⚠ Speculative reconstruction. This guide is built from a static snapshot of the game's data and may be incomplete or wrong. Gem names, success rates, costs, and formulas are inferred and can change with any game update. Treat everything here as a working model, not official data. Not affiliated with the developer.
Gem chest

The Gem System

A monetized gear-augment system: gems inlay into equipment for permanent stats, then merge upward through a probabilistic ladder. This guide documents the mechanics, the cost structure, and the point of diminishing returns.

Unlocks at player level 100 Live cap: T12 (T13+ unreachable) 8 families, 15 tiers in config
One merge

How gems work

A gem inlays for a permanent passive stat and grows by merging: a lower-rank feeder and the primary combine into the next rank.

Primary gem
the one you keep, rank n−1
+
Feeder gem
1 feeder (2 for T2 and T3)
↓  roll
Success
primary becomes rank n
Fail = Explode
only the feeder is lost; primary kept, pity +1
T1 to T5 Alchemy Pot, always succeeds. T6+ gem-smith, can fail. Pity guarantees success after K misses (story-gated)

Your primary is never destroyed and never de-ranks in the current version: every gamble rank's fail result is Explode (feeder only). A primary-de-rank outcome exists in an older (May 2025) build, but no current data selects it. SRC gem_merge.csv (FailedResult=Explode ×10) de-rank: old-build / server-side

When it opens. The whole gem system unlocks at player level 100, a flat level gate (the Silver to Gold rank boundary), not a zone or quest. Then each gear piece earns its first socket once that piece reaches equip level 100, and up to three more sockets as the piece hits equip level 130, 160, and 190. SRC function_limit.csv + conditions.csv
Pick your stat

The eight families

One stat theme each. Two-stat families carry a Final-grade secondary that is the real differentiator (see Picking a family). Tap a gem for detail.

Which gem fits which slot

The climb

The merge ladder

Rate, pity and feeder tier per rank. It collapses twice: T6 to T10 fed by T5 gems, then T11 to T15 fed by T10 gems (why the top explodes in cost). Pity caps the worst case at K+1 tries.

See it in game

The actual UI (with translation)

CN screenshots with English overlaid on each label. Older build (May 2025): layout is current, but values and one gem name predate a rebalance: use the calculator for live numbers.

In-game gem inlay panel Thunderous Spirit Staff T4 Ruby: ATK 6.3% Unequip T3 Ruby: ATK 4.5% Unequip T4 Black Crystal: DMG Boost 1.3% Unequip Slot enhance: unlocks at Lv120 T1 Ruby T1 Black Crystal T1 Amethyst Enhance Inlay Resonance Gift
Equipped sockets (3 filled)
4th socket, locked by level
Your gem inventory
Tab bar: Enhance / Inlay / Resonance / Gift
Inlay panel. 4 sockets/slot, 4th unlocks by level. Drift: “Black Crystal” is the old name for Obsidian; Ruby T3/T4 shown 4.5%/6.3% vs current 6%/12%.
In-game gem merge furnace Mechanical Furnace 225K Inlay T9 Yellow Gem T5 Yellow Gem Furnace charge (pity) 60% T9 T10 If it fails you lose the feeder. Choose carefully! Feeder candidates: “wrong type” (T7 gems) “wrong type” (T8 gems); checkmark = the one valid feeder Success rate 10% Synthesize (30K)
Primary gem (being upgraded)
Pity / charge bar
Feeder candidates
Success rate + Merge button
Merge furnace. The bar is the pity meter (matches the modeled charge value). Drift: T10 shown at 10% vs current 4% base.

Images from 3DM, local reference for personal study; English is overlaid, not edited onto the source. DERIVED translation UNVERIFIED values (older build)

Try it yourself

Furnace simulator

A functional replica of the Mechanical Furnace. Success rates, pity counters, feeder requirements, and failure behavior are taken directly from the game's data. Acquire gems through the modeled channels, then merge.

Inventory (click a stack to load it as the primary)

Get gems

The real acquisition channels. Each box opens to a random family with the game's real weights: 70% single-stat (“avoid”) families, 7.5% per two-stat family.

Sandbox (not in game):

Log

    Mechanical Furnace 机关熔炉

    Exact, from the game's data: success rates (100/36/18/8/4), pity counts (1/2/5/11/23), the charge meter (misses/pity, reset on success), feeder tier and count, same-family feeding, Explode-on-failure (feeder lost, primary retained), per-target-tier charge shared across families, the story gate, the T12 cap, pack contents and US prices, and the 700/300 family weights. Assumed: per-run dungeon drop quantity and the RNG draw, both server-side. SRC game data tables UNVERIFIED: drop count, draw

    Deep dive

    Cost calculator

    Priced in T1-gem-equivalents (the whole feeder + primary sub-tree, pity-capped). Drag a target tier.

    Total cost
    T1-gem-equivalents
    This step
    marginal T1-equiv
    Expected attempts
    at this rank
    Efficiency
    Ruby ATK% per 100 T1-equiv

    Full cost & efficiency table

    TierSuccessPity KExp. attempts FeederCumulative costRuby ATK%% / 100 cost

    Each tier approximately doubles both the stat gain and the cumulative cost; efficiency therefore declines monotonically, with no local optimum. DERIVED from gem_merge.csv + gems_catalog.csv

    Choosing a family

    Why the big number doesn't decide

    A T12 gem prints "+307% ATK", but the line scales only the character's base stat, a minor component of total attack at endgame. Its real contribution is approximately +10%. That figure is substantial, yet it is near-identical across all eight families, so it carries no weight in family selection.

    Every family prints the same headline

    Ruby, +307.5% ATK
    about +10% real ATK
    Obsidian, +301.5% ATK
    about +10% real ATK
    Amethyst, +301.5% SPD
    about +10% real SPD

    All eight families' main lines land within a hair of each other, so the headline is the same whichever you pick. It cannot be the thing that decides your choice. SRC gems_catalog.csv

    TL;DR: The headline is the same on every gem, so ignore it. Pick the family whose second stat your build wants.
    Where gems come from

    Acquisition

    There is no gem gacha, and merging consumes no currency. The only cost is the volume of low-rank gems consumed, so monetization operates through acquisition rate rather than direct purchase.

    Free / F2P
    • Co-op gem dungeon. 黄泉阁 (“Hades Pavilion”), a Silver-rank multiplayer daily with Normal→Hell tiers, the main drop source. SRC daily_dungeon + 3DM
    • Gem ore (宝石矿) from idle mop-up / material realms, “very low refresh chance.” SRC item.csv MopUpMaterial
    • Gems drop directly (CanDrop=TRUE); steady but slow and run-capped.
    Paid / P2W
    • Paid gift packs carry random-gem boxes: $4.99 → 8× T2 box, $14.99 → 9× T2 + 4× T3, $29.99 → 5× T3 + 3× T4 (US storefront; ¥30/98/198 CN). SRC limited_time_gift_pay_item + money_cost.csv
    • Cumulative-recharge reward tracks include gems. SRC accumulated_pay_award
    • No direct “buy a T12”: you buy throughput of low gems and feed the geometric merge cost (a T12 = ~4,583 T1-equiv).
    The draw is biased toward the weak families. Every random-gem source (boxes, ore, bags) applies family weights of 700/300: the four single-stat families account for 70% of draws, each two-stat family for 7.5%. No conversion path exists. Feeders must share the family; gems cannot be deconstructed, traded, or sold. Off-family gems serve only their own ladder, spare sockets, or a later progression-gated recycle. SRC reward_rule.csv 64500-64530, shop_goods.csv
    The three gem pack SKUs above are config-exact; other rotating bundles and the per-run dungeon drop count are season-bound / server-side, UNVERIFIED. Drop units from ore/bags: small ore 1× T1, large ore 1 to 2× T2, gem bags 1 to 6× T1. SRC reward_rule.csv 151-156
    The verdict

    How to invest

    • Spread before you peak. Cost compounds per gem; stats are linear. Many mid gems > one high gem.
    • T6 is the knee. Free 100% merge: get every slot to T6 or T7 first.
    • Pick by the secondary line, not the headline Base% number.
    • Don't gamble T7+ before the story unlock (no pity floor yet).

    Cost scales geometrically in T1-gem-equivalents. By T12 the stat per unit cost is roughly 1/7 that of T5, so breadth across slots, combined with the Final-grade secondaries, outperforms maximizing any single gem, including for maximal spenders.

    By spend bracket

    F2P
    • Stop at T6 or T7 on every slot.
    • Go no higher; the tail is not worth your gems.
    Light spender
    • T8 or T9 on your main weapon family.
    • T6 or T7 everywhere else.
    Whale
    • T10 to T12 on 2 or 3 highest-impact slots (weapon Obsidian / Amethyst).
    • Keep the rest at T7 or T8.
    Everyone
    • DPS: Obsidian + Amethyst.
    • Tank: Moonstone over pure Sapphire/Beryl.
    Cross-checked

    Verified against community sources

    Family names, stat lines, slot rules and merge structure all confirmed against the authoritative CN community for 杖剑传说 (the real CN title).

    TapTap (primary)

    Sorcerer gem guide

    Lists all eight gemstone families with their exact stat lines and the weapon/off-hand/boot/helmet/armor slot rules. In-game inlay screenshots. Matches our data point-for-point.
    3DM (mechanics)

    Gem inlay & merge guide

    Confirms the 3-into-1 then 2-into-1 merge, the tier-5/6 gamble boundary at the gem-smith, four sockets per slot gated by resonance, and failure that destroys the feeder and can drop the gem a rank.
    TapTap (build)

    Full build / gem / enhance guide

    Broader progression context with gem, gear and enhancement screenshots.
    Do not trust

    English SEO guides

    Third-party emulator-vendor blogs describe a generic gem taxonomy (Strength / Wisdom / Critical / Lifesteal / Haste…) that does not exist in this game. Templated AI content; disregard for gem facts.

    The community pages do not publish the exact success rates (36/18/8/4%), pity counts, or the T12 cap; those were derived from the game's own data. SRC game data tables