v2.3.1

[AlexWalp]

Hotfix

• Fixed an issue that caused the bot to be stuck after MD clear

---

This discussion was created from the release v2.3.1.

[AlexWalp]

Bruh

[gearnam2]

I tested the new version and this is what I think.
for the dev:

1. Tier 3 and higher Vestige are not eligible for gifts. (tier 1-2 worked)

2. The panel where you can choose gifts, it seems to me, does not fully work. As an example: the bot most often does not select floors where there are gifts that you have chosen, and even if it does, when choosing, it will still take something else (I understand most likely what this is connected with, but those gifts that are chosen should be higher than mounting trials because their impact exceeds even the worst trials, although this works in the opposite direction if a person chooses not the strongest gifts on this list.)
   Of course, choose these gifts for the sake of Clash Power, Coin Power or Base Power on later floors can be a fatal mistake for the run. Therefore, I propose an option where these gifts are taken in any case, even if these are the worst trials from the 1st to maybe the 6th or 7th floor, where the choice of trials will not be as critical as further from 8 to 9. And also between choosing a gift for status or keeping your chosen one, he chooses a status type rupture tier 2-3 instead of tier 3 Rusty Commemorative Coin or gift for 30%- buy gifts ?

3. The most difficult thing in terms of solution is reworking the improvement of the 3rd skill of the ID.
   What is the point, now as I understand the bot does this

• Spends all the money on gifts (making keyword rolls!) on certain gifts.
• Tries to make tier 4 gifts as early as possible and improves them.
• And only after all these actions, and provided that the player has 120+ cost left, only then does he buy 3 skills. (no matter what kind of ID it is, and even if it is not used) honestly, I don’t even know approximately how the code could be rewritten so that it would do these actions more profitably.
  therefore, what I write next may not be implemented. The logic is this:
• Add an icon to each sinner due to which the bot will buy 3 skills only from specially selected ones. Again, because of the 10 floors, most people take not 7 sinners, but all 12, and therefore the bot can buy skills for supports. For some, these IDs are most likely not upgraded and are needed to activate passives.
• A new problem appears. how will the bot buy these skills from the selected IDs if there are, say, 2-5 of them.
  while the bot is doing a store roll, it is also trying to buy the required skill (in short, the priority when a store appears/rerolls will be skill.)

4. reroll logic. as I wrote above, the bot often uses keyword rerolls instead of regular ones (by the way, the bot does not use the free roll at all if you took the starlight buff). These actions hit the cost very hard, and to reduce expenses, you need to use regular rolls. Let's say 3-4 maximum. and only when a super store appears, spend a keyword roll because there are more gifts there and will be many times more effective.

this solves the problem of buying 3 skills of a certain sinner, reduces the use of cost and increases the chance of the user receiving selected gifts, but at the same time reduces the chance of status ones.
(in most cases, when the bot does a keyword roll and sees a tier 1 gift, he also buys it by spending cost, when something more valuable might fall out in a normal one, knowing that we have 10 floors and there is no need to rush to buy all the status gifts except tier 3-4. although the bot fuses them as quickly as possible)

For players:
in two runs I used all the heishou with lord hong lu.

• I chose initial buffs 1 and 3.
• Card priority

1. starlight
2. gift
3. cost and gift chance
4. cost
5. resources (by the way, there is a bug when the bot picks up this card and not the ones selected above 4. as I understand it, because of the rarity. resources were tier 4-6? and others were 2-3 lower)

• To increase the chance of passing 10 floors, you must definitely prioritize packs and avoid some. I won’t write my own because the list is too long, but Glut. was the priority at least. I’ll leave someone’s list of what to avoid?
• The selected gifts are rusty coin, gifts to increase cost and discount gifts.
  each run lasted no more than two hours although it still requires control (on the 10th floor the boss had to be finished off manually)
  I forgot to add that I added gifts for slash and poise to the additional affinity

[AlexWalp]

1. I'll check it, but this wasn't an issue in my test runs as far as I can remember.
2. It doesn't see the gifts in floor selection, only gift affinities and whether those gifts are owned or not. If there are multiple floors to select from (no priority is set) then it chooses the one that has more unowned affinity gifts (rupture). I could improve it, but the small icons may plummet the detection rate which is not perfect even with bigger ones.
   And no, I think that mounting trials are MORE important than gifts (unless you have something like piece of relationship, the only exception). This is consistent with my testing as well. Mounting trials are unimportant in Hard MD, but CRUCIAL in Infinite mode.
   Also, yeah, currently the priority on affinity is higher than on keywordless. Perhaps I'll fix it in the next update.
3. It buys skill for all deployed identities. Having all 12 sinners selected does mean that it'll buy those for everyone. I can set it to target only the first 7 if that's a big issue. And then the s3 can be bought along with ego gifts instead.
4. Keyword refresh is very good for Normal MD, and the refresh was designed around that. I agree that keyword refresh is not optimal for getting gifts just to fuse them for tier 4. I can adjust that as well. But limiting keyword refresh to supershops may be too radical.

Regarding keyword refreshes, I would really like to see some math to back up the efficiency, but it's not even clear how more likely the keyword gift will appear with the keyword refresh...

Your runs are consistent with what I got in my testing.
I think that the biggest issue with the bot on f10 is not inefficient cost management, but its limitations in battle.
I'll try to address those limitations in the next update.