2 Commits

Author	SHA1	Message	Date
Crow	dca0be2932	Updates to SSC Strategies (#2138) ... # Pull Request Most of the changes are not functional but are to modify style based on comments received to the TK PR (e.g., eliminate nesting of if statements) and leverage general boss helpers. There is some reordering of returns and other changes to try to consolidate or clean-up the code (such as removing unnecessary parameters). The strategies themselves have only minor changes. - Main tank no longer uses tangential movement for Lurker Spout, unlike other bots. The MT will just call moves directly to a position behind Lurker. This is because I found tangential movement was taking too long for the MT to get in place since it starts right in front of Lurker. - Vashj MT group Shaman will now use Grounding Totem without actually switching to the Grounding Totem strategy. I have now eliminated all strategy swaps, which I dislike because they persist after the encounter, and it's better not to mess with player strategies since presumably people are generally using Windfury or Wrath of Air for the Air Totem. - I made a ton of changes to Vashj core passing as I noticed the existing logic is nonfunctional in several ways. It generally works fine ingame, but the changes should make things much smoother. For example, the storing of the nearest trigger NPC for generators in the existing strategy is useless because it relies on insert_or_assign for an unordered map that will continue to run during the course of the core passing logic, and a similar issue exists with respect to the map to store the last time a bot held a core. The result is that there is slight movement of bots when the core is flying through the air and not held by any bot because the trigger for core passing does not fire during that period. In practice, the time is brief enough that the sequence works OK, but it looks stupid because the bots should not be moving at all. So that should be fixed. There is a known issue re: core passing that would take extreme effort to fix and I am not going to do it because it is a fringe situation. There are a couple of spots where the Tainted Elemental can rarely spawn that can result in a straight-line passing sequence to the nearest generator that is blocked by LoS. Fixing this would be extremely difficult and niche, so what you will need to do if this happens is to just command your bots to destroy the core and try again with the next spawn. --- ## Design Philosophy We prioritize **stability, performance, and predictability** over behavioral realism. Complex player-mimicking logic is intentionally limited due to its negative impact on scalability, maintainability, and long-term robustness. Excessive processing overhead can lead to server hiccups, increased CPU usage, and degraded performance for all participants. Because every action and decision tree is executed **per bot and per trigger**, even small increases in logic complexity can scale poorly and negatively affect both players and world (random) bots. Bots are not expected to behave perfectly, and perfect simulation of human decision-making is not a project goal. Increased behavioral realism often introduces disproportionate cost, reduced predictability, and significantly higher maintenance overhead. Every additional branch of logic increases long-term responsibility. All decision paths must be tested, validated, and maintained continuously as the system evolves. If advanced or AI-intensive behavior is introduced, the **default configuration must remain the lightweight decision model**. More complex behavior should only be available as an **explicit opt-in option**, clearly documented as having a measurable performance cost. Principles: - **Stability before intelligence** A stable system is always preferred over a smarter one. - **Performance is a shared resource** Any increase in bot cost affects all players and all bots. - **Simple logic scales better than smart logic** Predictable behavior under load is more valuable than perfect decisions. - **Complexity must justify itself** If a feature cannot clearly explain its cost, it should not exist. - **Defaults must be cheap** Expensive behavior must always be optional and clearly communicated. - **Bots should look reasonable, not perfect** The goal is believable behavior, not human simulation. Before submitting, confirm that this change aligns with those principles. --- ## Feature Evaluation Please answer the following: - Describe the **minimum logic** required to achieve the intended behavior? - Describe the **cheapest implementation** that produces an acceptable result? - Describe the **runtime cost** when this logic executes across many bots? I have attempted to streamline methods and even remove some. The strategy is admittedly somewhat performance heavy due to the need for function calls such as iterating over inventory items. However, the new version should be less performance intensive than the merged strategy--for example, there were places where all members of the raid would have their inventory checked, but I've now limited the check to only the 5 core handler bots. I've run the instance with pmon on, and there are no methods that stand out as particularly resource intensive when not in a boss encounter, and I view that as the most important thing (though I make effort to reduce performance impact during encounters also). Expensive checks that are unavoidable for the strategy to work such as grid searches are gated behind cheaper checks. --- ## How to Test the Changes - Step-by-step instructions to test the change - Any required setup (e.g. multiple players, bots, specific configuration) - Expected behavior and how to verify it Enter SSC with a raid group and run the instance, including all bosses. Every boss should be killable, and every major mechanic should be addressed by bots. I will work with Dreathean to get the Wiki up soon so that should be a reference for testing strategies. ## Complexity & Impact Does this change add new decision branches? - - [ ] No - - [x] Yes (**explain below**) Only within the context of strategies, which are basically all new decision branches, and there are some tweaks here to what is currently merged. Does this change increase per-bot or per-tick processing? - - [x] No - - [ ] Yes (**describe and justify impact**) Could this logic scale poorly under load? - - [ ] No - - [x] Yes (**explain why**) I'm sure if you have a large server, with multiple raid groups running the instance at the same time, the performance impact could be significant. But I have done my best to limit it, and I think some notable performance impact is unavoidable with the current framework for raid strategies. --- ## Defaults & Configuration Does this change modify default bot behavior? - - [ ] No - - [x] Yes (**explain why**) Only for strategies in the instance. If this introduces more advanced or AI-heavy logic: - - [ ] Lightweight mode remains the default - - [x] More complex behavior is optional and thereby configurable There should be no impact if co +ssc and nc +ssc are not added to bots. Because raid strategies are currently not removed after leaving an instance, players should manually remove them (or reset botAI). This is a general issue that needs to be addressed with the module. --- ## AI Assistance Was AI assistance (e.g. ChatGPT or similar tools) used while working on this change? - - [ ] No - - [x] Yes (**explain below**) If yes, please specify: - AI tool or model used (e.g. ChatGPT, GPT-4, Claude, etc.) - Purpose of usage (e.g. brainstorming, refactoring, documentation, code generation) - Which parts of the change were influenced or generated - Whether the result was manually reviewed and adapted AI assistance is allowed, but all submitted code must be fully understood, reviewed, and owned by the contributor. Any AI-influenced changes must be verified against existing CORE and PB logic. We expect contributors to be honest about what they do and do not understand. GPT-4 because I don't like to use up my premium requests in CoPilot and I generally like it better than GPT-5 =P I use LLMs to draft code snippets but do review everything and have become less-and-less reliant over time. I don't use agent mode, only ask. For this PR, I had it do the updated version of AnyRecentCoreInInventory(), which is more complicated than before and uses indexing for each bot to consider status of only prior bots in the passing chain. Everything else either I wrote or could have written but had the AI help and just edited afterward to save time. --- ## Final Checklist - - [x] Stability is not compromised - - [x] Performance impact is understood, tested, and acceptable - - [x] Added logic complexity is justified and explained - - [x] Documentation updated if needed --- ## Notes for Reviewers Anything that significantly improves realism at the cost of stability or performance should be carefully discussed before merging. --------- Co-authored-by: Keleborn <22352763+Celandriel@users.noreply.github.com> Co-authored-by: bash <hermensb@gmail.com> Co-authored-by: Revision <tkn963@gmail.com> Co-authored-by: kadeshar <kadeshar@gmail.com>	2026-03-13 22:21:37 +01:00
Crow	8585f10f48	Implement Serpentshrine Cavern Strategies (#1888) ... Edited: Below description of methods were brought up to date as of the PR coming off of draft. ### General I've starting leveraging, to the extent possible, an out-of-combat method to erase map keys. This is mostly useful for timers that need to start upon the pull because I dislike having to rely on a check for a boss to be at 100% HP (or 99.9% or whatever) because it can be unreliable sometimes. ### Trash Underbog Colossi: Some Colossi leave behind a lake of toxin when they die that quickly kills any player that is standing in it. The pool is a dynamic-object-generated AoE, and bots will not avoid it on their own (I think because the AoE is out of combat, plus the radius is much larger than the default avoidance radius in the config). The method does not require bots to be in combat, and simply gets bots to run out of the toxin. You will probably still get a couple of idiots who drink in the middle of it, but in my experience, the vast majority of the raid gets out, and healers that escape can easily keep up a couple of fools until they've drank to full. Greyheart Tidecallers: Bots will mark and destroy Water Elemental Totems immediately. ### Hydross the Unstable The strategy uses 2 tanks, with the main tank assigned to the frost phase and the 1st assistant tank assigned to the nature phase. - The main tank will tank the frost phase, and the first assistant tank will tank the nature phase. They each have designated spots and will wait at their spots twiddling their thumbs while Hydross is in the other phase. - Hunters will misdirect to the applicable tank upon the pull and after each phase change. - The phase change process begins 1 second after Hydross reaches 100% Marks. The current tank will begin moving to the next phase tank's spot for the next tank to take over as soon as Hydross transitions. - DPS is ordered to stop after Hydross reaches 100% Marks until 5 seconds after he transitions. - Bots will prioritize the elementals adds after every phase change, unless Hydross is under 10% HP, in which case they should ignore the adds and burn the boss. - Ranged bots should spread during the frost phase to mitigate the impact of Water Tombs. ### The Lurker Below - There is a designated spot for the main tank. - Ranged DPS will fan out over a 120-degree arc that is centered directly across from the tank spot (to try to spread to reduce Geyser damage while also keeping them behind Lurker). - When Spout begins, all bots will run around behind Lurker. The intent is to keep a distance with a radius of 20 or 21 yards and within 45 degrees (either side) of directly behind him. Movement is specifically tangential along an arc so bots don't run in front of Lurker. - Spout's duration is tracked by a timer. The mechanics of the spell itself are rather unique and don't involve a continuous cast or aura to track easily so I settled for the timer. - If you have 3 (or more) tanks, each of the first 3 tanks will be assigned to one of the 3 Coilfang Guardians during the submerge phase. ### Leotheras the Blind The fight is designed for a Warlock tank. You can choose the Warlock tank by giving a Warlock the Assistant flag. If you don't do that, your highest HP Warlock will be picked. Do NOT switch the Warlock tank to a co +tank strategy--the designated Warlock is hardcoded to spam Searing Pain on Demon Leo and otherwise will engage in normal DPS strategies. If you don't have a Warlock at all, the strategy has some methods built in to try to make things work as best as possible with a melee tank. - The Spellbinders get marked with skulls and killed in order. - There is no designated spot or designated tank for the human phase. Your tanks will fight for aggro. Ranged bots will attempt to keep some distance, and when Whirlwind starts, everybody will run away from Leotheras. - During the demon phase, your melee tanks should take a backseat to your Warlock tank, who will receive help in the form of Misdirection. Bots will get the hell away from the Warlock tank so the Warlock tank should be taking every Chaos Blast alone. - During the final phase, your regular tanks will tank Leotheras, and the Warlock tank will tank his Shadow. The melee tanks will attempt to separate Leotheras from his Shadow so bots can focus down Leotheras without getting hit with Chaos Blasts. - Bots will wait 5 seconds to DPS after every transition into human phase, 12 seconds to DPS after every transition into demon phase, and 8 seconds to DPS after the transition into the final phase. There is no waiting on DPS after Whirlwinds, even though it would be ideal. It's not a big deal to live without, and for various reasons, it would have been a pain in the ass to deal with. - Bots will save Bloodlust/Heroism until after Spellbinders are down. - To deal with the Inner Demons, I disabled DPS assist for bots who are targeted and force them to focus only on their Inner Demons. This is sufficient in my experience for all DPS bots and Protection Warriors and Paladins to kill their Inner Demons, even at 50% damage. Feral Tank Druids and Healers still need help, so the strategy hardcodes their actions while fighting Inner Demons. For example, Resto Druids are coded to shift out of Tree Form, cast Barkskin on themselves, and just spam Wrath until the Inner Demon is dead. There are no bot strategy changes used for this method. ### Fathom-Lord Karathress You will need 4 tanks. Your main tank will tank Karathress, and an assistant tank will tank each Fathom Guard. If you have fewer than 4 tanks, then the priority order for tank assignment will be Karathress, Caribdis, Sharkkis, and then Tidalvess. - Roughly, the tank spots are (1) for Karathress, near where he starts but closer to the ledge for LoS reasons, (2) for Sharkkis, North from his starting location on the other side of the ramp, (3) for Tidalvess, Northwest from his starting location near the pillar, and (4) for Caribdis, far to the West of her starting position, near the corner. - Note that the tanks will probably clip through the terrain a bit when going to their positions. This is due to me implementing a forced MoveTo to the tank position coordinates. There is something weird about the maps in Karathress's room, and the tanks will take some really screwed up paths without making them go directly to the exact coordinates. So this looks stupid but is necessary. - One healer will be assigned to heal the Caribdis tank. Because AC Playerbots does not yet have a focus heal strategy, this just means that such healer has a designated location near the Caribdis tank's location. This healer can be selected with the Assistant flag. - Hunters will misdirect the Fathom Guards onto their applicable tanks. If you don't have three Hunters, the priority is Caribdis, Tidalvess, then Sharkkis. - DPS will wait 12 seconds to begin attacking. After that, they will prioritize targets as follows: - (1): Melee will always prioritize Spitfire Totems as soon as they spawn. This will continue through the duration of the fight. - (2): All bots will kill Tidalvess first. - (3): Melee bots will move to Sharkkis, and ranged bots will move to Caribdis. I understand this is not the standard kill order for players, which would have the entire raid kill Sharkkis next. The reasons I have done this differently are because melee DPS is much stronger with 3.3.5 talents vs. in retail TBC, and because bots get really thrown off by Cyclones and therefore they struggle to kill Caribdis quickly. You do not want Karathress below 75% HP before all Fathom-Guards are dead or he gets a huge damage buff. - (4) If Caribdis dies first, ranged bots will help with Sharkkis. - (5) Everybody kills Sharkkis's pet. - (6) Everybody kills Karathress. ### Morogrim Tidewalker - The main tank will pull the boss to the Northeast pillar, with the tank's back against the pillar. - A hunter will misdirect the boss onto the main tank upon the pull. - When the boss gets to 26% HP, the main tank will begin moving the boss to the Northeast corner of the room in preparation for Phase 2 (which begins at 25%). The tank will move in two steps to get around the pillar. - When the boss gets to 25% HP, ranged will follow the main tank to the corner and stack up right behind the boss. They will also move in two steps. - There is no method for melee since they will just naturally follow the boss anyway. ### Lady Vashj **Phase 1**: - The main tank will tank Vashj in the center of the arena. - If a Shaman is in the main tank's group, that Shaman will attempt to keep a Grounding Totem down in range of the main tank to absorb Shock Blast. This should continue in Phase 3. - Ranged bots will spread out in a semicircle around the center of the arena. - If any bot other than the main tank gets Static Charge, it will run away from other bots. If the main tank gets Static Charge, other bots will run away from the main tank. This method should continue in Phase 3. - If any bot is Entangled and has Static Charge, the bot will attempt to use Cloak of Shadows if it is a Rogue, and Paladins will attempt to use Hand of Freedom. This method should continue in Phase 3 (with some modifications). - Bots will not use Bloodlust or Heroism (saved for Phase 3). Bots will not use any other major cooldowns, either, such as Metamorphosis (saved for Phase 2 and 3). **Phase 2**: There are two central mechanics to this phase, both of which were challenging to get bots to execute properly. First is the system of prioritizing adds. The large playing field and multiple types of adds coming from random directions make this phase not doable with realistic DPS under the standard Playerbots target selection system. Therefore, I took inspiration from liyunfan's Naxx strategy for Phase 1 of Kel'Thuzad to disable dps assist and create a custom target selection system. First, a cheat with respect to the Coilfang Striders: - Tanks will permanently have the Fear Ward aura applied to them if you have raid cheats enabled. This allows them to tank the Coilfang Striders. The standard strategy was to have an Elemental Shaman kite the Strider around the perimeter of the arena, with ranged players (including healers) spamming DoTs on the Strider. If you can make bots do this, then great, but it's far beyond my capabilities. Therefore, with the cheat, the first assistant tank is responsible for tanking Striders and keeping them away from Core passers (described below) and Vashj. Evidently it was (and is, in TBC Classic) possible to tank (and melee DPS) Striders by wearing a Dire Maul Ogre Suit, which would give you enough reach to stay out of the Strider's fear. I actually tried that, and it does not work, either because AC's radiuses are not the same or just because bots do not maintain the same level of precise positioning. But anyway, the point is that technically the Striders are tankable by real players, so maybe that will make you feel better about using this cheat (it's fine enough rationalizing for me). I found this fight to be unmanageable without this cheat (i.e., using a method that would only have bots try to run away from Striders) because each Strider was guaranteed to wipe out a couple of bots, and you really cannot afford to lose anyone. YMMV though. - If cheats are enabled for Striders, Hunters will attempt to Misdirect the Striders to the first assist tank. - If cheats are not enabled, bots will attempt to use slows/roots to stop the Striders. I have some logic for them to use Netherweave Nets, but I suspect it does not actually work so I may remove it instead of trying to get it to function properly. Target priority is as follows: - Hunters and Mages: Enchanted Elementals, Coilfang Striders, Coilfang Elites. - Other Ranged Bots: Elites, Striders, Elementals. - Melee DPS: Elementals, Elites. - Tanks: Elites, Elementals (except if cheats are enabled, the first assistant tank will instead prioritize Striders and then Elementals) - Everybody else (basically means healers): Elementals, Elites, Striders - If there is more than one of the same target, bots will prioritize the one that is closer to Vashj. - In all cases, the valid attack ranged is limited so that bots should not leave the central platform. - If somehow a bot ends up too far from the center of the room and is not actively attacking anything, there is logic to make them run back. Handling Tainted Elementals and the Tainted Core: I will make another post about this later. It is easily the most complicated strategy I've ever worked on (far beyond anything on Kael'thas even) so will necessitate a long explanation. The tl;dr is that there is a chain of two-to-four bots that receive/pass the Tainted Core before using it on a Shield Generator, and if you are playing by yourself, you probably need to turn raid cheats on, in which case there will also be a bot that teleports to, kills, and loots the Tainted Elementals (i.e., the bots will then handle the entire sequence of shutting down Shield Generators). **Phase 3**: - The main tank will pick up Vashj immediately and try to keep her away from Enchanted Elementals. - DPS will burn down residual adds from Phase 2 in the order of (1) elementals, (2) strider for ranged only (if you have more than one up, you're dead), and (3) elites (hopefully you have only one up, but two with one almost dead is possible). - Hunters will kill Toxic Sporebats. This works quite well, but they (and anybody else if ordered to target Sporebats) have a tendency to levitate up into the pipes at the top of the room when killing the Sporebats. To counteract this, a method forcibly teleports bots to the ground if they get more than 2 yards above the ground. - The Phase 1 Cloak of Shadows/Hand of Freedom method is now expanded to include bots Entangled in the Sporebat poison pools (with Hand of Freedom usage prioritized on the main tank). - There is a specific method to avoid the Sporebat poison pools. The Vashj tank will move backwards when avoiding poison. --------- Co-authored-by: kadeshar <kadeshar@gmail.com>	2026-02-08 12:31:23 +01:00