What a position on PancakeSwap would have earned you
Pick a pool, a range width and an amount. Every figure is a replay of swaps that happened, so this signs nothing, holds nothing, and cannot cost anybody anything.
Set up a position
Nothing here is interpolated. Each combination below was replayed with the same engine that produces the published bands, and one that was never replayed refuses instead of guessing.
How much each pool can actually take. A1 caps a replayed position at 1.0% of the depth it sits in, and depth grows with width — so the range you choose is also a capacity decision, not only a yield one.
| Pool | At the narrowest | At the widest |
|---|---|---|
| PancakeSwap v3 WBNB/USDT 0.05% | 2.1071 WBNB · ±40 | 41.3529 WBNB · ±800 |
| PancakeSwap v3 WBNB/USDT 0.25% | 0.0128 WBNB · ±40 | 0.2518 WBNB · ±800 |
Worth reading against what each pool pays per unit of liquidity: the thinner pool earns more on every unit and is the one that can take least, which is the whole shape of the opportunity and is invisible in a fee APR.
0.00199 WBNB net over 15 days
1.00 WBNB in a ±40-tick range on PancakeSwap v3 WBNB/USDT 0.05%, opened Jul 17, held through 140,647 swaps. That is 4.8% annualised — a rate, not a return anybody realized over 363 hours.
Nobody is managing this position. It is opened once at the width you picked and held to the end of the window — never recentred, never widened, never pulled. That is deliberate: it is what doing it yourself looks like, and it is the baseline an agent has to beat. What recentring adds, measured against this exact policy →
- fees earned+0.022
- adverse selection (upper bound)−0.020
The same window at 0.010 WBNB earned 5.0%. This position is 100× the size and earns 0.24pp less on every unit of it — a bigger position takes a smaller share of each swap’s fee, because the share is L / (L_pool + L) and it is in the denominator. Both figures are replayed; neither is the other one scaled.
Ticks -63,410 to -63,330 — the range the accountant used, not one redrawn here. The pool’s own depth over it was worth 210.71 WBNB. A1’s ceiling on it is 2.1071 WBNB. LPs keep 66.0% of every fee on this pool; the rest is the protocol’s and this position never saw it.
The window above is one of 20 at this width. Here is the range they span, so a good one is visibly a draw rather than a promise.
+/-80 ticks leads at the median, but its P25-P75 band overlaps +/-130 — not separated at this sample size
Where the price went, and where the range was. 83.5% of the sampled path sat outside the range, earning nothing — and nothing moved it back, because nothing here is managing it. That gap is what an agent is for →
Jul 17ticks -63,410 – -63,330, sampled from 240 points across the tapeAug 1
What this is, and what it is not
- Nothing is managing the position. It is minted once at the width you chose and held to the end of the window. On a marketplace that sells agents, that is worth saying plainly: these figures are the do-it-yourself arm, and the report measures an agent against exactly this policy.
- It is a replay, not a forecast. Every number is what a position of that size, at that width, opened on that date, would have collected from swaps already on chain. Nothing here predicts the next window.
- Adverse selection is subtracted, as an upper bound. Assumption A10 publishes the convexity cost as a ceiling on what arbitrage took, so the net figure is a floor on what the position kept. Every other venue quotes the gross number.
- The protocol’s cut is taken from each swap, not modelled. PancakeSwap’s protocol fee is on by default and differs between tiers of the same pair, so it is read per swap from the event’s own fields. Why that matters, and what it costs →
- Only badged pools appear. 3 of 3 pools carry a due-diligence badge, and 2 have enough tape to simulate at all. What each was checked for →
1,400 replayed windows across 7 widths, from ±40 to ±800 ticks. Which width the evidence can separate →