number by the number of cards remaining in your deck. That's your A in B chance of drawing it. You do A / B to get the % chance of drawing the card. If you're looking for multiple cards, a rough approximation is to do that same thing for both types of cards and multiply the odds; e.g. 10% x 20% = 2%. Keep in mind that if there's overlap, e.g. you're looking for 2 Aces, you can't draw the same card twice so subtract 1 from the second count.

[dark pier]

Oh, then that's probably redundant information for you...

[frank harness]

useful info nontheless

even if not for me

i even attempt to do game wide ones in my head

like if i add any card to my hand what boost that does to my deck in the long run ect

[dark pier]

I personally do that part almost all by intuition. I don't bother doing much more than checking how many cards there are to satisfy a certain hand.

[dark pier]

The URL now updates with a query parameter that contains the information of the current hand setup, which can be copied into a new tab to create the same setup or shared with others for a faster exchange of ideas. (other improvements too, like disabling jokers, and the Flint)
https://efhiii.github.io/balatro-calculator/?hand=oADxiuW4AhD0QQAYAgADjPRdCAEgFg

I did what I could to try and make it so that the URL doesn't get too long.

[dark pier]

walkie talkie should be fixed as well.

[frank harness]

have you been working with the game itself with this or just seperately with a web app?

im currently trying to modify the tooltip to show a chance %

[dark pier]

I have looked at the game's code a little, but haven't messed with it besides using the debug mod.

[frank harness]

ive been studying the code for the last few hours trying to work out where/how i would mess with the tooltips

[lucid sparrow]

flush five base score is 160x16 and the planet gives +40 chips and +3 mult

[dark pier]

I made it so that for Raised Fist, it puts the lowest cards on the left (with preference for red stamps & red stamps with steel being on the right so that they trigger twice) so that Raised Fist triggers first, and it doesn't do so if you don't have raised fist so that Shoot the Moon can gets maximum effect from steel cards:

P.S. Raised Fist triggers (as far as I can tell) the right-most of the cards of the lowest value.

btw, I was already putting queens on the left of kings and aces for Shoot the Moon.

[dark pier]

Wait, I just realized that the one with Raised Fist disabled gets a higher score... I guess the two steel 4s getting to activate after Shoot the Moon is more valuable in that case than Raised Fist triggering before the other 3 steel cards. 😞

I think it'd be ideal if I did it the other way around; doing queens on the left, then low card(s), then the rest.

brb

much better
https://cdn.discordapp.com/attachments/563935472632725546/1213964061843394660/image.png?ex=65f76322&is=65e4ee22&hm=e3744bcc6d36ad5e3965327ff5edfb1d0b8bf049836d735c3ee00f34d810f02f&

[wind merlin]

godspeed edward, thank you for this wonderful creation

[exotic berry]

My goodness, @dark pier. This is absolutely incredible!

[dark pier]

Just a PSA that in the event of random events, the optimizer might not choose the actual best option; it'll choose the option with the highest worst-case, but the highest average case might be something else:
https://efhiii.github.io/balatro-calculator/?hand=oACMOJQACOMAECgAKag1BKA0BIA

This also made me notice a slight mistake with Sock and Buskin triggering too early

[full phoenix]

Tbh the average isn't a great metric either, is it ?
A median would be a better indicator since it wouldn't be skewed by higher scores

Tho it wouldn't be nice to compute especially for all the combinations

[dark pier]

When I said average, what I was thinking was really EV (expected value) which is basically the same thing as the mean weighted by how probable an outcome is.
The median would be the thing you want to optimize for a single hand, but the mean (EV) is what you'd want to optimize for many hands. For something in-between (the 3-6 hands you generally have), the actual thing to optimize is probably somewhere in-between 🤔.

[restive jolt]

maybe counting the random events averaging it?

[dark pier]

Oops! All 6s is annoying to have to consider. (especially since it's blueprint compatible)

[restive jolt]

true

[full phoenix]

Mhhh i mean to be fair I guess both indicators should be available

[dark pier]

For labeling them; does it make sense to call them: Long-term EV and Short-term EV for mean and median respectively? I imagine most people don't know or remember the difference between mean and median.

[full phoenix]

I think that's rather good names 👌

[bright cloud]

I tested today with brainstorm and it wasn't compatible

[dark pier]

Now I have to change things from having 1 optimizer (minimax) to having 4 (minimax, meanmax, medianmax, and maximax)

[golden sluice]

Wait what why is it blueprint compatible

[bright cloud]

Is it really blueprint compatible but not brainstorm?

I don't think it is

[dark pier]

People claimed it was in the discord, maybe they just assumed without testing?

[golden sluice]

Pretty sure that's outside the effects blueprint would affect

[bright cloud]

I'll test blueprint but I'm certain brainstorm

[full phoenix]

Well i hope its not too complex, i would try to help next week since I'll be on vacation

👍

[dark pier]

It'd be weird if it was the only card that is blueprint compatible and not brainstorm compatible

[dark pier]

There's also, IIRC, situations where you want the minimum payout, so a minimum variation of those might also be desirable in niche cases;
(minimin, meanmin, medianmin, and maximin)

[hushed cove]

ooh maybe I could make a thing on my save loader to make a url for the current save state

[dark pier]

That'd be epic, I can describe to you how it works; it's a bit complex because I wanted the URLs to be as short as possible.

[hushed cove]

I'm reading the comment and it seems to make sense

also my thing is also in js so I coild probaly just yoink the js file and use your code

I guess I need to see how jokers are formatted

[dark pier]

Each joker is indexed by x and y coordinate in the spritesheet image, joker value is what determines state and you can see that value in the web-app

[dark pier]

I had forgotten that I actually commented that bit; most of my code is pretty lacking in the comments department. 🙈

[hushed cove]

Yeah I feel that

Looks like I'll have to make some id map, but that shouldn't be too bad

and then I'll have to convert how things are storing their value

a few values seem like they might be a bit difficult

like card sharp

[dark pier]

The value for Card sharp is just a boolean; you can probably just default it to false until/unless you want to actually check for if it's enabled.

[hushed cove]

well the issue is I don't know what hand will be played

so I can't know if it should be enabled

[dark pier]

and neither do I. it's one of a few cards that's not implemented properly due to missing information.

[hushed cove]

yeah

the user being able to toggle it without readding it would be nice

oh wait did you add that?

oh you did

I missed that

in that case its not a big deal

the user can do that

[dark pier]

Swashbuckler and Obelisk are the other two that aren't really implemented correctly

[hushed cove]

yeah swashbuckler would be nice

if you implement sell value, the save reader could export that as well

same with plays on each hand

[dark pier]

I did not because Gift card is horrible

[hushed cove]

I mean iirc the discount voucher can even change sell values

and modifiers

its annoying

but it might be nice to be able to add it in the edit/add menu

just for when using swashbucler

side not it might be smart to add a version field to your binary format

so you can add stuff to it without breaking loading older url's

[dark pier]

Yeah, I strongly considered that; for now, I'll just change it to a different query parameter name if I have to make a breaking change.

[hushed cove]

yeah fair enough

but yeah you could keep backwards compat with an version field, just don't do the new stuff when parsing

otherwise, looking good

I guess you can still append data to the end of the format now

[dark pier]

indeed

[hushed cove]

ok I will see if I can make a quick proof of concept for this

[dark pier]

Okay, I guess blueprint isn't compatible (maybe it used to be?) but you can have multiple Opps! All 6s and they stack.

[bright cloud]

Yeah I know they stack but they can't be copied

[dark pier]

news to me

[hushed cove]

can confirm multiple oops all 6's stack

(they double each other's values)

[hushed cove]

just want to confirm, number of cards is cards held - cards selected , and number of cards in hand is cards selected, right?

also seems a bit odd you push the size of the hand then the size of the seleected but then you push the selected first and the hand second

[dark pier]

I just did some back-of-the-napkin math and, to find the median of this would require checking 2^26 = 67,108,864 possible outcomes. For context, calculating 40,000 takes over a second, lets call it 1 second, so that'd be roughly 27 minutes per hand-calculation plus the optimizer does upward of a few hundred hand-calculations per hand when one optimizer is enabled and upward of tens of thousands when both are enabled.

TLDR; I don't think we're getting a mean/median optimizer.

[bright cloud]

Let's first start with performance before we start calculating random effects

Like I ain't trying to rush or anything but imo it would be better to work on multithreading for example

Rather than the median of lucky card or something

[dark pier]

Fine, I'll do the boring thing and implement multithreading
grumble grumble programmer grumble

[hushed cove]

good news

I think cards should work too, but I wasn't in the right state to check

right now any custom values aren't implemented

[dark pier]

I like that you have 1 foil card mixed in in the middle

[hushed cove]

yeah I was just buying things in the shop

is nice it worked though

I will mess with it a bit more later

theres still a few todos

yeah cards seem to work

[bright cloud]

I mean it might be boring but I think most users who will actually want to get a score calculation will want the extra performance needed to get scores for high score runs

[dark pier]

I've found that the performance right now is more than good enough so long as you aren't trying to optimize both hand and jokers at the same time, but I said I'll work on it, I know how to do it, I'll work on it.

[bright cloud]

I mean it definitely tanks when I try calculating near inf scores

And I know that's the extreme but I think that's a use case that will be used more than you think

[dark pier]

send me a link

[bright cloud]

I mean I don't have links but it's easy to make

Just like 4 barons and 4 mimes and a bunch of red steel kings

[dark pier]

with that strategy, if you're trying to "optimize" the cards, you're looking at over a billion different hands (which are all one of 6 hands btw so really redundant) so of course it's slow; multithreading won't help in that case. What would help, though, is pruning duplicate hands from the calculation.

[bright cloud]

I mean I think performance optimization in general should be put first

Instead of such niche cases

I ain't a dev so I ain't gonna judge on the pace but it's just my 2 cents

[dark pier]

In general, multithreading is going to get you like one or two more cards. algorithmic improvements are going to either be unhelpful because the improvement isn't universal or orders of magnitude faster.

They're both worth doing.

[lunar topaz]

it doesn't correctly count the midas mask + vampire + pareidolia combo

[dark pier]

It doesn't take into account gold enhancements in general. I didn't think about that vampire effects them. That's annoying.

[lunar topaz]

Card Sharp isn't capitalized correctly

also why is 'Tarot' in superposition's description that color

[bright cloud]

Like I think if people want to help more to the site right now besides like finding cases where score isn't being calculated correctly (like vampire right now) i think they also need to try and "stress test" the site for different ways to help make it perform better under different situations

[lunar topaz]

if you use many blueprints and split them evenly among a retriggerer joker and a scoring joker it takes a noticeable amount of time to calculate

[dark pier]

Well yeah, scoring takes time & retriggering takes time; if the normal situation is 5 cards scored by 5 jokers for one trigger = 25 scorings, doing 5 cards scored by 5 jokers for 5 triggers = 125 scorings, it's going to take about 5 times longer, not to mention that blueprints have overheard in finding who they represent (that bit might be able to be done as a one-time thing per calculation but isn't a high priority).

[dark pier]

A lot of the lesser-used colors (words for suit colors, spectral, uncommon, planet, and tarot) I had just used approximations and never went in and made them color-accurate. I just went in and did that so next time I publish, they'll be the same as the in-game colors.

[dark pier]

It should now

[dark pier]

number of cards is the total number of cards. My thinking when doing number of cards first, the number of cards in hand, then pushing cards in hand, then pushing cards not in hand is that cards in hand is nested inside of the total cards:

all cards
  cards in hand
  cards not in hand (quantity can be implied by all-cards)
end  
```But I can see why it might seem like a weird order.

[dark pier]

Okay, sell value is now something that can be set and Swasbuckler takes that into account rather than a set value. I haven't put it in the URL thing yet though.

What are the default sell values of cards?

[rigid bone]

is there a way to input hiker chip bonuses on cards?

[dark pier]

no, I keep forgetting hiker even exists. You can currently approximate via putting bull as your first joker with whatever the sum of bonuses would be; it won't calculate quite the same though.

[hushed cove]

the cost of jokers is in the main game.lua where they are defined, and the sell cost is math.max(1, math.floor(self.cost/2)) I think

although iirc, the cost can change

like the discount voucher changes the sell cost

[dark pier]

Yes, the cost can change, but having a default value that's at least right some of the time is better than nothing.

[hushed cove]

I think this should be the defaults for each joker

being the buy cost not the sell cost

it seems inconsistant across rarieties

with the exception of legendaries

who are all 20 cost (10 sell cost)

[dark pier]

That's like 40 out of 150 jokers.

[dark pier]

Okay, I guess I was looking at the demo's code on accident. That explains why I kept finding discrepencies.

[hushed cove]

oh yeah that would do it

btw do you have a list of jokers your thing supports that support custom values?

[dark pier]

technically, they all do.

[hushed cove]

fun

do all of the ones that "actually" use it have it in their decription?

[dark pier]

I believe so; I was just getting a list via checking which use it in their description:
Stone Joker, Acrobat, Banner, Mystic Summit, Loyalty Card, Steel Joker, Glass Joker, Abstract Joker, Wee Joker, Supernova, Joker Stencil, Ceremonial Dagger, Fortune Teller, Hit the Road, Ride the Bus, Driver's License, Invisible Joker, Dusk, Throwback, The Idol, Satellite, Caino, Yorick, Bootstraps, Egg, Runner, Ice Cream, Blue Joker, Constellation, Green Joker, Card Sharp, Red Card, Madness, Square Joker, Vampire, Hologram, Cloud 9, Rocket, Obelisk, Erosion, Mail-In Rebate, Lucky Cat, Bull, Flash Card, Popcorn, Ramen, Seltzer, Spare Trousers, Campfire, Ancient Joker, Castle

[hushed cove]

ok that'll work

[dark pier]

Sell value is in the URL now; omitted when it's a default value but you can include it 100% of the time for simplicity if you want.
https://github.com/EFHIII/balatro-calculator/blob/5baa41860ccf1796adfee51fbd6f4907747f6e84/hand-url.js#L157

[hushed cove]

sounds good

I'll have to check how the save does it but I think it just tells me the sell value

[dark pier]

new url query parameter is h=, the old one is backwards compatible via using hand=

[hushed cove]

ok

also I think theres something weird with your index map for the jokers that have seperrate head and backgrounds

probably a missmatch between what you use for type and what the game uses

theres only 6 of them so I'll just fix it manually

[dark pier]

Yeah, I use the background as the canonical position, I guess they're using the head as the canonical position.

[hushed cove]

yeah I guess

I just copied from the code and did a bit of find replacing

[dark pier]

You found my WIP fallback, lol

[hushed cove]

imma put this here so I don't lose it js [ 'j_stone', 'j_acrobat', 'j_banner', 'j_mystic_summit', 'j_loyalty_card', 'j_steel_joker', 'j_glass', 'j_abstract', 'j_scholar', 'j_supernova', 'j_stencil', 'j_ceremonial', 'j_fortune_teller', 'j_hit_the_road', 'j_ride_the_bus', 'j_drivers_license', 'j_invisible', 'j_dusk', 'j_throwback', 'j_idol', 'j_satellite', 'j_caino', 'j_yorick', 'j_bootstraps', 'j_egg', 'j_runner', 'j_ice_cream', 'j_blue_joker', 'j_constellation', 'j_green_joker', 'j_card_sharp', 'j_red_card', 'j_madness', 'j_square', 'j_vampire', 'j_hologram', 'j_cloud_9', 'j_rocket', 'j_obelisk', 'j_erosion', 'j_mail', 'j_lucky_cat', 'j_bull', 'j_flash', 'j_popcorn', 'j_ramen', 'j_selzer', 'j_trousers', 'j_campfire', 'j_ancient', 'j_castle' ]

list of ones that use abilities

wait why is egg in here

with the sell value being supported, it shouldn't need that anymore

We've gotten out first joker value transfered over

oh ice cream has the (Currently +90 chips) but it should just say +90 chips at the top (like popcorn)

[dark pier]

It's a visual thing only

I put it there more to remind me that egg is anoying

[hushed cove]

ok fair enough

[dark pier]

I've removed the text from egg and fixed the text for ice cream

[hushed cove]

👍

man this game is inconsistant with how it handles custom values

[dark pier]

That doesn't surprise me. I tried to normalize how I interpreted them as much as I could.

[hushed cove]

yeah

[hushed cove]

you might need to fix it again

also bruh

oh bruh the planets even have a spot for total usage where the tarots have

[dark pier]

ehh, I forgot to delete a dollar sign. fixed now.

[dark pier]

It's just trying to be error resistant

[hushed cove]

No I meant the one uses the global varible and the other loops over and counts

[hushed cove]

ok a couple things

1. queens seem to be sorted as the highest cards, when they should be between kind and jack

2. I can't seem to select any kind of modifiers in the card GUI

[hushed cove]

if I zoom out a bit they show up but then when I zoom further out or back in they dissapear again

[hushed cove]

This says +mult but its actually +chips (it does the math right though

theres a slight inconsistancy here

let me investiagte

oh I think I made a mistake

oh it was floating point aritmatic

[dark pier]

That's for Baron, I just didn't bother to not do it when Baron isn't there.

[hushed cove]

why would you do it that way with baron?

[dark pier]

err not baron

Shoot the moon

[hushed cove]

ah

so more like the baron shoot the moon combo?

[dark pier]

Or just shoot the moon with steel aces/kings

[hushed cove]

okay makes sense

[dark pier]

Raised fist makes things more complicated to optimize though, which was annoying.

[dark pier]

It's not even about floating point; just the multiplier number it shows you is rounded, but in the math it uses the exact value.

[hushed cove]

no it was a floating point in my code

(joker.ability.x_mult - 1.0) / 0.1 when x_mult is 1.2 this is 1.9999999999999996

I then just converted that to an int which is 1 and not 2

[dark pier]

If you change the order of operations, you can get it to come out to 2 without rounding

[hushed cove]

ah thats smart

I just stuck a round on the function calling all of the conversions just in case

[dark pier]

That relies on the 1.2 not having floating point error already though, which might not be safe.

[hushed cove]

I belive this value should be whatever the card is visably displaying so it should be fine

also is abstract something the calculator can do itself?

oh also are hiker upgraded cards something I can do?

[dark pier]

I chose to have it as a value just for if you're putting in only some of the jokers

[hushed cove]

I guess

[dark pier]

If you have some cards that you know are useless, it makes the optimizer faster to omit them.

[hushed cove]

oh how do I do supernova?

[hushed cove]

oh also this is wrong

[dark pier]

I forgot that supernova's not implemented properly. It's currently a static mult, but there's no UI to add times hands have been played.

[dark pier]

Yeah, I had just noticed

[dark pier]

I'll append hand counts to the end of the URL format, here's an outline:

/*
include hand counts? - 1 bit
for each hand type:
  played this round? - 1 bit
  is non-zero? - 1 bit
  count if non-zero - 16 bit unsigned int
*/
  // hand counts (for supernova/obelisk/card sharp) - 1 bit
  if(false) {
    binary.push(1);

    for(let i = 0; i < hands.length; i++) {
      // played this round? - 1 bit
      binary.push(hands[i].playedThisRound ? 0 : 1);

      // is non-zero? - 1 bit
      binary.push(hands[i].played ? 1 : 0);
      if(hands[i].played) {
        // count if non-zero - 16 bit unsigned int
        binary.push(...intToBinary(hands[i].played, 16));
      }
    }
  }
  else {
    binary.push(0);
  }

[hushed cove]

oh this will allow obelisk to work better too

also it might be nice to get a hand thing for card shaerp

[dark pier]

Added an extra bit in that same area to the spec above

[hushed cove]

Why is seltzer an active/deactive joker?

also hows castle work with it having a chip and suit?

[dark pier]

Because I didn't know that it would disappear once used up when I started working on this

[dark pier]

The suit is purely aesthetic

[hushed cove]

I guess

oh I just pushed my changes

everything should work now

https://balatro.shorty.systems/

its a tiny bit clunky to use, but I'll fix that later

its bed time

[bright cloud]

5oak: 120x12 not 100x10
flush house: 140x14 not 120x12
flush 5 is correct

@dark pier just tested all in game

[raven thorn]

i could see this being in a balatro sandbox thing

[bright cloud]

i wish we had a sandbox but it ended up not being done for full game

so now we make the sandbox

[raven thorn]

well that's where mods come in

[bright cloud]

i guess

[dark pier]

Yeah, flush 5 is the only one I tested

fixed

[dreamy crescent]

How hard would it to be to port the efforts here into an in-game score preview? That's definitely the single thing I find most missing from the game that I'd love to see modded in.

[dark pier]

As a lazy approach, it's probably possible to convert the necessary part of the calculator into a node.js library to be called by lua from within the game.

[hushed cove]

I mean I think a save state kind of system would be possible

Just save the game, play the hand then save the score and load it

I guess you don't have the optimization then

[dreamy crescent]

I don't care about the optimization feature; I care about constantly having to do the math myself. Specifically, I would want a preview of exactly how much score I'm about to get with my currently selected hand + the current jokers I have automatically show up when I select the hand. Would literally save me a good 20 minutes per run compared to constantly doing the math manually. (Randomized factors can just show min/max as this Calc currently does.)

[left lark]

Does this account for DNA with Brainstorm causing 2 of said cards being in your hand during scoring?

[hushed cove]

if you want you can try using my save editor which I just put some support for putting the data from the game in the calculator.
https://balatro.shorty.systems/

[dark pier]

no

I hadn't considered that niche scenario

[dark pier]

I'm basically rewriting the entire simulation code, so it's going to be a while (probably like a week) before I have any significant changes, but when it's done, I should have significant raw performance improvement, significant algorithmic performance improvement, multithreading, and Monte Carlo based mean/median calculation. I'll also be implementing a few of the things currently unimplemented while I'm at it.

[bright cloud]

Letsgo

Hype

[dark pier]

I'm testing to see what the performance hit for using a big-float (mantissa & exponent stored separately as float64) would be.
There's a JS library, break_infinity.js that does that, but I only really need add and multiply, so I was able to make a custom implementation that's about twice as fast as it:

Still about 2.5 times slower than float64 (while not breaking until around 10 ^ (9 x 10 ^ 15)), though that's doing exclusively math operations so the full simulation probably won't quite be 2.5 times slower.

[dark pier]

wat

My array of test values was getting too long; I cleaned that up, and it's normal again.
Making my version only occasionally normalize the mantissa sped things up a little.

[bright cloud]

I mean if we have all the other performance improvements I doubt the slowness would be that much of a problem

[dark pier]

Maybe, but I expect this to be one of the main possible bottlenecks, so I want to make sure it's optimized before it would be a lot more annoying to change.

[bright cloud]

Makes sense

[left lark]

In your rewrite could you have it show us the E value as well so we know "Yeah... thats e13", without me having to count?

[dark pier]

It already does that

[left lark]

oh okay, last time I used it it didn't.

[dark pier]

I think the threshold is just higher than it is in-game

I also use x 10^exp since that's easier to understand

[versed roost]

e is basically a replacement for x10^

[left lark]

I understand that. Its just last time I used it I recall it having shown the whole number

[dark pier]

As I said, I think the threshold is higher than in-game

[left lark]

Yeah I heard you. Was talking to nick

[versed roost]

yea the switch to e numbers happens at trillion ingame

actually wait

the lowest one is e11 i believe, so that would be 100 billion

[dark pier]

There, now the threshold is 1e11

[dark pier]

My code was so unnecessarily verbose

[raven thorn]

i wanna take a guess at what the if statement means

"if card[3] is equal to 5 and is not card[6]"

i only have decent coding knowledge lmao

[junior cairn]

I assume that list stores attributes of a card

I’d guess “if card is steel and not debuffed”

[raven thorn]

makes sense

this is way out of my field

though i've never seen a triple =

[dark pier]

In Javascript, single = means assignment, double = means loose equality, and triple = means strict equality.

[dark pier]

Yeah, card is an array with the indecies meaning:

[rank, suit, edition, enhancement, seal, extraChips, disabled]

so card[3] === 5 means card enhancement equals 5 and 5 happens to represent steel. Similarly card[6] means card disabled is falsey aka card not debuffed.

I realize the readability isn't great, but I think it's fine for this.

Some notable differences (besides the card format being a 1D array of numbers and a Boolean rather than an object of objects of Booleans or numbers) is the triggerCardInHand function only taking 1 parameter instead of 9 and the steel card check taking 2 conditions instead of 4 (one of which was a slow indexOf() check).

[junior cairn]

lol when making my seed searcher I had to completely rework a lot

[dark pier]

Yeah, performance probably matters even more for that kind of application.

[junior cairn]

There’s probably a lot you can optimize still for that

Wouldn’t be surprised if someone comes along and makes a pull request that triples the speed of the searcher or something like that

[dark pier]

I've been meaning to check out you seed searcher, but I don't really want to do seed searching until I've beaten most of the game and I only just beat my first run on gold stake today.

[junior cairn]

makes sense

If I wasn’t involved in the technical aspect of the game I would probably be playing a lot of runs for 100% completion

[dark pier]

I swear I've spent more time programming this calculator than I have playing the game.

[hushed cove]

Yeah I feel that

[dark pier]

Multiple times, I've been hanging out with friends who all are playing the game and I'm just programming in the background.

[hushed cove]

I mean at least I enjoy programming

[dark pier]

I love programming so much that I quit a programming job and left the software field so that work wouldn't ruin the fun for me.

[hushed cove]

wow

qctually I can check how much time I've been programming this thing

oh wait actually theres a bit more if I expand the search

[junior cairn]

If I were to estimate I’ve probably spent 30-60 hours on Immolate

Haven’t kept track though and it was across a very long period of time (since the beginning of January, and going back to October if you consider really old versions)

Like it started out in October as a program that looked at the screen to manually search seeds

A literal hold R simulator

[raven thorn]

lel

[dark pier]

Incase anyone's curious; one of the major algorithmic improvements I'm making is, for optimization searches where cards played and joker order are being optimized, right now it checks every combination and permutation in a fairly straightforward way of just checking their score, but what I'm working on changing is making it so that it precomputes as much as possible, so:
First it'll pre-compute everything that it can knowing only what jokers are played (not joker order, cards played, or cards played-order)
then it'll loop through all permutations of jokers and pre-compute what it can using that new information, e.g. pre-determining what card blueprint and brainstorm are copying and how much Swashbuckler's multiplier is
Then it'll loop through all n choose k combinations of choosing cards to be played and pre-compute what it can with that information, e.g. what type of hand it is, vampire, obelisk. It'll also pre-compute triggering cards-in-hand effects at this time.
Lastly, it'll go through all the permutations of the order to play cards and compute the final score.

I also want to prune redundant permutations, but I haven't spent too much time thinking about that bit.

All told, I'm expecting that to increase performance by multiple order of magnitude in some cases.

[raven thorn]

im ashamed i got jumpscared by your paragraph

[dark pier]

Hey, it calculated a thing

[hushed cove]

nice

[dark pier]

Only took 1000 lines of code with none of the jokers implemented (a lot of the structure is done though; jokers are basically the only thing left)

[raven thorn]

thing is calculated, seal of approval granted

[swift barn]

can someone help me with that? i would like to have that info, but i dont get how to get it

[full phoenix]

I wonder if it'd be possible to use, for example const ENHANCEMENT = 3; const STEEL = 5 at top level and then use card[ENHANCEMENT] === STEEL
would JS be able to optimise that ? 🤔

but yeah I guess having them commented all at one place is easy and doesn't make reviewing the code too bad

pet peeve material right here

[dark pier]

I'd say probably?
It'd definitely be slower, but probably only on the order of microseconds, and probably only at compile time.

I do have them all defined in-code:

[full phoenix]

I see
I did specifically use consts because I do not trust optimisation being done if we try getting them from an object, I guess the correct method would be to measure it

[dark pier]

Yeah. I'm definitely not going to be using them from the object. The objects are there because I'm writing this in a way that it can be used as an npm library independent of my calculator and the objects can be exported in a module.

[dark pier]

This is what part of test.js looks like:

And the card function just builds an array.

In this context, Balatro.suits.clubs being unoptimized doesn't matter because it's only run once compared to the potentially many, many, many times it's used.

[dark pier]

I don't think I wont to do this for joker names, but this should be fine.

I always find it a little jarring that the ranks are offset by two, but whatever.

[junior cairn]

I did this for literally everything in Immolate lol

Just stuck it in a massive enum

Also you forgot Ace

[dark pier]

I noticed right when I posted the image

There is a conflict in that disabled joker is a different index than disabled card, but their edition is the same index.

there

[dark pier]

I lied; I have a few jokers implemented

regex replace is great

[hushed cove]

if you use 7 zip, you can extract the games code. This is in the file game.lua

[dark pier]

Improved editions styling

[warped walrus]

looks nice

[hushed cove]

that's pretty good

oh yeah next time you make a breaking change in the url format you should add gold card's

[dark pier]

It's already WIP

[hushed cove]

👍

[dark pier]

IIRC, gold matters in the event of vampire and that's it.

[versed roost]

the only other time it would trigger when played is with the gold ticket and that just lets you get the money

[dark pier]

That doesn't effect score though.

[hushed cove]

Unless you have bull/bootstraps

[dark pier]

That's such a niche case, why

[hushed cove]

I mean theres a few other jokers that could increase money before bull/bootstraps

[versed roost]

I didn’t even think of that lol

[hushed cove]

also gold seal

[versed roost]

Lucky cards

[hushed cove]

I think business card

[dark pier]

I'm not implementing money tracking for the sake of this niche case.
At least not right now.

[hushed cove]

tbh money tracking would be nice so you change all bull and bootstraps at once

[dark pier]

That's not even a good build.

I'll maybe look into after the big update I'm working on, but not before.

[golden sluice]

bull is a good build wym, bootstraps can sometimes be useful for the lack of other scaling

[bright cloud]

Bull + bootstraps only run or bust

[fickle marten]

how do you activate dusk on this

nvm

i figured it out

i crashed the calc by accident

[dark pier]

You probably just gave it something that takes a super long time to run. I've yet to see it actually crash, just hang due to large calculations.

I'll try and make it harder for that to happen later, but just try and be mindful for the time being.

[fickle marten]

nah, it just straight up crashed

its fine, it makes sense

it also may just be my laptop being shitty

[dark pier]

The engine rewrite is nearing completion. It's fully functional for pretty much everything except some jokers

[dark pier]

That includes all of the improvements I mentioned here except for the "pruning redundant permutations".

[heady swan]

Idk if you've done this already, but if you make it an api, it wouldn't be too hard to incorporate this into a mod to show calculations in game.

[cold raven]

I'm sure this has been asked before (and if so I'm sorry), but do you think an in-game calculator would ever be possible?

[golden sluice]

proving the haters wrong

[bright cloud]

@cold raven I remember him saying this

[full phoenix]

Did you really play a 5 card straight with 4 fingers

[golden sluice]

Habit - I would rather overshoot than miss by a few chips to play trash hand

[full phoenix]

That's fair yeah. 👌

[bright cloud]

seems like someone implemented a calculator in the form of a joker

https://discordapp.com/channels/1116389027176787968/1216754656617693255 really nice mod

[hushed cove]

Oh that's neat

If that is actually accurate, it shouldn't be too hard to add as a mod that shows it

[bright cloud]

It is accurate from what I saw

But at the same time I was doing inf tests so I didn't really see low numbers

[hushed cove]

I'm wondering if it's calculating things like gold seal incrementing bull a bunch of other smaller interactions

[dark pier]

https://efhiii.github.io/balatro-calculator/?h=FADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADiUAAcSgADmHMOYcw5hzDmHMOYcw5hzDmHMOYcw5hzDmHMMAAQFUCqBVA
Here's the score for this hand (with different numbers of 2s) going past naneinf

(also, the engine rewrite is done, now I just have to integrate it)

[bright cloud]

Hype

[junior cairn]

Nice

[brisk stump]

will you update it to calculate seeds?

[junior cairn]

If you integrate this with something like Immolate, then you’ve pretty much just created a clone of the game lol

So I don’t think that will be something that happens (although I wouldn’t mind helping if you are interested lol)

[brisk stump]

nah im good lol

actually is it possible to add immolate to it? @junior cairn

[junior cairn]

Possible but difficult

Since both tools are written in different languages (OpenCL vs JavaScript)

[brisk stump]

ah

well i wouldn't mind seeing it ;)

[junior cairn]

The best way to do that would probably be to port Immolate to WebCL

Since those could both run in a browser

[brisk stump]

im way too dumb to figure this out lol

[dark pier]

What would be the advantage of having them together in one tool that you don't get from using them separately?
They try to accomplish completely different goals; one being finding the best seed meeting certain conditions, the other being finding the best hand given certain cards/jokers.
I don't see how you could easily combine the two in a meaningful way besides just having the convenience of things being in one place.

[junior cairn]

At that point just recreate the entire game

[bright cloud]

soon we'll just make a tas to play the game for you

[junior cairn]

I was actually thinking of making a TAS mod at some point

[dark pier]

I've made game-playing AI before

https://efhiii.github.io/imgs/28pps-bot-POV_30s.mp4

[bright cloud]

Why play the game when you can get the game to play for you

[dark pier]

In Soviet Russia, the game plays you

[bright cloud]

btw, will the new engine show you the number if you get past inf? cuz i imagine if you have like a new library that can go past inf it wouldnt be hard to just show the number also on the site

[junior cairn]

@bright cloud appears so

[dark pier]

Yes, it'll go up to 1e(9e15)

aka 1 x 10 ^ 9,007,199,254,740,992

(The limiting factor there is that the exponent is stored as a JS number (aka 64 bit float) and 64 bit floats can only store up to 52 bit integers before losing precision. It won't completely break at 1e(9e15) but it won't reliably be correct)

[hushed cove]

The calculator doesn't seem to count DNA rn

it can matter if the cloned card is then has an in hand effect

[dark pier]

Yeah, I've been made aware of that.
I consider it low priority because I consider using DNA to be a throw-away hand.

[hushed cove]

fair

it does impact a few things you don't actually check

[dark pier]

high priority is getting the new engine integrated, which is proving anoying.

[hushed cove]

like cloud 9

[static aurora]

DNA matters with cards in hand shenanigans mostly I think

[golden sluice]

Steel, yeah

[dark pier]

Here's a demonstration of the performance improvements:

[bright cloud]

big hype

[hushed cove]

Huge

[dark pier]

The new engine is live.
I expect there to be lots of bugs

[bright cloud]

well i can easily tell the site is using my cpu by just the sound the fans make when i try to add more jokers lmao

[dark pier]

Yeah, before it would be nice and only use a small amount of resources but be slow, now it's a greedy little guy taking all it can get.

[bright cloud]

little is an understatement, i have a 7800x3d and 32gb of ram and its making chrome crash with like 6-8gb ram usage

its crashing when its getting near 100% cpu usage

[dark pier]

How many cards/jokers are you trying to optimize?

[bright cloud]

im doing a stress test but im having like 14 jokers

i know its alot but like thats what im interested in

https://efhiii.github.io/balatro-calculator/?h=cAASGAO4YA7hGDAHcMAdwwB3DuGAKAAptU2qbVNqm1WU this is what im trying to use optimize jokers on

[dark pier]

Well yeah, the amount of work there is follows a factorial curve based on the number of jokers so 14 jokers would be 87,178,291,200 permutations.

[bright cloud]

so what needs to be done is the permutation reduction thing you said?

[dark pier]

Yeah. I haven't implemented the pruning I talked about earlier.

[bright cloud]

yeah pruning that was it

[dark pier]

I wanted to get this out since it's already significantly faster.

[bright cloud]

also is it just me or is the breakdown looking weird in the site?

[bright cloud]

nah dont get me wrong im already thankful for this

[dark pier]

It's just you
The breakdown is completely borked right now since I had to rewrite that too, though I don't know what breakdown you're seeing.

[bright cloud]

alot of them are broken but this is the best example

it has jumps of like 26 it seems

it jumps from 0 to 26 to 52

yeah 26 each time

[dark pier]

Yeah, I forgot to normalize the exponent before rendering (The number is correct, it's just not normalized to between 0.1 e x - 1 e x)
I also need to add styling so that polychrome is transparent.

[bright cloud]

also idk if itll be easy to add or not, but the joker arrows you have seem so unintuitive, would it be hard or not to maybe do like joker dragging or something?

[dark pier]

It's something I have as a thing to think about, but it's just very low priority compared to things like making it not give incorrect numbers.

[bright cloud]

aight if its on the list its fine

[dark pier]

I've added most of the things I remember being things people want as issues on GitHub.
Feel free to add any more that you think are important.
https://github.com/EFHIII/balatro-calculator/issues

[dark pier]

should be fixed now

[dark pier]

Do you find yourself being too good?
Do you wish you could get another hand played for Ride The Bus?
Well have I got the toggle switch for you!
Introducing: Minimize Score!

[dark pier]

DNA should be fixed

[hushed cove]

are you sure about that?

oh wait I think it does but the breakdown doesn't seem to do it

https://efhiii.github.io/balatro-calculator/?h=wACCGAFpQABAAEMU

[dark pier]

I thought I had tested that and it was good, but I guess not. Should be fixed now.

Check this out

(EV calculations)
https://efhiii.github.io/balatro-calculator/?h=YAAQAgBAYAEA1AQAKAAAAIDgDAgYgJAQ

[deft palm]

this is a great discovery, I was always with my calculator on the side, specially when avoiding too many points with ride the bus or vagabond

[random grove]

I love Bloodstone (sometimes)

[lunar topaz]

is that median vs mean

[dark pier]

yes

(Long-term is mean, short-term is median)

[lunar topaz]

that could be improved by replacing two bloodstones with a retriggerer

[dark pier]

The point wasn't to be a perfect build; just to demo the new EV calculations

much better

[hushed cove]

oh does this use the new properites in the url now?

the one's for card sharp and uhhhh

that one card

[dark pier]

I haven't really thought about it, but I think so?
There's no UI for those properties though, so if you make a URL that sets them, they're stuck that way, lol

[hushed cove]

does it actually use them?

I guess I could test it

[dark pier]

I don't know

[dark pier]

Could you link to where I mentioned these properties because I don't even know what properties you're referring to. I haven't looked at the URL code in a bit.

[hushed cove]

#1212154374206259200 message

[dark pier]

I didn't have the parsing of it written, but I just added it, so it should work now (as in in about 2 minutes)

Actually, Card Sharp still uses the Boolean even though everything is in place for him not to.
I probably won't change that until I have the UI for him setup.

[hushed cove]

fair enough

ui is hard, all users should just write data into a custom binary format by hand

[dark pier]

Obelisk should work though

[hushed cove]

oh neat

I swear there was another

[dark pier]

Right now, Obelisk always is X1 because the "most played hand" is always a tie with every hand.

[hushed cove]

oh yeah supernova

[dark pier]

Supernova

Yeah, it should also work

[hushed cove]

cool

I guess this is the part where I see if my code works

[dark pier]

Supernova currently (for normal users) always gives +1 Mult because every hand type has been played 0 times by default.

[hushed cove]

tbh just add the ui in the hands section like how the game already has it

[dark pier]

I want to, but I also want to maintain a semblance of mobile support

And there's no room

[hushed cove]

scroll bar go brrr

[dark pier]

On mobile, you already can't read the score at level 500

[hushed cove]

supernova seems to work

card sharp and obelisc don't seem to

also supernova's description doesn't show properly

https://efhiii.github.io/balatro-calculator/?h=wAAkGsDlIABkAEsBmBMDoBQEgZBiA4FwHgbAyAECICQPASA4

This is what the above's game state looks like

pair was already played this blind

theres also a non-zero chance I messed up something somewhere

I guess this is the string I generated https://efhiii.github.io/balatro-calculator/?h=wAAkKAAaxwAA5SAAkABkAAzAmB0AoCQMgxAcC4DwNgZACBEBIHgLASA4AAANVwABVSgAHQ

in case me messing with something made the calculator mess something up

Yeah looks like the played this round is properly parsed

[dark pier]

I know

[hushed cove]

also my adding cards ui is still broken unless i zoom

[dark pier]

Here's the data it got

There's no reason that the times played (the 3rd number in the arrays) should be less than the times played this round

[hushed cove]

wait did I switch the played this round or is that just how its parsed?

I thought played this round was a bool

[dark pier]

It is

[hushed cove]

ok so I must've made my bool backwards

pair is the only one played this round

[dark pier]

It looks like here, I inverted the "played this round" for some reason

[hushed cove]

oh yeah I just got to that bit

let me fix that

That looks better

ok fixed build pushed

https://efhiii.github.io/balatro-calculator/?h=wAAkGsDlIABkAAzAmB0AoCQMgxAcC4DwNgZACBEBIHgLASA4

new url

[dark pier]

I'm getting all 0s from that URL

[hushed cove]

this URL? https://efhiii.github.io/balatro-calculator/?h=wAAkKAAaxwAA5SAAkABkAAzAmB0AoCQMgxAcC4DwNgZACBEBIHgLASA4AAAIAwAAAGgADQ

I think the calculator might've deleted those parts

[dark pier]

no, this one

[hushed cove]

I mean try that one

[hushed cove]

yeah it looks like the calcuaotr does not save the changes of those values rn

oh obolisc works now

wait

no it just seems to always reset now?

oh wait

nvm this is the wrong thing

:3head:

obolisc seems to not reset on the one with proper values, but does on the one with all 0 values

[dark pier]

Yeah, since it wasn't used, I hard-coded it to never put them in the URL.
I just changed it so that it should now (again, it may take a minute to update).

[dark pier]

Yeah, if everything's zero, you're always going to be playing your most played hand so it always resets.

[hushed cove]

yeah but if everythng is not, then I paly my most played hand, it doesn't reset

[dark pier]

In this for loop, I forgot to put the .length in this.hands.length

It should be fixed now

[hushed cove]

https://github.com/EFHIII/balatro-calculator/issues/21

[dark pier]

I have no idea what's going on there.

[hushed cove]

truely a css moment

I just thought of a really niche interaction

actually I don't think it matters

yeah

adding width: 100%; here https://github.com/EFHIII/balatro-calculator/blob/main/index.html#L99 seems to fix it and not break it on chrome, but idk why this is happening

[dark pier]

Either I'm look at the wrong line or CSS is being wonky.

[hushed cove]

Does the URL support gold cards?

doesn't look like it cause if I put one in and refresh it loses it's gold status

vampire and midas mask don't seem to behave quite right

iif the mask is after the vampire, it makes them gold after the vampire strips them, so the vamprie doesn't get xmult and the cards stay gold

but your's just has the vampire get xmult no matter which side

[dark pier]

Vampire and Midas have their own phase? arg...

That means that, hypothetically, if you had multiple vampires and midases, you could alternate them:

And the successive vampires would all get additional X_mult

[hushed cove]

yeah

it also means if you have a bunch of gold cards already, you can jput vampire first and eat them, then re gold them

actually

thats odd

only one ate it

So apparently the vampire only eats once

Sounds like a bug to me

[hollow oak]

https://efhiii.github.io/balatro-calculator/?h=YAB-S1wVyiQCCEAO4Gi7ACQAKGo4BwAwGgzAMAoBgERAAyABIADYAEgALAARAAkABYA
Anyone know why in this link my vampire adds zero (says x12) and my brainstorm adds 0.

[dark pier]

the vampire was because I wasn't adding the existing value of vampire.
The brainstorm was because brainstorm (and blueprint) were just behaving like the other joker, not using its existing value.
Both should be fixed now?

[hollow oak]

on the website here?
https://efhiii.github.io/balatro-calculator/?h=UAB-VrgVyiQCCHcrXANF2ABorXANF2ABouwANF2ABouwAMAChqDAMswgRMwHANAYCpZIAQIiABkACQAGwAJAAWAAiABIACw

@dark pier

oh yes thanks!

i hit refresh

love your work on this btw, so helpful!! 🙂

[dark pier]

I do wonder if this is intended.
Let me know if you make a bug report about it somewhere.

Even just getting LocalThunk's response from seeing that would be interesting.

[hollow oak]

hm interesting, optimize hand is pretty broken with steel cards
optimize jokers puts my vampire first and copies it, and it seems to be huge

https://efhiii.github.io/balatro-calculator/?h=oAA0XYAQQfla4FcokA7gYAFDAYQIgIDgBqGXmA0BgKlkgREADIAEgANgASAAsABEACQAFg
you can optimize hand on this if it helps

[hushed cove]

Looks like there is a bug report, for something slightly different, but someone brought this case up #1215730247434240040 message

not thunk response though

[hollow oak]

I did the thing and pressed play and got a billion points 🙂
The calculator was slightly lower than actual, I will post screenshots in case it helps.
before pressing go - 5 cards selected.

my points gained off that press, and the calculator (calculator didn't have +3 gold from playing Ace with golden coin on it, but that doesn't make up the difference. I The numbers seem to work out if this is the bug: One of my vampires gained the +5 ranks from 5 cards played properly in the calculator and the other one did not (Brainstorm)

[dark pier]

I'm having trouble replicating your calculator state, do you have the link?

[hollow oak]

https://efhiii.github.io/balatro-calculator/?h=IAA0XYAPyNuCCHcNAApkQLIFoDYSQNQyzQGAFAaBEQAMgASAA2ABIACwAEQAJAAW

i had some cards that were double buffed so that is hard to see visually

[warped walrus]

your link that you sent is missing a fortune teller

[hollow oak]

oops i will fix

[warped walrus]

i think i see the issue

vampire isn't sucking the enhancements? no indicator

the enhancements are not scoring as if they weren't there tho

[hollow oak]

https://efhiii.github.io/balatro-calculator/?h=oAA0XYAPytcCCFcokA7hoAFMiBZAtAbCSBqGWaAwAoDQIiABkACQAGwAJAAWAAiABIACwA

i think one of the 2 vampires was getting the buffs in the calc but not both of them

[warped walrus]

oh yea

brainstorm shows x12 but vampire shows x13

[hollow oak]

yep that is the bug

[hollow oak]

I learned a lot about how the math works from this website, so handy

[dark pier]

Yeah, that seems to make up the difference

[hollow oak]

yep the only other differnce was the 3 gold I gained on the minotaur

[warped walrus]

why is mime retriggering those cards

[warped walrus]

*bull

wait im dumb

imma get some sleep

getting some abilities confused

[golden sluice]

It's not

[dark pier]

It retriggers all card held in hand abilities and those are cards held in hand.

[golden sluice]

You can only remove enhancement once

After it's removed, another copy of vampire cannot remove it

[hollow oak]

brainstorm copies the vampire....

[golden sluice]

Brainstorm can only copy the Xmult effect

[hollow oak]

right

why are you named stupid?

[golden sluice]

[hollow oak]

the point is vampire grows to 13 when it eats and then we copy the 13

[dark pier]

That was nasty to try and fix, but it should be fixed.
I'm slightly concerned that there are other minor problems related to Blueprint/Brainstorm though.

[golden sluice]

like BP on runner and you play straight?

actually it does not even scale from playing straight

[dark pier]

Yeah, I had a typo in the part of the code that says "this hand has a straight/flush" (it's fixed now)

[dark pier]

Ignoring the typo, that's not an issue because that just checks a variable that's set for if the hand has a straight, and if true adds 10.

[hushed cove]

Right but then I add another enhancement, which you would think the other vampire can remove

[golden sluice]

yea and that's what happens, is it not?

[hushed cove]

No

[golden sluice]

so if you have midas -> vampire -> midas -> vampire
first midas +1.0X to first vampire
second midas +1.0X to second vampire. So they scale separately

that's the intended behaviour

[hushed cove]

Yeah but in the video, only the first vampire removes the gold and scales, and the others do nothing

[golden sluice]

oh, my bad, I didn't see the video, thought it was unrelated probably

[hollow oak]

is the intent to put the calculator into the game to automatically do this stuff eventually? or just on the website?

[dark pier]

Just the website. The main intended use-case is for theory-crafting strategies.
Putting a version of it in the game would be a different project.

[hollow oak]

it has been really useful for telling my friends what i was doing, a screen shot doesn't really help people learn since you can't mouse over anything and read it and see the math in action etc.
I also learned a lot of the rules. I originally didn't realize that xmult was not all counted up like mult is, but indeed multiplied together like real math.

[dark pier]

Yeah, it's definitely become a good tool for understanding how the game works, especially with the breakdown

[dark pier]

Here's my idea for how to have the UI fit

[dark pier]

Okay, it's live.
Card Sharp, Obelisk, and Supernova should all work now.

[dark pier]

These buttons are hopefully helpful

The Play hand button should increment the values of relevant jokers (that don't involve tracking money).

[golden sluice]

does idol not support rank change?

oh I see

they're going in order

[dark pier]

Yeah. That one's admittadly a little clunky

[dark pier]

Ever wonder what would happen if you had every joker?

Baseball Card is a significant contributor to the score.

[bright cloud]

i wonder how long will it take to optimize jokers xd

[dark pier]

150! calculations, aka longer than the heat death of the universe.

optimizing around 81 jokers is where it gets to taking longer than the heat death of the universe.

[bright cloud]

When there are videos on how 52! Is big and here we got 150! Lmao

[lunar topaz]

i feel like you could just put all the +mult before all the xmult, put the uncommon +mults after the other +mults, sort the uncommon +mults by their +mult value, put blueprint and brainstorm both copying baseball card. this is easy

[dark pier]

There are a few more optimizations to make than that; I'm working on it rn

[hushed cove]

I was thinking how theres quite a few jokers that don't affect anything position wise, (things like oops! all 6's, egg, gift card, golden joker, burnt joker, smeared, etc.) and most of these don't matter if blueprinted, so you could just always put them on the left, (so they can still be swashbuklered)

but I worry if theres more cases I can't think of

[dark pier]

https://efhiii.github.io/balatro-calculator/?h=qQB2AAAgZgbgjAEAYAIBQBgHCAEQIoQwjhAiJCQEQIISRiDCGUMEZIIAUAkFAK4NAZA2BpHAOgCwHgoBSC0FYLwUQuDcGkN4ag2BsgwBsLkHAvg4B4DoHwOQ-B4j2A4BIUI-gcjAHcKkVAtAAj-EIAkRggggBMCEE4ZAxhnAsBECoEwLgQgpFIKwUQqhTBgDEGoMwYRyDsHEOocw4R0gQAwBQDQEQMhYCoFkaA2hrDRBoDwGQOgfAxFgLgXgshdC2FiHoPwcQ8jwHoPweQ-RAAE

Oh wait, that's 149 jokers, what did I miss?

[hushed cove]

wee?

[warped walrus]

i see wee tho

[hushed cove]

I'm probbaly blind then

[warped walrus]

notice the gap

[hushed cove]

ah yeah

[warped walrus]

also the four sinful joker art is outdated

[dark pier]

Hack was missing

https://efhiii.github.io/balatro-calculator/?h=aQB2AAAgZgbgjAEAYAIBQBgHCAEQIoQwjhAiJCQEQIISRiDCGUMEZIIAUAkFAK4NAZA2BpHAOgCwHgoBSC0FYLwUQuDcGkN4ag2BsgwBsLkHAvg4B4DoHwOQ-B4j2A4BIUI-gcjAHcKkVAtAAj-EIAkRggggBMCEE4ZAxhnAsBECoEwLgQgpFIKwUQqhTBgDEGoMwYRyDsHEOocw4R0gQAwBQDQEQMhYCoFkaA2hrDRBoDwGQOgfAxFgLgXgshdC2FiHoPwcQ8jwHoPweQ-QlCAAIA

[dark pier]

Thanks; I had grabbed the sprite sheet from the demo on accident when starting the project and forgot to change that.

Should be fixed now.

[lunar topaz]

long term expected value doesn't consider that in the long term scaling jokers will scale

[dark pier]

long term expected value is the mean

It's not supposed to take that into account; I just figured most people aren't going to remember the difference between mean and median.

The long term EV of anything with scaling jokers taking into account scaling would be Infinity.

[dark pier]

I finally got a solid run, but forgot to reroll the boss.

My highest hand:
https://efhiii.github.io/balatro-calculator/?h=YADAUAAYCgADAUAAYCgAA2UAALSgADwAJ6v1fq_V-r9X6v1fq_V-r9X6v1fqwAGI

[dense elk]

I believe there is an issue with raised fist - iirc, debuffed cards don't work with it, so the value is 0 when the lowest card held in hand is debuffed

Also, when adding cards, some of the cards held in hand don't show up at all (but will reappear once they would be in the played hand).

[dense elk]

to mee, it looks like it's using the art for the scholar right now, not sure why

[dark pier]

Oh yeah, I changed the sprite sheet but I was using a modified sprite sheet because the sprite sheet in the game (for some reason that confounds me) has a duplicate of Scholar's sprite in Wee joker's spot.

should be fixed now

[golden sluice]

https://efhiii.github.io/balatro-calculator/?h=gABBCAAQDgDw
does not show debuffed 10 in hand for some reason

https://efhiii.github.io/balatro-calculator/?h=wABBDIDYEABAOAPCXgbw
Now debuffed king is also not being shown

plus all debuffed cards still trigger held in hand jokers

[dark pier]

This is known, see the github issues.

[golden sluice]

Is it possible that for jokers like square, castle etc. you specify value as the final mult/chips value? manually removing the base and dividing chips on the joker is super annoying

and first time I used it, got super confused, since I assumed it already takes the final chips value ^ so the results were incorrect