Mechanics history and comparison guide
The local download has now been decoded beyond the original samples. See downloaded AEGIS material for Alpha maps, offline database exports, full workbook extraction and the remaining outside-evidence gaps. These are candidate inputs; their presence does not establish Thai effective dates.
This reference follows recovery, EXP, monsters, drops and consumables from the international Alpha/Beta preludes through Thailand’s last pre-Renewal era. It supports the developer and player notes attached to each patch. Thai beta’s natural HP/SP recovery formulas remain unresolved. Original Thai operator guides now supply passive-skill tables, but their publication and effective dates are unknown. Historical formulas below are candidates or observations of particular emulator revisions, with their scope stated beside them.
The timeline ledger owns patch dates and Episode labels. A guide’s publication date, an archive capture, a scheduled maintenance and a confirmed deployment are different events. Content can arrive before or after its associated balance change. Thai-language coverage of an English or Korean server does not establish a Thai release.
Progression overview
| Stage | Mechanics evidence and practical consequence |
|---|---|
| International Alpha | A guide updated September 22, 2001 describes early healing items and qualitative monster drops. The launch date and numerical regeneration, EXP and drop rules are unverified. This is a separate reconstruction before the Thai service. |
| International Beta 1 | Skills and Mage progression appear in contemporary guides and community chronology. Early SP evidence describes time per recovered point; passive SP Recovery is a separate system. |
| International Beta 2 and July fixes | Preserved material describes recovery-skill, potion, death-penalty and monster changes. Original Thai operator pages explicitly discuss the English Beta. Their rules must not silently become Thai CBT defaults. |
| Thai CBT and OBT, provisional Episode 1.0 | CBT began September 16, 2002. The launch records establish service and wipe policy; they do not establish the full mechanics. OBT’s October 25 date remains a mirror-supported lead. |
| Clock Tower and Glast Heim, provisional Episode 1.5 | New hunting areas require their own monster, spawn and reward records. A later map’s availability does not date every monster’s balance revision. |
| Pets and rebalance, provisional Episode 2.0 | A notice mirror reports selective monster/card-drop changes. The affected records and absolute rates remain incomplete. |
| Comodo, Episode 3.0 label lead | Skill, status, party and refining changes need Thai-specific deltas. The reported June 30 test and July 16, 2003 production dates are not verified dates for every mechanic. |
| Turtle Island and WoE, Episode 4.0 label lead | Monster skills, boss behavior and job-change quests become separate comparison targets. Thai guides captured in December 2003 document weight and EXP examples, without identifying when those rules started. |
| Juno, Episode 5.0 label lead | Later guides associate a different recovery system with Juno. Thailand’s reported January 26, 2004 schedule does not yet date the recovery transition. Monster rewards, placement, item weights and loot rights require separate checks. |
| Rebirth, Episode 9.0 label lead | Keep transcendent EXP, HP/SP scaling and job skills separate from ordinary classes. Temporary EXP events and Sakray rates are additional dated settings. |
| Lighthalzen, Episode 10.2, onward | Homunculus, later job skills, equipment and consumables introduce more recovery and reward interactions. Their appearance does not prove that the underlying player formula changed. |
| Veins, Episode 11.2 | Contemporary Thai coverage warns that boss HP may already differ from the advertised foreign changes. Avoid applying one blanket boss adjustment. |
| Morroc, Episode 12.1, and Memorial Dungeons, Episode 12.2 | World replacement, battleground rewards and instance access add distinct progression rules. Ordinary field spawns, instance populations and reward tables need separate records. |
| Encounter, Episode 13.2, to final pre-Renewal | Preserve the last major content stage separately from later weekly patches. The cutoff is immediately before the November 9, 2010 Renewal maintenance; the exact final preceding server snapshot remains unresolved. |
Episode labels in this table inherit the qualifications on the linked patch pages. They are organizational references, not proof that another region’s identically numbered episode had the same rules.
The recovered July 22, 2002 operator FAQ, section 6, says the planned initial Thai service would be Beta 2 and free during beta. Its tentative mid-August opening was a plan, superseded by the September launch evidence. It identifies an intended build family; it does not make every English Beta 2 rule an authenticated Thai rule.
Natural HP and SP recovery
Natural regeneration, passive recovery skills, healing consumables and recovery modifiers must be tracked independently. They can run on different timers and obey different restrictions.
Early observations and unresolved formulas
| Evidence | What it supports | What it does not settle |
|---|---|---|
| March 1, 2002 Japanese discussion, post 8 | Approximate standing SP interval: 7.54 − 0.06 × INT seconds per point; the table doubles the recovery rate while sitting. | Server/build, Thai applicability, high-INT cap, rounding and timer behavior. |
| March 22 discussion, post 18 | Alternative rate expression: 60 / (3.76 − 0.03 × INT). Its denominator corresponds to a slightly different 7.52 standing offset if interpreted as the sitting rate. | Pose and units are not restated in that post. The small offset disagreement cannot establish a patch change. |
| April 18 discussion, post 25 | A relayed Mage-thread report says recovery remained positive at INT 126. | Exact cap or recovery speed; the original measurement is missing. An unclamped interval crossing zero is unsafe to extrapolate. |
| February 13–14, 2003 discussion, posts 114–115 | Participants ask for the HP formula and describe it as still undetermined, tentatively mentioning MaxHP, level and VIT. | The tentative dependencies are not a formula. This is uncertainty in one contemporary discussion, not proof that no formula was known elsewhere. |
| Thai operator leveling guide, captured December 12, 2003 | Red weight stops natural HP/SP recovery; wording says “over 50%.” | Exact > versus >= boundary, introduction date, CBT applicability and passive-skill exceptions. |
The Japanese thread title refers to mathematical formulas; it is not an official operator thread. A later post about congestion also warns that observed recovery timing could be distorted by server load. Memory of slow recovery alone cannot distinguish an intended rule from delayed updates.
More Thai guides with unknown effective dates
Four additional original operator guides were recovered in round six. Their December 2003 captures do not establish that the described rules were active during Thai CBT in September 2002.
| Guide | What the page says | What remains open |
|---|---|---|
| Interface | Weight over 50% stops HP/SP from increasing on their own; items or skills may restore it. Over 90% prevents attacks and skill use. | Natural amount and interval are absent. Exact equality at 50%/90%, passive-skill exceptions and introduction dates remain unknown. |
| Basic play | Discusses returning to the saved point or waiting for another player to revive you, then gives 50% HP for Novice / 1 HP for other classes without separating the routes. It describes 1% Base/Job loss on death. | This is ambiguous death/revival wording, not a universal Resurrection skill rule or natural regeneration formula. EXP denominator, rounding and effective date are not supplied. It cannot date a change from the Beta 2 mirror's 5% claim. |
| Training Ground | Registration grants 50 zeny; players may skip to a random town; three training stages are described. | This guide is separate from the missing September 18, 2002 Training Ground problem notice. No beta-effective date is given. |
| Training course | Fabre in the second training area advances Base progress but not Job progress. | No numerical EXP reward or effective date. Do not copy this tutorial rule into the ordinary field Fabre record. |
These pages narrow the questions to test later, but Thai beta natural HP/SP amount and interval formulas remain unverified.
Historical Athena candidate
Athena Dev 2.1.1, attributed to July 2003, contains this inspectable implementation. Its archive date is not independently authenticated, and its behavior is not an official Thai specification. The extraction, hash and source passages are in athena-2003-candidate.json.
| Component | Implementation in that artifact |
|---|---|
| HP denominator | max(300, 3000 − 15 × effective_VIT − 14 × BaseLevel) |
| SP denominator | max(300, 7500 − 60 × effective_INT) |
| Sitting | Integer-divide the clamped denominator by two. |
| Every 250 ms | Add floor(1000000 / denominator) + 1 to the relevant accumulator. |
| Applying recovery | Each 4000 accumulated units permits one point; apply/report on every fourth tick. |
| Function gates | Dead or weight at least 50%: stop; HP additionally requires not walking. |
There is no natural-recovery MaxHP/MaxSP term in these functions. The SP denominator reaches its minimum at INT 120. These details identify this candidate precisely; they do not resolve the different contemporary estimates. Replacing a modern eight-second setting with a smaller number would not reproduce its accumulator and publication behavior.
Pinned modern emulator comparison
These are unmodified player bases before recovery modifiers, taken from rAthena d4b8e7b and Hercules 04a422ce. They are comparison baselines, not a dated Thai production snapshot. floor means discard the fractional part.
| Rule | rAthena | Hercules |
|---|---|---|
| Natural HP amount | floor(VIT/5) + max(1, floor(MaxHP/200)) | 1 + floor(VIT/5) + floor(MaxHP/200) |
| Natural SP amount | 1 + floor(INT/6) + floor(MaxSP/100) | Same |
| Extra SP at INT ≥120 | Add floor((INT−120)/2) + 4 | Same |
| Default natural interval | HP 6000 ms; SP 8000 ms | Same |
| Default recovery-skill interval | 10000 ms | Same |
At VIT 10 and MaxHP 1000, the HP amounts are respectively 7 and 8. Both give 1 at VIT 1 and MaxHP 100. This difference needs historical evidence to adjudicate; choosing pre-Renewal mode does not make the two emulators identical. Defaults, modifiers and timer scheduling are separate from these amount expressions. See the source audit and arithmetic comparison.
The transition date remains open
A period Knight guide labels later recovery formulas as Juno, while a Hunter guide describes movement from individual points to larger chunks. A May 13, 2003 Japanese discussion explicitly references a Korean SP revision. None supplies the Thai maintenance that activated the replacement. Keep the before/after policies unassigned to an exact Thai date until a notice or dated Thai measurement connects them.
Passive recovery skills
An early Mage guide and July 2003 Priest guide describe a flat SP Recovery candidate of 3 × skill level per ten seconds. The Priest guide’s forecast of a future formula is not a measured Thai rule.
Original Thai operator skill tables
The newly recovered Swordman guide and Mage guide were captured on December 13, 2003. Neither page gives a publication date or activation date. Both describe passive pulses every ten seconds while sitting or standing still.
| Skill level | Swordman HP table | Mage SP table |
|---|---|---|
| 1 | 5 | 3 |
| 2 | 10 | 6 |
| 3 | 15 | 9 |
| 4 | 20 | 12 |
| 5 | 25 | 15 |
| 6 | 30 | 18 |
| 7 | 35 | 21 |
| 8 | 40 | 24 |
| 9 | 45 | 27 |
| 10 | 50 | 30 |
These printed tables correspond to 5 × skill_level HP and 3 × skill_level SP. The Swordman text also mentions VIT influence without a coefficient and a potion bonus of 10% per skill level. Neither table prints a MaxHP/MaxSP term. An omitted term does not prove that the live server lacked it. These are skill descriptions, not natural-regeneration formulas.
There is a conflict to resolve: the preserved BEEhive Beta 2 notice describes Swordman recovery gaining MaxHP scaling from a Beta 1 flat 5–50 table. It is a fansite republication, with no verified Thai activation date. The operator table may be stale or simplified; its later archive capture does not settle which behavior Thai CBT used. Keep both claims and their provenance. The first-job guide also contains conditional wording about future quest functionality, reinforcing the need to date the guide content independently.
Modern skill comparison
At the pinned modern revisions, the base skill amounts are:
- HP:
5 × skill_level + floor(skill_level × MaxHP / 500). - SP:
3 × skill_level + floor(skill_level × MaxSP / 500).
Thus level-10 SP Recovery at MaxSP 500 gives 40 per skill tick under the modern expression, versus 30 under the flat candidate. Standing/sitting, walking, weight, poison, buffs and timer resets still need separate validation. A correct natural-regeneration policy can coexist with an incorrect passive skill.
EXP requirements and awarded EXP
Level requirements determine how much EXP is needed; monsters and allocation rules determine how quickly it is earned. They must have separate version histories.
The historical Athena artifact has four columns: Base, Novice Job, First Job and Second Job. Its consumers index the requirement by the current level, not a cumulative total. Among the compared playable requirements, both modern ordinary pre-Renewal tables differ only at Base level 98 (99,999,999 in the archive versus 99,999,998) and Novice Job level 2 (20 versus 18). First- and second-job comparisons match. These differences might be implementation errors; they are not proven official balance changes. Cap sentinels must not become real level requirements. Comparison records.
The Thai operator guide captured in December 2003 gives two useful allocation examples: equal damage awards 66.6 versus 33.3 to the first and second attacker, and each additional attacker increases monster EXP by 10%. Their interaction and rounding are unspecified. The party prose also mixes equal and level-dependent division. These examples are test targets; they do not define a complete algorithm.
The Beta 2 mirror’s 5% death-EXP loss and dungeon item-loss claim remain foreign-era candidates with unclear denominator and Thai applicability. For Rebirth, keep transcendent tables distinct and record temporary double EXP or Sakray multipliers as dated server events. They cannot explain ordinary beta rates without separate evidence.
Before a stage can claim reproducible leveling, record its level requirements, monster BaseEXP/JobEXP, contribution and party rules, death penalties, MVP/quest awards, level-up overflow behavior, and event multipliers.
Monster stats and drops
RagnaDB exposes episode-labelled datasets, but the inspected bundle supplies no beta-specific monster dataset or authenticated Thai distribution. Its Episode-1 candidate is attributed to a December 2002 archive; later sets include reconstructed packs. Those labels are provenance leads. The extracted candidate records retain their attribution and raw fields.
| Monster | Episode-1 candidate HP / BaseEXP / JobEXP | Pinned modern pre-Renewal HP / BaseEXP / JobEXP |
|---|---|---|
| Poring | 50 / 2 / 1 | 50 / 2 / 1 |
| Fabre | 63 / 3 / 2 | 63 / 3 / 2 |
| Mummy | 4837 / 717 / 448 | 5176 / 800 / 602 |
| Doppelganger | 12000 / 10680 / 3872 | 249000 / 51480 / 10725 |
| Savage | 1901 / 357 / 226 | 2092 / 521 / 248 |
The selective differences rule out reproducing this candidate with one universal EXP or HP multiplier. They also require combat attributes, race/size/element, AI, skills, movement/attack timing and spawn records to be compared alongside rewards.
Several import traps remain:
- A field called
MaxAtkcan be smaller thanMinAtk: candidate Mummy has 328/55, while historical Athena stores endpoints 328/383. Base-plus-spread is a plausible interpretation, not yet an authenticated schema. - A sampled Episode-8.2 dataset gives Poring EXP 14/7 rather than 2/1. Its sevenfold difference has no established cause; normalization would conceal the uncertainty.
- Preserve repeated item drop slots and zero-rate placeholders in raw evidence. A map keyed only by item ID loses slot structure; whether slots roll independently still requires implementation evidence.
- Raw rates need their denominator, inclusive/exclusive random comparison, clamps, server multipliers, bonuses, steal rules, MVP ordering and loot-ownership timers. Equal stored numbers can produce different effective behavior.
The Thai announcement mirror reports a tenfold Poring-card increase in the early rebalance. It does not establish the absolute before/after rates or justify a global tenfold card increase. Veins coverage likewise warns that Thai boss HP may already have changed. Retain these as individual change claims until the affected tables are recovered.
Official Rachel quest and boss values
Two newly recovered operator guides add specific values to the Rachel chapter. Their pages have no publication dates; archive captures on June 28 and July 4, 2007 establish only when these copies existed. They do not settle the Thai 11.1 launch date or establish earlier beta values.
| Operator guide | Values printed | Limits |
|---|---|---|
| Quest 373: Ice Necklace | Hammer[2] ×1, Rough Wind ×5, Empty Scroll ×1; four Freezing Snow Powder when helping Maheo, then 700,000 EXP on returning the necklace to Urstialla. Mohadian buys Glacial Heart for 375 zeny each. | EXP is not labelled Base/Job; no minimum level is stated in the readable guide. The website uses airplane.gat as an explicit placeholder because its map list lacks Rachel. |
| Quest 380: Ktullanux | Complete the preceding quest and use four powders. The four fires return two hours after the boss is killed. Ktullanux: 4,417,000 HP, Water level 4, Brute. | No complete combat/drop table. The two guides differ slightly in entrance coordinates; preserve each guide's values until checked. |
For later implementation, keep the quest reward, boss record and fire-reset timer separate. Do not convert the unspecified EXP into both Base and Job rewards, or treat the fire reset as an ordinary monster respawn interval.
Potions and recovery cards
The English Alpha guide, updated September 22, 2001, reports Red Potion at 30 zeny for 20–30 HP and Blue Potion at 1000 zeny for 50–60 SP, with potions weighing 5. These belong to the international Alpha observation, not Thai beta. The English Beta fix note reports an 8-zeny Milk price change on July 22, 2002.
Do not add a Red Potion change from 50 to 60 zeny based on the July 22 FAQ: section 3 explicitly uses that change as a hypothetical example of why client patching matters.
The Thai recovery-item guide captured December 13, 2003 instead lists:
| Item | Approximate recovery | Listed price in zeny | Listed weight |
|---|---|---|---|
| Red Potion | 30 HP | 50 | 10 |
| Orange Potion | 50 HP | 200 | 13 |
| Yellow Potion | 80 HP | 550 | 16 |
| White Potion | 120 HP | 1200 | 20 |
| Blue Potion | 40 SP | 5000 | 20 |
| Milk | 15 HP | 25 | 5 |
These approximate descriptions are not healing ranges, and listed prices do not establish stock at a particular NPC. The content may predate its capture. No synchronized server version, VIT/INT scaling or skill multiplier accompanies the table.
The neighboring card guide lists Sohee MaxSP +25%, Eggyra SP recovery +25% and Willow MaxSP +30. Their exact activation dates, complete effects and rounding remain unresolved. Card effect changes must be separated from card drop-rate changes.
Official iRO Juno notes describe reduced healing-item weights and monster/reward/loot changes. This is useful foreign comparison material, without a Thai effective date. Weight changes affect carrying capacity and the point at which regeneration stops; potion economics and recovery therefore need to be checked together.
Requirements before future emulator work
Each selected patch needs a mechanics record with its predecessor, region/server, effective-date status, evidence links, exact numerical rules and any explicitly chosen reconstruction assumptions. A patch may carry forward a previous rule only as a stated reconstruction decision when continuity is unproven. Never convert “no notice recovered” into “unchanged.”
| Comparison | Minimum useful cases | Result to retain |
|---|---|---|
| Natural recovery | Vary VIT/INT, BaseLevel and MaxHP/MaxSP independently; standing, sitting and moving; weight below/at/above 50%; INT around 120–126. | Timestamped amounts, internal ticks if available, client display timing, rounding, clamps and uncertainty. |
| Recovery transitions | Sit/stand midway through a tick; start/stop walking; change equipment, map, weight or status; reach full HP/SP. | Whether progress is kept, reset, suspended or recalculated. |
| Passive skills and items | Skill levels 0/1/10; maxima around formula boundaries; identical stats with different cards; potion samples with and without skill bonuses. | Separate natural and skill contributions, item ranges and the order of multipliers. |
| EXP allocation | Solo; equal damage with reversed first hit; unequal damage; third attacker; party/non-party; overkill; death and level-up boundaries. | Awarded Base/Job EXP for each participant and all relevant rates. |
| Monsters and drops | Compare the sample monsters plus each patch’s changed bosses; preserve every drop slot; inspect rate processing before probability sampling. | Per-record diffs, source schema, AI/spawn behavior and effective probabilities. Rare drops need a statistical plan, not a short kill session. |
| Patch rollout | Open an old save under the next stage; check classes, map access, inventory, cards, quests and active timers. | Migration decisions, player-visible changes, rollback behavior and unresolved historical choices. |
For players, publish only the mechanics selected for that rollout, clearly marking any approximation: how resting changes, what earns more or less EXP, which monsters or drops change, and whether existing characters or items are affected. Keep research candidates out of definitive player-facing patch claims.
Detailed evidence is maintained in recovery-evidence.json, the source register, the additional contemporary recovery findings, and the recovered original notices and skill guides. The Round 5 packet includes saved HTML, retrieval URLs, redirect destinations, timestamps, encodings and SHA-256 hashes. No emulator changes are part of this reference.
Extracted local database candidates
The deep inspection decoded 83 unique MDF variants into 1,017 table catalogs and 75,348 primary-row candidates across snapshots, with 5,882 candidate procedure/view texts. These totals include repeated schemas and administration/account data, not independent historical gameplay records. Full offline exports are retained locally; the database ledger exposes only metadata and selected gameplay samples.
The candidates materially disagree: Poring raw EXP 14 versus 2, Mummy raw EXP 800 versus 488, Poring-card raw rate 20 versus 1, and Red Potion raw weight 10 versus 70 occur in different packages. ASB4's database Poring HP 50 also conflicts with its loose text value 25. Raw units, loader usage, region and effective revision must be established before converting these into stage rules.
Three real-file decoder controls and 90 cross-format monster-field comparisons passed. LDF log replay, engine recovery, uncommon type/LOB cases, unnamed physical tails and one card-row cardinality disagreement remain explicit local verification work. Natural HP/SP recovery formulas and the Thai beta profile are not established by these exports. See the extraction appendix and validation.
Native recovery formulas now traced in later candidates
The native follow-up establishes both NEWHPSP branches in a matched 2006 Russian candidate and compares the core with a 2013 Malaysia/Singapore candidate. With acceleration bonuses zero and ordinary nonnegative inputs, new-mode natural HP is max(1,floor(MaxHP/200))+floor(VIT/5); SP is floor(MaxSP/100)+1+floor(INT/6), with an extra floor((INT-112)/2) when INT is at least 120. Ordinary thresholds are HP 6,000/3,000 ms and SP 8,000/4,000 ms for standing/sitting, before special modifiers. Old mode restores one point with VIT/level- or INT-dependent thresholds. These are candidate binary semantics, without a Thai effective date.
The full arithmetic and state qualifications retain fractional HP terms through acceleration, the high-INT bonus order, strict elapsed-time comparison, movement/weight/Monk exceptions and conditional natural caps. Passive skills use separate paths with a supplied-maximum clamp. A 2006 old-mode branch is not evidence of Alpha behavior. The 100-span/two-pair/14-arithmetic validation does not execute the target or authenticate period deployment.
All 83 LDF variants now have offline record recovery, with 8,226 bounded row images matching paired candidate MDFs. This closes hash-only coverage; redo/undo, active-log selection, checksum/GUID checks and SQL-engine consistency remain local work. Historical Thai formula/configuration binding, early-period engine behavior and actual transition dates remain outside-source gaps.