Misquote
agent cards loaded.

Agent marketplace · BNB Chain

Hire an agent to run your liquidity position.

Four agents, each replayed over 30 days of chain history, each quoted as a range with the losses shown. You grant a session key that expires and that you can revoke in one transaction. Every marketplace misquotes you; this one shows its arithmetic — every number traces to chain state or a published assumption, and where the evidence is thin it says nothing instead.

Find an agentWatch it run, no wallet

Or read the arithmetic first: does hiring one beat doing it yourself? · how a quote is made

Already on chain: a payment token escrowed on BSC mainnet, and reclaimed · a second agent that signs its own quotes

  • Wardenin range 93.9% · 20 moves · 0.019 WBNB
  • Gridin range 100.0% · 13 moves · 0.018 WBNB
  • Sentinelin range 87.0% · 1,191 moves · -0.042 WBNB

3 agents over one tape, on one scale — longest bar is the largest amount. In range is called against a 70% floor.

Router is not on this table: it holds no range of its own, so the columns here measure a position it never takes.

Rebalancing

Warden

COUNTERFACTUAL — this position was not heldchain tape
3 of 60 observations fall within 0.25% of the median. Every observation is above zero, which is off this axis.
The bands are separated — the gap survives the spread, not just the medians. +17.21pp at the median.

P25–P75 across 20 sub-windows of ~362.9h each, annualised. 60 of 60 observations finished in profit. On 1.00 WBNB of capital; a larger position earns a lower rate →Why that differs from the 725.7h of tape →Tape runs 17 Jul 2026 to 16 Aug 2026.

Worst window 15.00%, best 39.99% · 24 of 60 observations are distinct results

In range: PASS (94% of 522520)Beats holding: PASS (100% of 60)Beats doing it yourself: 60 of 60

In range: threshold 70% · beats holding: threshold 50%

vs doing it yourself: +17.21pp — agent beats DIY by 17.21pp, bands do not overlap

  • fees earned+0.058
  • adverse selection (upper bound)−0.023
  • costs−0.016
net — bars scaled against 0.058 WBNB, the largest component0.019 WBNB
in range93.9%
decisions522,520 over 725.7h
moves1 mint · 19 recentre · 0 pull

To hire: an escrow you set, opening at 0.1 of the payment token · 0.000755 BNB in key and gas fees. What that pays for →

Hire WardenQuote my positionFull tearsheet →
Market making

Grid

COUNTERFACTUAL — this position was not heldchain tape
2 of 60 observations fall within 0.32% of the median. Every observation is above zero, which is off this axis.
The bands are separated — the gap survives the spread, not just the medians. +12.71pp at the median.

P25–P75 across 20 sub-windows of ~362.9h each, annualised. 60 of 60 observations finished in profit. On 1.00 WBNB of capital; a larger position earns a lower rate →Why that differs from the 725.7h of tape →Tape runs 17 Jul 2026 to 16 Aug 2026.

Worst window 11.44%, best 43.52% · 60 of 60 observations are distinct results

In range: PASS (100% of 522520)Beats holding: PASS (100% of 60)Beats doing it yourself: 56 of 60

In range: threshold 70% · beats holding: threshold 50%

vs doing it yourself: +12.71pp — agent beats DIY by 12.71pp, bands do not overlap

  • fees earned+0.048
  • adverse selection (upper bound)−0.020
  • costs−0.010
net — bars scaled against 0.048 WBNB, the largest component0.018 WBNB
in range100.0%
decisions522,520 over 725.7h
moves1 mint · 12 recentre · 0 pull

To hire: an escrow you set, opening at 0.1 of the payment token · 0.000755 BNB in key and gas fees. What that pays for →

Hire GridQuote my positionFull tearsheet →
Health

Sentinel

COUNTERFACTUAL — this position was not heldchain tape
5 of 60 observations fall within 0.19% of the median.
The bands are separated — the gap survives the spread, not just the medians. −67.71pp at the median.

P25–P75 across 20 sub-windows of ~362.9h each, annualised. 0 of 60 observations finished in profit. On 1.00 WBNB of capital; a larger position earns a lower rate →Why that differs from the 725.7h of tape →Tape runs 17 Jul 2026 to 16 Aug 2026.

Worst window -67.56%, best -48.65% · 60 of 60 observations are distinct results

In range: PASS (87% of 522520)Beats holding: FAIL (0% of 60)Beats doing it yourself: 0 of 60

In range: threshold 70% · beats holding: threshold 50%

vs doing it yourself: −67.71pp — agent loses to DIY by 67.71pp, bands do not overlap

  • fees earned+0.017
  • adverse selection (upper bound)−0.0048
  • costs−0.054
net — bars scaled against 0.054 WBNB, the largest component-0.042 WBNB
in range87.0%
decisions522,520 over 725.7h
moves596 mint · 0 recentre · 595 pull

To hire: an escrow you set, opening at 0.1 of the payment token · 0.000755 BNB in key and gas fees. What that pays for →

Hire SentinelQuote my positionFull tearsheet →
Yield · Venus Core Pool (BSC) dollar markets and 2 PancakeSwap v3 range(s) at +/-80, +/-400 ticks

Router

COUNTERFACTUAL — this position was not heldchain tape

net return on supplied capital (realized yield − switch costs), over 125h — too short to annualise honestly

31 of 60 observations fall within 0.00% of the median. Every observation is above zero, which is off this axis.
The bands overlap — at this sample size the two are not distinguishable, whatever the gap between their medians.

median 0.03% · 60 of 60 windows finished in profit

Best rate seen
2.98%
Round-trip hurdle
1.06%
Moves
1 enter · 1 switch · 0 exit
Decisions
251
  • vUSDT1,954,290.91 USD ceiling · held 78
  • vUSDC404,339.03 USD ceiling · held 171
  • PancakeSwap v3 WBNB/USDT 0.05% · ±80 ticks1,751.67 USD ceiling · never entered
  • PancakeSwap v3 WBNB/USDT 0.25% · ±400 ticks91.57 USD ceiling · never entered

vs doing it yourself: ±0.00pp — indistinguishable: -0.00pp is below the 0.1pp materiality floor

Router moved 1 time(s) over 250h. Best realized rate 2.98%; the widest venue gap reached 1.074pp against a 1.059pp hurdle — the edge cleared, and only just. It held the better venue on 68% of samples.

Router measured 2 PancakeSwap range(s) and entered none — size, not yield. A1 caps a position at 1% of its venue, so against 10,000 the ceilings are: PancakeSwap v3 WBNB/USDT 0.05% at +/-80 ticks tops out at 1,752; PancakeSwap v3 WBNB/USDT 0.25% at +/-400 ticks tops out at 92.

Hire Router

How it chose, and what that rests on →

What there is to check

8 pages, each generated from an artifact this repository can regenerate. The figures below are counted from those files, not typed.