Uminchu - Auto Fishing Bot

[scowlyas]

just wanted to pop in and say thanks again for the help earlier, and to also say that i'd be excited beyond words to see the Gold Saucer automation (namely chocobo racing) mentioned in the OP come to this bot! just in case u wanted a reason to try it ^^;

[Wilds76]

I so excited for the Gold Sausage minigames ... i love sausages !

[Lyllandor]

It's still automatically doing stuff I don't want it to automatically do.

I have this in my ini file:

[Settings]
OptMooch=0
OptChum=0
OptPatience=1
OptCollect=321
OptTimeout=0
OptStartTime=0
OptEndTime=0
OptUpdatesOn=1
OptAutoDesynth=1
OptAutoFish=1
OptAutoGather=1
OptAutoAcceptQuests=0
OptAutoRetainerReassign=0
OptAutoSkipCutscenes=0
OptAutoSkipDialog=0
OptAutoTurnin=0
VarNotified=16186391
OptDebug=0

Went to go start a quest that I was excited to read the lore for and before I can blink it's gone. Had to quickly close the bot to stop this behavior.

[jelio]

I'm very sorry to hear of your negative experience. Knowing which settings combinations you are using should be helpful in debugging. Can you tell me a bit more about the circumstances?

What exactly was skipped - a line of dialog, many lines of dialog repeatedly, a quest accept window? Were you in a cutscene? Does this always happen or do you know the circumstances where it happens?

Because the desynth and gather features currently require a copy of unmacroed sprint somewhere to work properly, can you tell me if you have that set up now? Also, had you just been doing either?

I have been making small changes every day to these routines and making them more self contained. They should behave a bit better next update but if I can reproduce your situation I can squash it for sure.

Were you able to rewatch that part of a quest or cutscene later?

[Next Post]

New release. I learned a lot about classes, recursion, caching and optimization by making this cactpot solver. It was so much fun. The important algorithm to recommend the next tile was developed by years ago. I actually made a good heuristic solver originally, but he made the perfect solver by brute forcing all possible board states. Please check it out and let me know if it's buggy. It's a bit hard to test the UI stuff with only 3 chances per day per character, but the solver itself seems to have no errors.

I re-added mod-key pausing for all users again. If it's doing something weird, you can press this to pause the bot temporarily.

• 2017-04-16 Update
  • Added Mini Cactpot automation! Just start playing and the bot will automatically select the optimal tiles and rows to reveal. It will play all remaining games available.
  • You can now configure everything within the UI. See the updated More button.
  • You can now set the minimum percent chance of success to keep desynthing. Default 40 (percent).
  • You can now automatically desynth items that require you to confirm by checkbox. Default off - use caution.
  • Many changes were made to improve the behavior of the automation routines. They will give up automation due to user interference more quickly.
  • Retainer Reassign now also moves the cursor to the left on the second (Assign) window.
  • Skip Cutscenes now captures virtually all cutscenes. It will still mash cancel to some extent on those cutscenes which can't be skipped, but should give up. This feature needs special logic to ignore especially problematic cutscenes such as the Aesthetician and the Squadron recruits. Please let me know if you encounter especially problematic scenes.
  • Holding down a modifier with the game in focus will delay keypresses on any system again. This is to allow the user to quickly pause the bot if necessary.

[Lyllandor]

I had just been fishing. I dinged 60 and went to the quest giver. Started reading the quest description window and it accepted the quest and started paging through the dialog of the quest as fast as it could go before I closed the bot. I went back and read the lines I missed in the event log, but it was still a hindrance, especially considering that I added those lines to the ini file and made sure they were set to 0.

I added sprint without a macro to hotbar 6, and auto gather worked later on. I'm just scared to start any quests with the bot running since I can't figure out how to turn those features off. Maybe now that they can be configured in the UI that'll fix it, dunno. Will probably be a few days before I play FF14 again though, busy with other things. Just happened to get on today. Keep up the good work!

~

Think I figured it out. I could've swore that I scrolled up and down the config file before adding the lines myself, but apparently there's now settings at the end in a section called 'automation' that has them all turned to =1. Sigh.

[jelio]

Sorry about that. That's right, it needs to be in the automation section. The new version should also clean up the ini file and resort if you have some old/strange settings in there too. But you shouldn't need to fiddle with that anymore. I'm glad it's resolved.

By the way, the options window must be closed for changes to take effect.

[agentken007]

Pretty sure that automation section has always been there... thats how I changed my settings, never had any problems. Thanks jelio for the update, works great!

[Wilds76]

Thank you @ !

[Lyllandor]

Just to confirm, because there's multiple versions of the 'perfect solver' algorithm for cactpot out there, that you are using the one from this one?



This is the 'latest version' fork, as far as I can find.

[jelio]

Essentially yes. It should produce identical results to that as we are both using the same algorithm to pick the next tile. This is done by going through what the board could look like if we choose each tile recursively until board state 4, then picking the best row. The average value of those hundreds of thousands of final choices is then compared.

But the code is not quite the same in all places. For example, that version brute forces all possible future states to calculate the expected winning for any given *row* which is hundreds of thousands of calculations in the early game. (In practice he only calculates this at step 4 which is fast enough in javascript but a bit too slow for me.) But you only need at most 8C3=56 states to calculate the expected winnings for a completely blank row at step one, which can even be reused for any fully blank row.

[Lyllandor]

That went a little over my head lol. I just know when I went looking around I found 3 of those githubs that all look nearly identical but the one that I linked was the 'latest' that incorporated all the others.

Essentially it calculates not just the highest average row value, but which 'next pick' is most likely to reveal the highest average row from all possibilities.

The first 81 variations (the revealed number basically) are hard coded since the calculations have already been done and don't need to be recalculated, and the rest are done on the fly.

I don't know what the differences are in the other variations though but as this one was created most recently and seems to be perfect, I would assume any variation would take it away from perfect.

[jelio]

Yes, I am calculating the next tile in the same way. However, I don't waste thousands of cycles figuring out something that can be done in tens of cycles. I also use a pre solved dictionary for the hardest 82 board states (including all zeroes, though this cannot be played in game).

However, the UI navigation I'm not so confident on. It looks like it will sometimes just press confirm through to pick the first diagonal highlighted. Tile picks look perfect. I think I will temporarily disable the confirm button until I am more confident in the UI nav.

[Lyllandor]

If you have some way of the bot indicating what it WOULD pick, without it actually picking it, that'd be cool. I'll compare with the site I'm using and see if I notice any discrepancies. If after a couple weeks there's no discrepancies I'd feel safe enough turning it to automated. ^_^

[jelio]

If you are interested in what I was trying to explain, consider this board state:

X23
45X
***

The way he is calculating the expected winnings for X23 for example is like this

calculate the average winnings for the this row in the following board states:

123   623   723   823   923   123   623   723   823   923
456   451   456   456   456   459   459   459   459   451
789   789   189   719   781   678   178   618   671   678

... and so on for 120 board states.

I solve it in 5 states: 123, 423, 523, 623, 723, 823, 923. We get the same answer.

As far as the tiles, it already seems quite good. I will disable the final confirm for now in case it's really messing up on the rows. Been hitting plenty of jackpots though.

Hmm. I can have it display the solutions using a Tooltip like the ones when you hover over an option.

[Lyllandor]

I think the reason it does 120 board states is to average all 8 rows simultaneously. Whereas you could just average each row's potential states, then compare those averages. Two ways to get to the same point.