Path of Exile 2’s support pool is 557 gems — 150 more than the 407 skill gems in the game. Of those, 530 carry an attribute colour (194 red, 183 blue, 153 green) and 190 sit in level band 1, so the first third of the pool is available almost immediately and the last 71 entries are the late-game tail.
That shape explains most support gem confusion. Players treat the pool as one flat list of modifiers and then wonder why a support does nothing on their skill. It is not a flat list: it is five level bands, three attribute colours and roughly two dozen keyword families, and a support only applies when its keywords overlap with your skill’s keywords.
The five level bands
| Band | Support gems | Share of pool | Cumulative |
|---|---|---|---|
| 1 | 190 | 34.1% | 190 |
| 2 | 107 | 19.2% | 297 |
| 3 | 109 | 19.6% | 406 |
| 4 | 80 | 14.4% | 486 |
| 5 | 71 | 12.7% | 557 |
By the end of band 4 you have access to 486 of 557 supports — 87.3% of the pool. Band 5 adds only 71 entries, and those are the ones that tend to be named, specific and build-defining: Minion Mastery sits in band 5, as do the rank-III versions of several support lines.
The split between band 2 and band 3 is almost exactly even — 107 against 109 — which is unusual for a progression table and worth knowing if you are deciding when to stop re-rolling a character. There is no quiet stretch where nothing new appears.
Attribute colour: red is the widest support pool
| Attribute | Support gems | Skill gems for comparison |
|---|---|---|
| Red (strength) | 194 | 31 |
| Blue (intelligence) | 183 | 107 |
| Green (dexterity) | 153 | 54 |
| No attribute recorded | 27 | 215 |
The comparison column is the interesting one. Strength has the smallest active skill pool in the game (31) and the largest support pool (194). Intelligence is the mirror image: 107 skill gems, and 183 supports to choose from but a far deeper set of things to attach them to.
Practically, a strength build is not short of supports — it is short of skills. That is why strength archetypes cluster around a handful of slams and strikes pushed hard, while intelligence builds spread across dozens of spells.
Tag families, and why they decide everything
| Tag | Support gems |
|---|---|
| Attack | 108 |
| Lineage | 76 |
| AoE | 66 |
| Projectile | 58 |
| Physical | 57 |
| Melee | 56 |
| Trigger | 54 |
| Duration | 51 |
| Spell | 47 |
| Fire | 40 |
| Minion | 38 |
| Persistent | 37 |
| Lightning | 33 |
| Buff | 30 |
| Chaos | 29 |
| Cold | 25 |
Attack at 108 is nearly double Spell at 47, and Melee at 56 is more than twice Cold at 25. Those ratios are the reason melee and attack builds feel support-rich while cold builds feel like they are choosing from a short list — they are.
Lineage at 76 is the odd entry. It is the third-largest tag in the pool and it does not describe a damage type or a delivery method, which means a player filtering by Attack or Projectile will never see those 76 gems at all. Whatever a Lineage support does, it is doing it outside the normal keyword path.
The clearest illustration of keyword gating is the support text itself:
“Meat Shield I — Supports skills which create Minions, granting them less damage taken but making them deal less damage. Cannot support skills which create undamageable Minions.” — poe2db.tw support gem entry, cached in
src/data/poe2-support-gems.json
Note the double restriction: the skill must create Minions, and it must create damageable Minions. A support can pass the keyword check and still be excluded by a clause. That is why reading the gem text beats reading the tag alone, and why 349 of the 557 entries carrying a description matters — for the other 208, the tag is all you have.
The socket budget is the real constraint
Support gems do not exist in a vacuum; they compete for sockets with the skill itself. The gemcutting table gives the breakpoints:
| Gem level | Character level | Sockets |
|---|---|---|
| 1 | 1 | 2 |
| 10 | 36 | 3 |
| 15 | 64 | 4 |
| 20 | 90 | 5 |
At gem level 1 you have two sockets, which means a skill plus one support. At gem level 10 — character level 36, roughly the end of the campaign — you have three. You do not see a four-socket group until character level 64, and five sockets until character level 90.
So the practical support question for most of the game is “which one support”, not “which five”. With 190 supports in band 1 to choose from and a single free socket, the decision is worth more thought than the size of the pool suggests.
Building a support plan that survives the bands
Start from the tag intersection, not the damage number. Filter by the two or three tags your main skill actually carries. A Lightning Arrow build is looking at Projectile (58), Lightning (33) and Chaining (6) supports — a pool of roughly 60 candidates before deduplication, not 557.
Keep one band-1 support per archetype in reserve. 190 entries at band 1 means a replacement is always available when you swap skills mid-campaign. Do not over-invest in a band-4 support you cannot yet use.
Do not assume two similar supports stack. Several support lines have rank I, II and III versions, and several different supports describe similar effects. Check the wording before spending the socket.
Respect the exclusion clauses. Meat Shield I is the model case: matching the Minion tag is necessary but not sufficient.
If you want the active side of the same system, the skill gem unlock guide covers all 407 skill gems by band and attribute, and the uncut gems guide explains the 20-level gem ladder those sockets come from. The full table is filterable on the support gems database, and the skill gems database holds the matching active pool.
For worked examples, the minion build guide uses all 38 Minion-tagged supports, and the crossbow build guide covers the 11 Ammunition supports. If you are still leveling, the beginner guide puts the band structure in campaign order.