https://fallenlondon.wiki/wiki/Blog:Naoh/history?feed=atom&Blog:Naoh - Revision history2024-03-29T13:31:09ZRevision history for this page on the wikiMediaWiki 1.35.6https://fallenlondon.wiki/w/index.php?title=Blog:Naoh&diff=1086215&oldid=prevNaoh: blog entry2021-04-04T19:18:41Z<p>blog entry</p>
<p><b>New page</b></p><div>=== The Emissary, part 2: Hunting and Searching in Parabola ===<br />
<br />
Some aspects of Fallen London are very linear and easy to model using the linear programming approach used in my grind optimisation software, [https://github.com/klenze/emissary the emissary]. Others are more difficult. The tasks around [[Lost in Strange Lands]], e.g. hunting animals and searching places in Parabola, possibly one of the most Byzantine grinds to analyse. So to analyse that, I did what any normal person would do and wrote a small framework for a state-based simulation of that part of Fallen London.<br />
<br />
For Parabola, I have implemented the relevant actions [https://github.com/klenze/emissary/blob/main/generators/hunting_generator.py#L40 here].<br />
I tried to make the framework (sim.py) as complex as needed to keep these definitions as simple as reasonably possible.<br />
<br />
The standard example considered here will be the hunt of the Pinewood Shark with ferocity 18, generally the most difficult repeatable hunt. For simplicity, we consider a player character inspired by my primary, e.g. [[Dangerous]] 261, [[Glasswork]] 8, [[Monstrous Anatomy]] 7. The simulation optimises for action costs only, so the costs for getting rid of menaces, buying or grinding items or second chances must also be stated in terms of action costs. See [https://github.com/klenze/emissary/blob/main/char.py char.py] for details.<br />
<br />
Without any complications, for the given stats, the best way to hunt is quite straightforward:<br />
- [[Push onward]] (until you succeed)<br />
- [[Trick the Parabolan Quarry into chasing you]] (with Nightmares 5)<br />
- Kill it<br />
<br />
The probability of success for the first and the last option is rather low, so you will probably fail a lot and get lots of wounds. My simulation suggests that you will spend 13 actions per shark hunted this way '''(0)'''. As mentioned, this includes taking care of wounds you will acquire.<br />
<br />
What about second chances? They can, obviously, help a lot. If you do not assign an action cost value to them at all (e.g. you encounter "Hard earned lessons has not increased because it is larger than 20" twice a day), spending them liberally will reduce the cost to hunt a shark to 7.1 actions '''(1)'''. A more sensible use would be to use them only for the kill action. This will require around 2 hard earned lessons per shark and decrease the action cost to 8.5 actions '''(2)'''. If you value your second chances at 1 action (reasonable), this is also the optimal path for then 10.3 actions -- the same as before, plus the valuation of almost 2 second chances at one action each. '''(3)'''<br />
<br />
If you have the Ursine Honey Connoisseur, you probably have Glasswork 10 rather than 8. This means that you can save a whole 0.17 actions on the hunt by taking "Apply your knowledge" until the success probability is less than 0.7 '''(4)'''. Characters with Glasswork 11 (Silverers, some Ambitions) even get a further discount for effectively 10 actions per shark. <br />
<br />
Okay, if the Silverers Glasswork bonus is not all that helpful, what about the Monster-Hunter? The key action here seems to be "Wound Quarry", which reduces its ferocity by 1+[[Monstrous Anatomy]]. We will assume our monster hunter has a MA value of 10. (Furthermore, I assume that the Wound Quarry action does not change the storylet on success or failure, e.g. stays "Pursuing the Parabolan Quarry". The wiki suggests that, and I am in no position to test it.) The outcome for the Monster Hunter is 9.4 actions per shark (minus 0.2 with Glasswork 10) '''(6)'''. <br />
<br />
Ambushing a quarry in its lair does not seem to be a good option for the given quarries. We can find out how ferocious a quarry would have to be for this to become the best option. (We will take the stats from option '''(3)''', above).<br />
A ferocity 30 shark would take almost 16 actions to catch, but would follow just the same strategy. <br />
A ferocity 40 shark would take about 20 actions, still no change in strategy.<br />
With ferocity 50, it will take some 25 actions. Eventually, we will spent hard-earned lessons also on "Push onward", but the tracking to home option still fails to convince.<br />
<br />
If we manually disallow the Trick quarry into following you option, things change. the action cost for a hunt of the real (as far as these things can be considered real) shark (ferocity 18) becomes 16.5 '''(7)'''. The optimal path here is now: Apply, Gazette, Push, Lair, Continue, Strike, 4x Ambush, Kill. Lets unpack this a bit.<br />
If we track the quarry to its home, we have to succeed twice at tracking: "Push onward" and "Continue to its home". This makes scouting a lot more useful, to the point that even the Glass Gazette is worth its (effectively) two actions (subject to [[Explore the dreams of the Church in the Wild|some requirements]]). Another key insight is that "Continue towards (Quarry Home)" resets scouting. This is the reason why we ambush that poor shark to death after we spent all of the energy to track it to its home. Otherwise, we would just take our chances (0.56 with a hard earned lesson).<br />
<br />
No Strategy Requirements Actions per shark<br />
0 Trick Nightmares 5 13.0<br />
1 Trick+2nd NM5, Lots of hard earned lessons 7.1<br />
2 Trick+2nd NM5, some lessons 8.5<br />
3 Trick+2nd NM5, 2nd chance = 1 action 10.3<br />
4 Apply+Trick+2nd GW10, NM5, see above 10.1<br />
5 Apply+Trick+2nd GW10, NM5, see above 10.0<br />
6 Wound+Trick Monster-Hunter, NM5, MA10 9.4<br />
7 Apply+Gazette+Lair+4x Ambush high player pain tolerance 16.5<br />
<br />
With regard to other preys, it turns out that spending a Goldfish / Nothing (or one action worth of honey) to catch the albatross / bear is indeed worth it. <br />
<br />
Action costs to capture/recapture/kill various quarries (and search their lairs)<br />
Quarry Hunt Search Lair<br />
warbler 5.2 5.6<br />
albatross 5.2 (5.8)<br />
storm-bird 8.9 7.0<br />
terror bird 6.0 8.5<br />
auroch 5.7 (8.0)<br />
bear 7.6 (7.5)<br />
shark 10.3 (8.1)<br />
<br />
While hunting in Parabola is mostly interesting to get the [[Dangerous Gains]], the simpler [[Search Parabola for a Place in the Neath]] has a wider range of applications. The top five (mostly cardless) Echo grinds discussed in the [[Blog:Naoh/Introducing_The_Emissary:_a_python_script_for_finding_efficient_item_grinds|last part]] involve [[Find the weakness in an opponent's defences]]. [https://github.com/klenze/emissary/blob/main/sample_outputs/gen_hunting_generator.out.txt#L1316 For Glasswork 8], the best action is to [[Apply your own knowledge]] once, then [[Sneak forward]] until you succeed, average 3.76 actions. People with GW<=7 should start with [[Sneak forward]] for 3.86 actions instead. <br />
For Glasswork 10 (e.g. my char, once he has recovered from the storm bird), we are at 3.40 actions, GW 11 will take 3.25 actions (so a Profession change to Silverer might be worth it in the long run). With GW 12 (or more), it will only take some 3.13 actions. This will in turn raise yields from [[Ambush a courier]] to 5.70 EPA (instead of 4.92 EPA with GW 8). <br />
<br />
Thats all for tonight.</div>Naoh