GearTools

0.5.2 — 2026-08-18

First public release. Everything below came out of testing the addon in a live client, one report at a time.

New: search for the stats you want

A second window takes stat targets and a priority order and finds the gear that reaches them.

• Targets outrank priority. The search spends as much of your stat ordering as a target costs, and only then uses the ordering to decide what to do with what is left over. A target it cannot reach is reported, with the shortfall, rather than quietly answered with something close.
• Advanced mode replaces the ordering with four explicit weights, the way a simulator reports them. The ordering generates weights; the weights are the real input, and each place's multiplier is shown on its button.
• Search your bags or the season. "Equipped and bags" answers what can I wear right now and is the default; "season database" answers what should I aim for. Bank and void storage are not readable, so the option says so.
• The rows start from what you are wearing, so the first search is a comparison rather than a blank slate. Lock any slot you have decided, change a target, and search again.
• Crafted gear's two secondaries are a choice the search makes, and the result names the pair to ask the crafter for.
• A progress bar, and a cancel button. The search runs on a frame budget and yields rather than freezing the client.

The numbers now match your character sheet

Four separate causes, all of which made GearTools read lower than the game.

• Gems and enchants were not counted at all. The client reports an item's own stat allocation; a gem is a separate item in a socket and an enchant a separate effect. The gap is now measured against your real totals and added to both columns, so the current column reproduces the character sheet exactly.
• The percentages were the part the rating contributes, not the sheet's number. Crit adds a class base, haste compounds with everything else granting it, and mastery is reported as an effect rather than as points. Percent targets mean the sheet's number too now.
• Gear from an earlier season was dropped from the plan while the current column counted it. Everything you are wearing is imported; a piece this season's database does not list is flagged and counted, just not re-rankable.
• Diminishing returns are modelled, not only flagged: percent targets convert through the curve, and the search scores effective rating.

The season database

• Delves and world bosses are back, banded at Hero 6/6. They were excluded because an unbanded delve item would have been offered at Myth 6/6 — a combination that cannot exist. 599 items to 753.
• The primary stat is filtered. A Retribution Paladin is no longer offered an Agility polearm, which the stat scoring could not see for itself.
• Weapon rules are per spec. No one-handers or shields for a two-hand spec, and no two-handers for a shield tank.

Fixes

• The Target item level could be badly wrong and look right. Two above-track items rejected upstream contributed 0 apiece while the average still divided by all sixteen slots: 291.9 where the real answer was 334.9. The average now reports how many slots are behind it, so an incomplete plan reads Item Level 14/16 rather than a figure that looks complete.
• Above-track items were unplannable. The slot row compared the chosen rank against the group's maximum, so rank 9 against a maximum of 6 was rejected: no tooltip, and the item silently absent from the stat totals. Only the missing tooltip was visible.
• Item tooltips on the search rows, at the rank the search actually chose.
• The gear planner is always the window that opens.
• A crafted item picked with chosen stats no longer showed "random stats".
• Locked slots reach the search's totals. Trinkets are always locked, so every search had been ignoring their stats.
• The skipped-items count says which kind of skipped, and the per-slot candidate cap is reported instead of silently truncating.
• Save-as-preset asks for a name and confirms an overwrite. It used to write to a fixed name, destroying the previous result and any preset that shared it.
• Russian covers every string the interface draws, enforced by a test.
• The stat table fits its window, its columns hold Russian labels, and the search panel no longer sits behind the action bars.

Known limitations

• Not a simulator. Printed stats only: no DPS, set bonuses, trinket procs, enchants or gems. Trinkets are picked by hand for that reason.
• Percentage buffs are approximated. A "+5% stats" aura scales with your gear, so the offset measured on your current set is slightly off for a different one. No API separates a base from a multiplier in one reading.
• The diminishing-returns curve models the three verified breakpoints and a flat penalty above the third. If the real curve has more brackets, the very top end is under-penalised.
• The search is a local search, not a proof. It can stop where no single swap helps though a combination would, and it says which happened.
• A new patch's items are invisible until the database is regenerated. GearTools compares build stamps at login and says so once.
• The SimulationCraft class keyword for Death Knight and Demon Hunter is still unconfirmed against Raidbots.