choosing a rung in four steps
The procedure runs period, rung, availability, cost, and the last step turns on one number: the two fee bases cross at 23.1x taker and 69.4x maker.
Figures rebuilt 2026-08-27 by the MarketMoves editorial desk. Every number on this page is derived from the sources named in the methodology.
Four steps, in this order, because each one constrains the next. Reversing the first two is the common error: a multiplier chosen before a holding period is a number with no threshold behind it, and the modal rung falls 1.82 steps for every tenfold increase in the length of the hold, from 500x over one minute to 5x over one week.
step one, pick the holding period
The period sets the threshold. Each holding period has its own distribution of adverse excursion, and the rung is whatever clears the 95th percentile of that distribution. The modal rungs across the 5 periods measured here are 500x at one minute, 200x at five minutes, 20x at one hour, 20x at one session, 5x at one week, so the same market and the same conviction produce 4 different answers depending only on how long the position stays open. A period that is not decided in advance defaults in practice to the longest one, because an unclosed position keeps accumulating excursion.
step two, read the rung for that market and period
The table answers per market, not per class. The modal rung is a summary and the spread behind it is wide: at one hour the highest survivable rung across 46 markets was 200x and the lowest 10x, with the distribution running 10x on 1, 20x on 29, 50x on 9, 100x on 6, 200x on 1. The per-market rungs sit on the scalping page for one minute and five minutes, the day-trading page for one hour and one session, and the swing page for one week. Each row carries its own window count, so the sample behind the rung is visible next to it.
step three, check the venue offers that rung
A rung the data supports is useless if the venue does not list it. The flat-fee venue applies one ceiling of 1000x to every market it lists, across 5 asset classes, which means the ceiling never binds before the data does. The notional-fee venue sets a ceiling per asset, from 2x to 40x, and lists 18 of the 50 markets tracked here, while the flat-fee venue lists 50. On Bitcoin the ceilings are 1000x and 40x, so a rung above the second figure exists on one venue only. Instrument-level verification on the flat-fee venue covers 0 of 50 tracked markets, and the record for Bitcoin reads unverified, which makes this step a check a reader performs on the venue rather than a claim this site makes.
step four, price that rung on each venue
The fee basis decides which venue is cheaper, and the crossover is a single number. A fee charged on the wager does not move with the multiplier; a fee charged on notional scales linearly with it. Solving the two against each other gives 23.1x taker and 69.4x maker, so below the first figure the notional-fee venue opens for less and above it the flat-fee venue does. At 2x the cheaper opening cost belongs to Hyperliquid and at 40x it belongs to Moon, on the same $100 wager.
| Rung | Moon opens for | Hyperliquid opens for | Cheaper to open | Moon room | Hyperliquid room |
|---|---|---|---|---|---|
| 5x | $1.00 | $0.22 | Hyperliquid | 20.0% | 18.8% |
| 20x | $1.00 | $0.86 | Hyperliquid | 5.00% | 3.75% |
| 200x | $1.00 | not offered | Moon | 0.50% | not offered |
| 500x | $1.00 | not offered | Moon | 0.20% | not offered |
The referral discount on the notional-fee venue is worth 0.0018% of notional, which is why it moves the crossover very little and is stated here rather than advertised: fee schedule. Opening cost is also not the binding cost at the top of the ladder. On the flat-fee venue the opening fee is always 1.00 per cent of the distance to liquidation, at 2x and at 1000x alike, because both the fee and the room scale with the wager and the multiplier cancels.
worked example: Bitcoin over one session
Taking the four steps in order on one market. The period is one session. The rung for Bitcoin at that period is 50x on the flat-fee venue, whose room is 2.00 per cent against a 95th-percentile session excursion of 1.55 per cent measured across 511 windows, with the long side at 1.55 per cent and the short side at 1.23 per cent and the rung set by the worse of the two. The median session excursion on the same sample was 0.78 per cent and the 99th percentile 1.68 per cent, so the rung clears the ordinary session with room to spare and does not clear the worst window in a hundred.
Step three: both venues list Bitcoin, at 1000x and 40x respectively, so 50x is available on both. Step four: a $100 wager at 50x opens for $1.00 on the flat-fee venue and $2.16 on the notional-fee venue, which is the crossover doing its work, since 50 exceeds 23.1. Room differs as well as cost: 2.00 per cent against 0.75 per cent at the same rung, because one venue holds a maintenance buffer and the other does not.
The rung above, 100x, leaves 1.00 per cent of room against the same 1.55 per cent of session excursion, which is why the procedure stops at 50x. Its session rung on the notional-fee venue is 25x with 2.75 per cent of room, a different answer from the same excursion because the ladder and the maintenance rule differ, and neither figure includes carry.
Indicative example based on third-party market data. Moon's reference price, spread, fees, settlement and liquidation rules may produce a different result.
Step three exists because Moon documents asset classes and not instruments. 0 of 50 tracked markets are confirmed at instrument level and 50 are unverified, so availability is settled on the venue. Venue terms from registry 2026-08-27a.
Step four prices the opening cost only. Carry is excluded because one of the two rates is unpublished: the flat-fee venue charges every 8 hours at an undisclosed rate, and the other publishes hourly funding capped at 4.00% per hour: funding documentation.
Hub and tables: leverage by holding period. Working: how leverage works, what liquidation is, where the flat fee lands, the calculator. Stamp 2026-08-27 07:56 UTC UTC, registry 2026-08-27a.