Endfield Gear Enhancement: 5★ Stats Rise 30%, Lower Gear Can't

Two hundred forty-three gear records sit in Endfield's equipment file. Inside them are 680 stat lines beyond the flat base value every piece carries, and those 680 lines split cleanly in two. Four hundred thirty-three of them record three enhanced values in sequence. The other 247 record nothing at all, and nothing can change them.

The split is not about the slot, the set or the wear level. It is about rarity, and it is absolute in the data: every single line on a five-star piece has enhancement values, and not one line on a one- to four-star piece does. This page lays out that split, what a full run of enhancement is worth on the lines that do scale (-5.1% to 30% of the starting number), and the two lines in the whole library that get worse as they climb.

What the three values actually are

Each modifier line is stored as a starting value plus an enhancedAttrValues array holding three more numbers. On a percentage line the pattern is a clean compounding 10%: the LYNX Gauntlets T1 third line opens at 0.205357 and runs 0.225893, 0.246429, 0.266964. On flat values the same 10% is applied and then rounded to a whole number, which is why a line starting at 65 ends at 84 rather than 86.5.

Two things are worth saying before the numbers get used for anything. The file never labels the three steps, so this page counts them by position, not by an in-game level number. And the flat base value that every piece carries (attribute id 3 on all 243 records) has no enhancement array at all. Whatever the game calls the process of raising a line, the data only describes it for the modifiers stacked on top of that base.

The split is rarity, and nothing else

Sorting the 680 lines by the rarity of the piece they sit on produces a table with no overlap. Five-star gear accounts for 433 lines and all 433 scale. Everything below five stars accounts for 247 lines and none of them do.

It also holds piece by piece. There is no record in the library where one line scales and another on the same item does not, so you never have to read an item sheet line by line to know whether it is worth taking further. The rarity on the card answers it. Everything below five stars lands at zero: Aburrey, Armored MSGR, Catastrophe, Mordvolt Insulation, Mordvolt Resistant, Roving MSGR and the two AIC starter sets, 61 pieces between them.

What a full enhancement is worth

Across the 431 scalable lines that rise at every step, the total gain from the starting value to the third recorded value lands between 27.78% and 30%. The fraction-valued lines land on exactly 30% almost every time (average 29.47% over 133 lines). The integer-valued lines average 28.92% because each step is rounded down, and that rounding is the only reason any line misses the 30% mark.

The ids are printed rather than names on purpose: the source data carries two conflicting id spaces for gear attributes, so any label this page invented would be a guess. What the ids do show is that the 28-30% band is not an average of wildly different curves. It is the same curve, written on flat numbers and on percentages.

Why you will see 8.33% and 9.09% instead of 10%

Take one step at a time and the rounding becomes visible. A line that starts at 21 gains two points per step, which is 9.52%, then 8.7%, then 8% as the two-point jump matters less against a larger base. A line that starts at 65 gains 6, 7 and 6 points, which reads as 9.23%, 9.86% and 7.69%.

Only 157 of the 1299 individual steps hit exactly 10%. The rest are the same rule seen through whole numbers. If you are estimating what a piece will look like after enhancement, work from the third recorded value rather than multiplying by 1.1 three times.

Every scalable line carries a pity rule

The correspondence is one-to-one, and it is the cleanest signal in the file: a line has enhancement values if and only if it also names an enhancement rule. All 433 scalable lines have one of three, and the 247 lines without enhancement values have none.

Double is the largest group at 258 lines and Extra covers 152, with 23 left on Single. The names line up with the official description of artificing as a stored-progress system where a capped bar guarantees the next attempt, which the gear sets and artificing guide quotes from the release notes. What the data adds is where those rules sit: not on the piece, not on the set, but on individual attribute lines.

Two lines in the library go backwards

Every scalable line except two rises on each step. The exceptions are both attribute id 0, both on five-star Lv.70 pieces, and both carry the Extra pity rule.

Losses of 5.06% and 4.89% are small, but they are losses, and they run against every other id-0 line in the file. Those open anywhere from 0.0885 to 74 and behave like the rest, gaining about 28.5%. Both outliers start at 0.82 and 0.83, and five other id-0 fraction lines start higher (the top is 0.96), so the likeliest reading is a cap applied in the source table rather than a property of those two pieces. Treat it as a data quirk to know about, not as advice to skip either item.

Which sets this leaves behind

Because rarity decides, the set table splits the same way. Fifteen suit ids come out at 100%, eight come out at zero, and one group (the 43 records with no suit id) sits at 26 of 104 lines, because some of those pieces are five-star and most are not.

That puts the two halves of the gear library in very different places. A five-star set is a set you can keep improving: MI Security alone holds 45 lines that all move. A set below five stars is a fixed object, and the only thing that changes about it is which piece you wear. If you are choosing what to spend assembly materials on, the rarity on the card is the whole decision.

What these numbers do not cover

No costs. The file records what a line becomes, not what an attempt consumes, so nothing here can tell you whether the third step is worth the material.

No success rates. The pity rules imply a progress bar somewhere, but the number of attempts each rule needs is not in the equipment data.

No in-game labels. The file stores ids, and the two id spaces behind gear attributes do not reconcile, so this page prints numbers. Names of stats would need a source the data does not have.

No set bonuses. Whatever a full set grants on top of its lines is not recorded in these records at all.

Frequently asked

Can four-star gear be enhanced? Not according to the data. All 247 lines on one- to four-star pieces carry no enhancement values, and no piece in the library mixes scalable and fixed lines.

How much does a full enhancement add? Between 27.78% and 30% of the line's starting value, with percentage lines landing on 30% and integer lines a little under it because of rounding. Two lines in the library lose about 5% instead.

Is the base value enhanced too? No. The flat base on all 243 records has no enhancement array, so only the modifier lines on top of it change.

Do the three steps ever hurt? Twice in the whole library, both on attribute id 0 and both by about 5%. Every other scalable line rises at every step.

Sources

Equipment records from the site's game-data mirror: 243 pieces, 680 modifier lines, 243 base values. Artificing wording from the official release preview (2026-01-17) and the 1.4 [Homecoming] notes (2026-07-16), as quoted in the gear sets and artificing guide.

Related reading: every five-star gear set, Frontiers and its 14 pieces, Qingbo and its five copies, AIC Fieldwork at Lv60, Type 50 Yinglung. Browse the full equipment database for any piece named here.