Allocation

This section describes how the USP board allocates collateral between the whitelisted yield sources. These are classified according to some liquidity tiers

• T1: is a cash equivalent product, instantly redeemable (0-2 days) --> $w_s$

• T2: is a short-term yield product, redeemable in less than 1 week ($\leq$ 7 days) --> $w_7$

• T3: is a long-term yield product, redeemable in less than 1 month (8-30 days) --> $w_{30}$

The weights should satisfy the condition $w_s + w_{7} + w_{30} \;=\; 1.$

Allocation rules

1. Sort vaults by descending yield vs duration score $S_i$ (see Formulae)

2. Fill Tier 2 weight up to the global cap $c_7$. One-fourth of T2 allocations unlock every week and are constantly moved back to T1

3. Allocate the remainder to Tier 3 until the target length (12d) is reached

where $c_7$ limits the exposure to T2 products: too small allocation decreases the yield, but too large one risks draining liquidity if many users redeem within one week.

Formulae

Given some key parameters:

• $A$ = total collateral (USDC + USDS)

• $R_d$ = net USP redemptions on the day $d$

• $APR_{i, \tau_i}$= APR and epoch (days) of vault $i$

• $\sigma_w$ = daily standard deviation of redemptions (look-back $W \approx 90d$)

• $p$ = instant-service percentile (e.g. 97.5%)

---

To compute the instant-liquidity buffer (T1)

1. Statistical need --> $L = z_p\,\sigma_w\,\sqrt{H}$

   where $H =$ days promised instant liquidity (usually 1) and $z_p \approx 1.96 \; \text{for} \: p = 97.5\%$

2. Add operational cushion --> $\text{buffer}_{\min}=L+\kappa_A$ where $\kappa \approx 1\%$ of $A$ (oracle lag, gas spikes)

3. Target balance --> $B_{\text{sUSDS}}=\max \left(\text{buffer}_{\min},\,B_{\min}\right)$

4. Convert to portfolio weight $w_s^{*}=\dfrac{B_{\text{sUSDS}}}{A}$

If the current $w_s < w_s^*$ then move funds into T1, otherwise sweep the surplus out of the best-scoring vault.

To allocate the surplus, we compute a yield vs duration scoring for every vault $i$ using the formula

$S_i = \dfrac{\text{APR}_i - f_i}{1 + \lambda \tau_i}$

where $f_i$ is the annualized fee, and $\lambda \approx 0.04$ is a factor used to penalize long-term strategies.

T3 allocation should be larger than the target length ($\tau_{\text{target}} \approx 12$ days)

$\dfrac{\sum_i w_i \tau_i}{1 - w_s^{*}} \;\le\; \tau_{\text{target}}$

where $w_{i, \max}$ is an optional exposure limit to any single vault or protocol. It should be used only when two or more independent vaults exist.

A keeper routine should be set to rebalance funds either weekly or whenever $| (w_s - w_{s_\text{target}}) > \varepsilon |$

def rebalance():
    A = current_AUM()
    sigma_w = stdev(redemptions, window=W)
    L = z_p * sigma_w * sqrt(H)
    B_s = max(L + kappa * A, B_min)
    w_s_target = B_s / A

    scores = [
        (i, (apr[i] - fee[i]) / (1 + lambda_ * tau[i]))
        for i in vaults
    ]
    scores.sort(key=lambda x: x[1], reverse=True)

    alloc = {"sUSDS": w_s_target}
    remaining = 1 - w_s_target
    w7 = 0

    for i, _ in scores:
        if tau[i] <= 7 and w7 < c7:                 # fill T2
            x = min(c7 - w7, remaining)
            alloc[i] = x
            w7 += x
            remaining -= x
        elif remaining > 0:                         # fill T3
            alloc[i] = remaining
            remaining = 0
        if remaining <= 0:
            break

    execute_onchain_swaps(alloc)

Initial parameters

The USP board can modify the allocation parameters. The default values are:

Variable	Description	Default value
$p$	Instant-redeem service level	97–99%
$\kappa$	Operational cushion	1–2 % A
$\lambda$	Converts “epoch-day” into APR penalty	0.03–0.05
$\tau_{\text{target}}$	Max weighted epoch	10–14 days
$c_7$	Upper bound on the total weight of the 7-day vault sleeve	20–30 % A
$w_{i, \max}$	Optional per-vault cap. Use once the number of vaults is $\geq 3$	off / 30 % A