Most perp traders open a position on three numbers: entry, size, leverage. You decide what to trade, you pick a clip size, you bump leverage until the margin fits, and you click confirm. The order ticket doesn't ask for anything else. The funding dashboard is one tab over. The liquidation price is one click in. The expected dollar loss if your stop fires is in your head — or it isn't anywhere at all.
The problem isn't that those three numbers are wrong. It's that they don't describe what you're actually exposed to. Two positions with the same notional and the same leverage can have completely different risk profiles depending on the venue's funding regime, the asset's realized vol, and where your stop is parked. The clip-size-and-leverage habit hides that asymmetry, and the order ticket is designed to keep hiding it.
We have watched enough Hyperliquid live runs — and read enough post-mortems on the ones that went sideways — to know that opening a position on four numbers instead of three is one of the highest-leverage process changes a perp trader can make. It is the difference between knowing what you put on and hoping you sized it right. The four numbers are not exotic. They are not signal-secrets. They are the basic accounting any reasonably designed strategy file should compute for you, before the order is sent.
This is what those four numbers are, how to compute them on a Hyperliquid perp with real values, and the shape of the strategy-file block that does the math automatically so you cannot forget any of them.
Number 1: R — the dollar amount you lose if the stop fires.
The first number, and the only one with a universal name in trading literature, is R: the dollar amount you lose if your stop hits exactly as planned.
R is not your notional. R is not your leverage. R is position_size × stop_distance × price, expressed in dollars, or equivalently as a percent of NAV. On a 2-ETH long entered at $3,800 with a stop at $3,760, R is 2 × $40 = $80. If your account is $4,000, that's 2.0% of NAV. That number — 2.0% — is the only sizing number that really matters once your stop is set. The notional ($7,600) and the leverage (1.9× on a $4,000 account) are downstream of R; they describe the exposure, not the loss budget.
The reason to compute R explicitly is that, on perps, sizing-by-leverage hides it. "I'm using 5× leverage" tells you almost nothing about your downside. A 5× position with a stop 0.5% from entry has R = 2.5% of NAV. A 5× position with a stop 4% from entry has R = 20% of NAV. Same leverage, eight times the loss budget. People who size by leverage make this mistake constantly, and the screenshot of the resulting trade looks identical to a more careful one until the bad day arrives.
The fix is to flip the formula. Start with the R you are willing to lose on this trade — most accounts that survive a year run R in the 0.5%–2% NAV band — and back into the position size from the stop distance. position_size = R_dollars / (stop_distance × price). The leverage that falls out is whatever the math says it is. If it comes out above the venue cap, or above your own comfort line, the right place to fix it is at the R input or the stop, not by quietly cutting leverage after the fact and pretending the trade is the same trade.
Number 2: Notional, and the leverage it implies.
Once R is fixed, your notional is determined. You should still write it down explicitly, because notional is the number that decides what the market impact at fill and the funding accrual at hold will cost you. R doesn't tell you either of those.
On Hyperliquid the maker/taker fees are 0.0144% / 0.045% as of writing (and a tier or two cheaper if you're earning the maker rebate), and a typical IOC fill on a top-three perp eats roughly another 2–6 bps to crossing the spread — more if your clip is larger than the top of book. Multiply those by your notional, not by R, to see your round-trip cost. On the $7,600 ETH long above, that's roughly 2 × $7,600 × 0.0007 ≈ $11 of friction on a clean entry-and-exit, or about 14% of the R you're betting. Most traders mentally write that off as "fees" without realizing it's eaten a seventh of their loss budget before the position has moved.
Leverage is the same number from a different angle: leverage = notional / NAV. Compute it even when you sized by R, for two reasons. The first is liquidation distance, which we get to next, and which is a function of leverage rather than R. The second is that the venue will refuse the order if leverage exceeds the per-asset cap (50× for majors, lower for long-tail assets on Hyperliquid), and you would rather see that arithmetic at the strategy layer than at the order layer.
Number 3: Liquidation distance, measured in daily-vol units.
The liquidation price is the price at which the venue force-closes your position because the margin has been eaten. On a Hyperliquid isolated long, it's roughly entry × (1 - 1/leverage + maintenance_margin). On the 1.9× ETH long above, liquidation is somewhere around $3,800 × 0.48 ≈ $1,820. Comfortable.
But comfortable by what measure? The dollar distance to liquidation is meaningless without a vol context. The right unit is daily realized vol. If ETH's 30-day realized vol is annualized 60%, the one-day standard deviation is 60% / √365 ≈ 3.14%, or about $119 at $3,800. Your $1,980 distance to liquidation is $1,980 / $119 ≈ 16.6σ of a one-day move. That's not a number you ever need to think about; the position will not be force-closed by ordinary price action.
The same arithmetic on a 20× position changes the picture entirely. Liquidation moves to roughly $3,620, your distance is $180, and that's about 1.5σ of a one-day move. Inside any normal week you should expect to see a 1.5σ move several times. The position will get force-closed on a Tuesday for ordinary reasons. The number that surfaces this risk is the vol-normalized distance, not the dollar distance — the same dollar distance looks fine on a quiet asset and reckless on an excitable one.
A rough live-trade heuristic, with all the usual caveats about heuristics: anything inside 3σ of one-day vol should be re-examined. Below 5σ is loaded — you are choosing to take a real chance of being closed by venue mechanics rather than by your own stop. Above 8σ you are mostly insulated from the routine vol distribution. The exact bands depend on how long you plan to hold and what regime you're in. The point isn't the bands; the point is that the question gets asked at all, in units that mean something.
Number 4: Funding cost or carry, per day.
The fourth number is the one almost nobody computes pre-trade, and the one that most quietly decides whether a slow-moving position is making or losing money while it sits.
On Hyperliquid, funding is paid hourly between longs and shorts. The published rate (say, +0.0125% per hour on ETH-PERP) is what the long side pays the short side every hour the position is open. Annualized, +0.0125% per hour is roughly 110%. Per day on a $7,600 long that's $7,600 × 0.0125% × 24 ≈ $22.80. If you hold the position for a week, you've spent ~$160 in funding alone — twice the R budget — purely to keep the trade on. The trade has to clear that bleed before it can call itself profitable.
The number to look at is funding_per_day_dollars, signed: positive if you're earning, negative if you're paying. Compute it from the funding rate prevailing when you click, not the seven-day average — the spot rate is what you start paying at the next hour boundary. If it's negative and large enough that your planned holding period would burn more than 0.3 × R, the trade is borrowing money from the loss budget to exist. That is sometimes the right thing to do — when your edge has a fast time-to-realize — and sometimes a tell that the trade is on the wrong side of the venue's structural pressure and the carry will keep paying for the other side to wait you out.
The flip side is also worth surfacing. If you're earning carry, write the daily dollar accrual down explicitly. It is, in some strategies, the entire P&L. The Funding Harvester strategy file we ship treats funding_per_day_dollars as a first-class signal output, not as flavor — you should not have to back into the carry from a funding screenshot in a tab.
How Engine writes the four numbers into a strategy file.
Inside Engine, the four-number check lives in the ## Risk block of the strategy file. The agent computes all four before any order is signed, and refuses the trade if any number is missing or outside the configured bounds. A small slice of the strategy file looks like this:
## Risk
# loss budget
R_max: 1.5% of NAV
# liquidation buffer
min_liq_distance: 4σ daily realized vol (30d window)
# funding bleed cap
max_funding_cost_per_day: 0.3 × R_max
# pre-trade emit (every field is required to sign)
emit:
- R_dollars
- notional, leverage
- liq_distance_sigma
- funding_per_day_dollars
When the agent considers a position, it doesn't write a "size 2.4% NAV" line into the decision log and call it done. It emits four lines, one per number, with the actual computed values for the contemplated order. If liq_distance_sigma comes back below 4, the trade is vetoed and the rule that vetoed it is named in the log. You can scroll back a week later and read why a position you wanted didn't go on. That property — the veto explains itself — is the whole reason the file is shaped this way.
The four numbers aren't a discipline pose. They're auditable. Anyone — you, a teammate, a skeptical reader of your published strategy on the marketplace — can read them, recompute them, and tell whether the trade was sane. Most trading-product interfaces ask for three numbers because three is what fits on the order ticket. Four is what fits on a strategy file you can actually inspect.
That's the asymmetry to lean on. The order ticket is for the venue. The strategy file is for you. The four numbers belong on the file.