PoE2 Crafting Cost Variance: Average, Median and Unlucky 10%
Looking only at the average (expected) cost of a PoE2 craft is misleading. In five examples run on the ExileDesk engine, half the runs finished at about 70% of the average, and one in ten paid more than twice the average. A few unlucky runs pull the average up, so it sits above what most people actually experience.
Zundamon
If a post says "N div per item," does everyone craft it for that price?
Metan
Not at all. With the same method, one person finishes in 50 div and another needs 500.
Zundamon
That big a gap?! You'd never see it from the average alone.
Metan
So we took the engine's records of hundreds of runs and laid out the spread of luck.
Half the Runs Finish at About 70% of the Average
Here are the average, median and 90th percentile for five examples. The top three are examples where every run finished; the bottom two count finished runs only.
Zundamon
If the average is 220 div, I'll finish for about 220 div, right?
Metan
Look at the table. In every example, the middle run paid less than the average.
| Example | Runs | Average | Half the runs (median) | One in ten (90th percentile) | What the numbers are |
|---|---|---|---|---|---|
| Gold Ring, three suffixes (annul misses at once) | 300 | 220 div | 166 div | 451 div | Every run finished |
| ES gloves, three prefixes (Omen of Sinistral Exaltation + Greater Exalted Orb) | 300 | 105 div | 74 div | 250 div | Every run finished |
| Other gloves (cheapest route) | 1,000 | 94 div | 64 div | 223 div | Every run finished |
| ES gloves suffixes | 300 | 747 div | 631 div | 1,533 div | Finished runs only (looks cheap) |
| Mana ring | 300 | 865 div | 794 div | 1,634 div | Finished runs only (looks cheap) |
In every example, the middle run paid less than the average, and one in ten paid about twice the average.
Zundamon
The middle run is cheaper than the average!
Metan
In the top three, the middle run paid 0.68 to 0.75 of the average.
Metan
On the other hand, one in ten paid 2.0 to 2.4 times the average.
Zundamon
How much does the unluckiest person pay?
Metan
On the Gold Ring, one run in a hundred cost 980 div. That's 4.5 times the average.
A Few Unlucky Runs Push the Average Up
Crafting cost works like "keep going until it lands." So there is no ceiling for an unlucky run.
Zundamon
Why don't the middle run and the average line up?
Metan
A few runs miss over and over, lose every target and have to start again.
Metan
Those few runs drag the average up.
Zundamon
Most runs don't miss that much, so the middle sits below the average.
Metan
Exactly. On the Gold Ring, runs lost every target and restarted from Chaos Orbs 25 times on average.
Metan
More restarts make the cost balloon. Fewer restarts come in well under the average.
Zundamon
So how much should I budget?
Metan
If you hear "220 div on average" and bring 220 div, more than half of you will have some left over.
Metan
But if you want to be nearly sure to finish, plan for twice that or more.
With a Cap, Look at Cost per Item
One more thing to watch: the numbers you get when you set a cap (for example, giving up at 2,000 div per run).
Zundamon
What changes when there's a cap?
Metan
The number changes depending on whether the runs that never finished are counted.
| Number | What it counts | Mana ring | How it reads |
|---|---|---|---|
| Average of finished runs | Only runs that finished within the cap | 865 div | Leaves out runs that never finished, so it looks cheap |
| Cost per item | Everything spent by every run, divided by the number finished | 1,991 div | Includes runs that never finished (everyone's spending spread over the finished items) |
On the mana ring, 64% finished, and the cost per item was more than twice the finished-run average.
Zundamon
More than twice?!
Metan
When the finish rate is low, the finished-run average alone makes it look more than twice as cheap as it is.
Metan
That's why ExileDesk articles show both side by side.
Annulling Misses at Once Halves the Spread
The method changes the spread too. For the Gold Ring suffixes, we compared holding misses until three are filled (B) with annulling them at once (C).
Zundamon
Saving misses and annulling them together sounds like less hassle.
Metan
It's the other way around. Look at the table.
| Method | Average | Median | 90th percentile | Which is better |
|---|---|---|---|---|
| B fill to three | 394 div | 272 div | 955 div | Worse (Annulments hit targets more often) |
| C annul at once | 220 div | 166 div | 451 div | Better (cheaper on average and for the unlucky) |
Annulling at once (C) cut not just the average but also the 90th percentile to less than half.
Zundamon
Even the unlucky runs dropped from 955 div to 451 div!
Metan
Fewer Annulments hitting targets means fewer "lost everything, start again" runs.
Metan
The details are in Don't Hoard Your Misses.
Not yet verified (2)
- Every number comes from the ExileDesk engine. None were compared with Craft of Exile or in-game results.
- Prices are from the ExileDesk server market (2026-10-20 to 21). If the market moves, the costs change too.
Look at "half the runs" and "one in ten" alongside the average, and you won't misread crafting costs. When there's a cap, put the cost per item next to them too.
Where the Numbers Come From
Show sources (3)
- Average, median, 90th percentile, one in a hundred, number of restarts, and methods B and C for the five examples: ExileDesk engine. Run counts as in the tables. Capped at 2,000 div per run (only the other gloves used a different setup: 1,000 runs with a cap on the number of crafts instead of on div)
- Mana ring 64% finished, 1,991 div per item: ExileDesk engine, 2,000 div cap
- Market: ExileDesk server market (2026-10-20 to 21)
Try It Yourself
Apply the same moves a few times on a Gold Ring on the crafting bench and you'll feel the spread for yourself.
Related posts
VOICEVOX: Zundamon / Shikoku Metan
Character art: Shikoku Metan / Zundamon standing art by Sakamoto Ahiru