CIP-0050 — Pledge Leverage

CIP-0050 · Pledge Leverage-Based Staking Rewards · 2022, updated 2025 · Liesenfelt et al. · adds one parameter L · hard fork required · No-go as a standalone — coherent only paired with a fee-layer viability floor

official page · original PR #242 · 2025 update PR #1042

CIP-0050 attacks the diagnostic's central pledge pathology — 95.6 % of the pledge-bonus budget returns to reserve unclaimed (POL.O1.F3), 78 % of staked ADA sits in pools with pledge ratio below 1 % (POL.O2.F1), and pledged ADA earns 0.68 %/yr against 2.3 %/yr passive (POL.O2.F2). The proposal converts pledge from a 22 % yield nudge into a hard cap on reward-eligible stake: a pool collects rewards on at most $L \cdot p$, with $L = 100$ the recommended endpoint.

CIP-0050 is the sharpest pledge-as-signal lever in the candidate bundle, suboptimal as a standalone — coherent only paired with a fee-layer viability floor. The cap delivers two structural properties by algebra (zero pledge → zero reward; pool-splitting revenue-neutral), but at $L = 100$ it clips ~84 % of productive stake — including pools that produce blocks reliably and serve delegators well — while leaving the root cause of the broken pledge signal — the bonus function A(ν, π) in the SL-D1 reward formula — entirely unrepaired.

Three findings frame the verdict:

• The cap is mechanically sharp — both design properties hold as theorems, not predictions.
• The change is too radical: ~84 % of productive stake clipped at L = 100, including pools that produce blocks reliably and serve delegators well — and the root cause of the broken pledge signal (A(ν, π)) is left untouched.
• In today's regime, most of the pool pot returns to reserve unused: distribution efficiency falls from 44 % to ~8 %, and network-wide delegator yield collapses from ~2.27 % to ~0.44 % — an 80 % drop. The closing-incentive-gap pathology the diagnostic flags (POL.O1) gets dramatically worse, not better.

The instrument names the right target and lands cleanly on it — but lands on a population the V2 spec is supposed to protect, not discipline.

Table of Contents

1. What CIP-0050 proposes

CIP-0050 turns pledge into a hard cap on pool earnings. The new rule: a pool collects rewards on at most $L$ times its operator's own pledge, on top of the existing saturation cap.

Three consequences fall out immediately:

• A pool with zero pledge earns zero reward.
• Splitting one pool into many leaves the total reward cap unchanged.
• Pledge becomes a binding constraint on what the protocol pays out.

The instrument adds a single dimensionless parameter $L$ (the proposal targets $L = 100$). It needs a hard fork to install the new ledger variable; no pool re-registration. It is the sharpest pledge-as-signal proposal in the candidate bundle.

2. The problem it tries to fix

Today, pledge barely affects what an operator earns. The mainnet evidence is clear:

• 78 % of staked ADA sits in pools with pledge ratio under 1 %.
• 42 of the 48 largest multi-pool operators forfeit the pledge bonus entirely.
• Pledged ADA yields ~0.68 %/yr; the same ADA placed as passive delegation yields ~2.3 %/yr.

Operators have rationally chosen to ignore the pledge bonus — it costs them more than it pays. The reward formula prices pledge as a soft 22 % nudge, and the operator population treats it as cosmetic.

CIP-0050 converts that nudge into a constraint: no pledge, no reward.

Sybil resistance — what is actually being defended.

A Sybil attack is when one person pretends to be many. In Cardano staking, that means a single operator running many separate pools to capture more rewards than they would running just one — even though the protocol's design assumes those pools belong to different operators.

Today, registering a pool costs about 500 ADA of refundable deposit. The protocol cannot tell on-chain whether two pools belong to the same person, unless that person declares it. This is how Cardano ends up with 449 productive pools but only 83 distinct entities behind them (POL.O5.F1) — the protocol target says ~500 slots, the reality says fewer than 100 hands.

CIP-0050's pledge cap is, at its core, an anti-Sybil instrument. The reasoning: if every pool needs its own pledge to earn rewards, then splitting one pool into N copies costs N times the pledge. Sybil becomes capital-bound, not just registration-bound.

This is the intuition the proposal's advocates build their case on. It is correct as far as it goes — and the diagnostic agrees with the mechanics. The disagreement that follows is about who on mainnet actually pledges under the new rule, and who ends up paying the cost regardless.

3. Verdict — three reasons it fails as a standalone

1. The cap is mechanically sharp on pledge-as-signal — zero pledge, zero reward.

A zero-pledge pool earns zero reward; a fleet split across $N$ pools earns the same total cap as one. Both properties hold by algebra, not by behavioural assumption — pledge becomes a binding constraint on the reward-eligible stake.

This is the sharpest pledge-as-signal expression in the candidate bundle, and the only one that makes pool-splitting strictly revenue-neutral at the pool level. → The cap as pledge-as-signal — full mechanism

2. The change is too radical — it removes V1 rewards from a large majority of currently-productive pools without fixing the root cause in the formula.

At $L = 100$ and the stake-weighted-median retail pledge ratio of 0.07 %, the cap binds at $0.07\sigma$ and pool reward drops to ~7 % of its V1 baseline; 78 % of staked ADA sits in pools below the 1 % compliance threshold. The custodial CEX / IVaaS segment — 21 % of productive stake — cannot self-pledge at all and collapses to zero reward by construction. ~84 % of productive stake — including pools that produce blocks reliably and serve delegators well — sees a material cut.

And the mechanism that produces today's broken pledge signal — the bonus function A(ν, π) inside the SL-D1 reward envelope — is left entirely untouched. The σ′ clip is a new gate before the formula runs; it does not repair what A does to the pledge signal once a pool is past the gate. → Too radical, root cause unfixed — full quantification

3. Most of the pool pot returns to reserve and network-wide delegator yield collapses.

Today, the diagnostic shows ~56 % of the pool pot already returns to reserve unused every epoch — the single largest addressable inefficiency in the system (POL.O1.F3: 95.6 % of the pledge-bonus budget already wasted). CIP-0050 at L = 100, applied to today's pledge distribution, clips eligible σ′ on ~84 % of productive stake — collapsing distribution efficiency from 44 % to ~8 % and pushing return-to-reserve to ~92 %. Network-wide stake-weighted delegator yield drops from ~2.27 % to ~0.44 % — an 80 % collapse.

The closing-incentive-gap pathology the diagnostic flags as the largest addressable inefficiency in the system gets dramatically worse, not better. → Pool pot returns to reserve — yield collapses

The remainder of the document walks the proposal in three steps: §4 quantifies what changes on mainnet today; Appendix A unpacks the formula, the three binding regimes, and the four operator response paths; Appendix B documents the per-finding evidence with verdict tags.

4. What it does to mainnet today

CIP-0050.4.1 — At the proposal's recommended $L = 100$, ~18 B ADA — 84 % of productive stake — sits in pools that would be clipped or collapsed. Red = clipped, green = unaffected, orange = mixed.

84 % of productive stake clipped or collapsed at the long-run target is the headline asymmetry.

The CIP's staged ramp ($L = 10\,000 \to 1\,000 \to 100$) is meant to soften this. The ramp bets that operators raise pledge between steps.

The diagnostic finds little evidence that operators do this. And the 2.04 B custodial-by-extraction segment cannot pledge at all — they hold custodied retail funds, not their own capital.

What the proposal's own forward-looking simulation shows.

The CIP-0050 advocates published a forward-looking simulation using the Edinburgh Reward-Sharing Simulation engine, projecting network behaviour at $k = 2\,000$. Two of their scenarios are directly comparable:

A one-entity improvement in their own model. The advocates frame this as "network pledge rises slightly under CIP-0050" — accurate. But the headline measure of decentralisation (the count of distinct entities holding the network) is essentially flat when L is the only thing that changes.

The larger improvement they cite (~116 entities at $a_0 = 0.1$ versus ~160 at $a_0 = 0.3$) comes from restoring a_0 to 0.3 — its current mainnet value. That part of the gain is already in place; it is not a CIP-0050 contribution.

Read together with the mainnet snapshot above (~84 % of productive stake clipped or collapsed at L = 100), the picture is consistent: the cap reshuffles money among the same handful of large operators that produce today's concentration. It does not break the concentration regime; it just changes who qualifies for the V1 reward.

5. Read more

• The proposal itself — CIP-0050 on cardano-cips.org
• How it fits the four-CIP bundle — Intro & Conclusion of the 4 CIPs
• Stake-cap layer comparison with CIP-0037 — Stake-Cap synthesis
• Mechanism in detail (formula, three scenarios, the cliff at the pledge-ratio threshold, operator decision tree, deployment ramp) — Appendix A
• Full findings list (S1 mechanical sharpness, S2 too radical / root cause unfixed, S3 entity-level gap) — Appendix B
• Origin, V2-milestone mapping, and diagnostic-finding anchors — Appendix C

Appendix A — Mechanism in detail

This appendix gives the full mechanical decomposition of CIP-0050: the formula, the binding regimes, three worked scenarios, and the operator/delegator response surface. The opener summarises the conclusions; this appendix carries the derivations and figures that back them.

A.1. The formula

CIP-0050 modifies the reward-eligible stake $\sigma'$ entering the SL-D1 reward function (full treatment in pools-distribution §2.3):

$$\sigma'^{(50)}_{i} = \min\!\left(\sigma_{i},\ \frac{1}{k},\ L \cdot p_{i}\right)$$

Symbols — inherited from the SL-D1 formula as simplified in the sub-flow. The RSS protocol normalises stake and pledge as fractions of circulating supply, not absolute ADA. Under that convention:

• $\sigma_i$ — pool $i$'s total stake (pledge + delegated), as a fraction of circulating supply.
• $p_i$ — pool $i$'s pledge, as a fraction of circulating supply.
• $k$ — target-pool count protocol parameter.
• $z_0 = 1/k$ — the saturation threshold as a fraction of circulating supply. At $k = 500$ and today's mainnet supply, $z_0 \cdot \text{Supply} \approx 77$ M ADA in absolute terms.
• $L$ — new leverage cap, dimensionless ($L \geq 1$).

The reward curve is expressed in two normalised coordinates (from pools-distribution §2.3):

• $\nu = \sigma / z_0$ — stake saturation level: what fraction of one fully-saturated V1 pool the total stake represents. $\nu = 1$ means the pool is at V1 saturation; $\nu > 1$ means oversaturated.
• $\pi = s / \sigma$ — within-pool pledge ratio: the fraction of the pool's stake the operator commits as their own. $\pi = 0$ means hollow pool; $\pi = 1$ means full self-pledge.

These are independent coordinates over $[0, 1] \times [0, 1]$. The pool reward is $\hat f'(\nu, \pi, \bar p) = \bar p \cdot P_{\max} \cdot E(\nu, \pi)$ where $P_{\max} = R/k$ (reward ceiling) and $E(\nu, \pi)$ is the envelope function (see pools-distribution §2.3).

CIP-0050 in normalised coordinates. Substituting $\sigma = \nu z_0$ and $p = s = \pi \nu z_0$ into the CIP formula and dividing by $z_0$:

$$\nu'^{(50)}_{i} \;=\; \min\!\left(\nu_{i},\ 1,\ L \cdot \pi_{i} \cdot \nu_{i}\right) \;=\; \nu_{i} \cdot \min\!\left(1,\ \tfrac{1}{\nu_{i}},\ L \cdot \pi_{i}\right)$$

For pools below or at V1 saturation ($\nu \leq 1$) the second term $1/\nu \geq 1$ is dominated, so the formula simplifies to $\nu' = \nu \cdot \min(1, L \cdot \pi)$. The reward becomes $\hat f' = \bar p \cdot P_{\max} \cdot E(\nu', \pi)$.

Reading the formula — the three binding cases:

For $L = 100$: the new cap binds whenever the pool's within-pool pledge ratio $\pi$ falls below 1 %. Above 1 %, CIP-0050 is invisible; at or below 1 %, the effective saturation level $\nu'$ is clipped to $L \cdot \pi \cdot \nu$ and the reward shrinks accordingly. The compliance condition reduces to a direct threshold on the pledge ratio.

Visualising the three caps — which one wins, for the median retail pool.

A.1 — Three competing caps for the median retail pool ($\sigma = 15$ M, $p = 10.5$ k): the new pledge-leverage cap $L \cdot p = 1.05$ M ADA wins the min, slicing the reward-eligible stake to 7 % of the pool's actual size.

Design surface.

A.2. Three worked scenarios

The three min-arguments ($\nu$, 1, $L\cdot\pi\cdot\nu$) — equivalently $(\sigma, 1/k, L\cdot s)$ in absolute units — carve the state space into three regimes. Each scenario below shows the four quantities $(\sigma, s, \nu, \pi)$ alongside the binding cap. $z_0 \approx 77$ M ADA at today's mainnet, $L = 100$.

Scenario A — Compliant small pool (within-pool pledge ratio $\pi \geq 1/L = 1\,\%$).

Scenario B — Zero-pledge mainnet-median pool (within-pool pledge ratio 0.07 %).

Scenario C — Saturated pool with thin pledge.

The core consequence. For a given $L$, the reform partitions the pool population by a single threshold on the within-pool pledge ratio $\pi$: at or above $1/L$, CIP-0050 is inactive; below it, the effective saturation level $\nu'$ is clipped to $L \cdot \pi \cdot \nu$ and the pool reward scales with the clip ratio $\nu'/\nu = L \cdot \pi$. Scenarios A and C both illustrate "cap does not bind" but for different reasons — A is compliant on pledge ratio; C is bounded by V1 saturation before the pledge cap can bite.

The three scenarios side by side.

A.2 — Reward fraction preserved across the three scenarios at $L = 100$: compliant (A) and saturated (C) pools keep 100 %; the zero-pledge median retail pool (B) drops to ~7 % — a binary line at $\pi = 1/L$.

A.3. How much pledge does an operator need

For any pool with stake $\sigma$, the minimum pledge to escape clipping at $L = 100$ is:

$$p_{\min} = \frac{\sigma}{L} = \frac{\sigma}{100}$$

Worked across the canonical nine-tier taxonomy that the diagnostic uses to bracket pools by stake size — running from Dormant (≈ 50 K ADA, too small to produce blocks reliably) through Sub-block, Sub-reliable, Healthy, Large healthy, Near-saturation, Saturated (~77 M ADA, at the V1 cap) to Oversaturated. Full definitions in pools-distribution §4.1.3; the seven middle tiers are the productive range used here:

At $L = 100$, a Healthy-tier pool needs **150 k ADA** (≈ \$37 k USD at \$0.25/ADA) to escape clipping. A Saturated-tier pool needs 770 k ADA (≈ \$192 k USD). At the tighter $L = 10$ the CIP considers, the pledge requirement is **10× higher** — a Healthy pool would need 1.5 M ADA (≈ \$375 k USD).

A.4. The cliff at the pledge-ratio threshold

How the σ' is computed — the three-way min.

A.3 — Decision flow for the three-way min computing $\sigma'$: each pool falls into one of three regimes — Clipped (pledge cap binds), Unaffected (actual stake binds), or V1 saturation ($1/k$ binds, CIP-0050 is a no-op).

Reward preserved vs pledge ratio — the cliff shape at different $L$ values.

A.4 — Reward fraction preserved as a function of within-pool pledge ratio $\pi$ for four leverage values: linear ramp up to the cliff at $\pi = 1/L$, flat at 100 % above. Median-pool marker at $\pi = 0.07\,\%$ falls in the clipped regime for every $L \geq 10$.

The stake-weighted-median retail pledge ratio is 0.07 %. That is marked on the X axis above and falls in the clipped regime at every $L \geq 10$. At the CIP's recommended $L = 100$, the median pool preserves only 7 % of its V1 reward.

The ramp seen from the median mainnet pool. Reading the same curve at the median pledge ratio (π = 0.07 %) as the CIP's staged $L$ tightens:

Table A.4 — Reward fraction preserved at the median retail pledge ratio at each step of the CIP's recommended ramp.

A.5. Structural properties (theorems, not predictions)

Four properties follow directly from the formula — independent of any behavioural assumption.

Table A.5 — The four structural properties of the CIP-0050 σ′ rule. The first two are the design's main strengths and are quantified at finding-level in Appendix B.1.

A.6. What this means for different operator types

A concrete example. Consider three mainnet operator archetypes (profiles inferred from operator-delegator §4.3.3):

The operator-level implication. Compliance under CIP-0050 at $L = 100$ requires the operator to lock 1 % of the pool's total stake as pledge. Three populations face very different economics:

• Custodial-by-extraction (CEX / IVaaS) — cannot self-pledge (funds are custodied retail balances, legally separate from operator capital). σ collapses to σ' = 0. Their only recourse is to attract pledge-holding partners or exit.
• Retail small-pool operators — the median pool needs hundreds of thousands of ADA in additional self-pledge. Most operators in this segment do not have this capital liquid. Their realistic options are: (a) shrink the pool by refusing delegation beyond $L \cdot p$, (b) accept the reward cut, (c) exit.
• Treasury / foundation pools — already compliant by construction. The reform is a no-op for this segment.

The reform's signal to the network is: "to operate a pool that captures delegation, you must prove 1 % of the pool's stake in personal capital". That signal is clean in principle — it is exactly the pledge-as-binding-commitment V2 §3.2 asks for — but it acts as a new gate before the reward formula runs, while leaving the formula itself (the A(ν, π) bonus function that produces today's non-pledging equilibrium) untouched (see Appendix B.2 — Too radical, root cause unfixed).

Minimum pledge to escape clipping, by tier.

A.6 — Minimum pledge needed per tier to escape clipping: a Healthy pool requires 150 k ADA at $L = 100$ and 1.5 M ADA at $L = 10$ — linear in pool stake and in $1/L$.

The pool-size sweep at median pledge. At the stake-weighted-median pledge ratio of 0.07 %, the fraction of reward preserved is the same for every pool size (7 %) — but the absolute stake clipped scales with σ:

The relative impact is uniform (7 % preserved everywhere at the same pledge ratio), but the absolute "stake in exile" grows linearly with the pool. A Saturated-tier pool with the median pledge ratio loses 71.6 M ₳ of reward-earning stake — the bulk of its delegation effectively becomes unrewarded overnight unless the operator raises pledge.

MPO fleet example — the revenue-neutrality of pool-splitting. Consider an operator with 1 M ₳ of pledge capital and a fleet of delegation-attracting pools.

A.7 — Fleet reward-eligible capacity vs pool count for an operator with 1 M ADA pledge: capacity plateaus at $L \cdot P =$ 100 M ADA once $L \cdot p < 1/k$, making pool-splitting revenue-neutral.

Above the threshold where $L \cdot p < 1/k$, fleet capacity = $L \cdot P$ constant. The MPO fleet-expansion incentive disappears at the pool level.

A.7. Operator decision tree

A.8 — Four response paths an operator faces under the new cap (accept the cut, raise pledge, shrink, exit), gated by the question "do you have $1/L$ of pool stake in liquid capital to lock?".

A zero-pledge operator facing the cap has four response paths. Using the Healthy-tier 15 M ₳ pool as the reference (10.5 k ₳ pledge = 0.07 % ratio):

Only options 1 and 4 are accessible to every operator. The diagnostic answers — for 78 % of stake, and for 42 of 48 saturation-scale MPOs — no, by revealed preference — to the central question "do you have 1/L of your pool's stake in liquid capital willing to lock it?".

A.8. The delegator perspective

A delegator holding 10 000 ₳ in a pool that gets clipped experiences the cut as a per-ADA ROS reduction. Using the three scenarios:

A delegator in Scenario B — a typical retail pool on mainnet today — would see their yield drop from ~227 ₳/yr to ~16 ₳/yr per 10 k ADA staked. On 10 k ADA over a year, that is the difference between \$56 vs \$4 at \$0.25/ADA.

A.9 — Annual yield on a 10 000 ADA stake across the three scenarios: compliant pools preserve ~227 ADA/yr; the zero-pledge median pool collapses to ~16 ADA/yr (a ~93 % cut).

At this level of compression, delegators have strong incentive to re-delegate — but per OPE.O7.F1, the observed delegator response to yield signals is not reliable, so whether the re-delegation actually occurs is the empirical open question (see Appendix B.3 — Pool pot returns to reserve, yield collapses).

A.9. The CIP's recommended deployment ramp

The CIP acknowledges that moving from today's regime (where pledge is cosmetic) to $L = 100$ is a shock and proposes a staged activation:

The ramp's implicit assumption: each step gives operators time to adjust their pledge upward before the next step. The reform rests on the assumption that this adjustment occurs — otherwise the system moves through successive regressive states. Whether the adjustment actually happens is the empirical question analysed in Appendix B.2.

A.10. Composition with a k-raise — when does a stake-cap actually deconcentrate?

The proposal's advocates argue L works in synergy with raising k. Their logic, walked through plainly:

• Today, raising k from 500 to a higher value just creates more pool slots that existing multi-pool operators absorb at near-zero marginal cost (the diagnostic confirms this — see cip-0082 §B.3);
• With L active, splitting a fixed pledge across more pools shrinks each pool's reward envelope. Existing fleets cannot absorb the new slots without raising real pledge.

The mechanical part is correct. A stake-cap layer is a genuine precondition for a constructive k-raise. Without it, every k-raise to date has produced multi-pool absorption (the August 2020 k: 150 → 500 raise produced today's MPO landscape).

What the advocates' argument does not address is the population that gets stranded by the cap before any k-raise happens. At L = 100 today, ~84 % of productive stake earns rewards through pools that fall outside the cap. Raising k on top of that does not help any of those pools — they were already clipped. So under (CIP-0050 alone) + (k-raise) on today's mainnet:

• the upper-tail concentration shifts toward operators with both real pledge and capacity for more pools — a small population in practice;
• the small-pool tail still bleeds out: both operator income and delegator yield fall;
• the custodial-by-extraction segment (~21 % of productive stake) still collapses, because the operators of those pools legally cannot self-pledge.

The deconcentration goal is reachable along this path only if a fee-layer viability instrument runs in parallel — something to keep the small-pool tail alive while the stake-cap pressures the upper tail and the k-raise opens new slots for new entrants. This is the fee-layer → stake-cap → k-recalibration sequencing the solution-evaluation conclusion names.

Without the fee-layer step, CIP-0050 + k-raise reaches a smaller productive population, not a more decentralised one.

Appendix B — Findings

Three cards below organise the analysis: what the cap actually delivers (S1), why it is too radical and leaves the root cause A(ν, π) unfixed (S2), and the macro-level damage to pool-pot distribution efficiency and network-wide delegator yield (S3).

Both properties hold by algebra, not by assumption about how operators will behave. That is what makes CIP-0050 the sharpest pledge-as-signal instrument in the bundle.

Reading the three findings together.

CIP-0050 is a radical reform. It removes V1 rewards from a large majority of currently-productive pools — including pools that produce blocks reliably and serve delegators well. The reform then bets that operators will respond by pledging up, restoring rewards through compliance.

The diagnostic shows the bet does not hold, because the formula bottleneck A(ν, π) that produces today's non-pledging equilibrium carries through unchanged. The σ′ clip is a new gate before the formula; it does not repair the gradient inside the formula. So the dominance relation that already convinced 78 % of stake to ignore pledge survives the reform.

A genuine V2 stake-cap reform must redesign A itself — smoother operator onset at low ν, no design preference for fully-private pools, explicit reward for the balanced-commitment regime. See the stake-cap layer synthesis for the three properties an A-redesign must deliver. The σ′ clip is at most a complement to that redesign, not a substitute.

Reading the three findings together.

CIP-0050 frames itself as a "soft cut-off". On today's mainnet — where the SPO landscape has already shown it cannot or will not pledge at compliance scale — it is anything but soft.

The dominance relation that produces today's non-pledging equilibrium is unchanged by the reform. So the cap acts as a binding clip on the majority of productive stake. Most of the pool pot returns to the reserve, and delegator yield collapses across the network. The closing-incentive-gap pathology the diagnostic names — the single largest addressable inefficiency in the system — gets dramatically worse, not better.

The fix is not at the cap. The fix is at the gradient: redesign A(ν, π) so pledge stops being a dominated strategy (stake-cap layer synthesis), and secure operator viability via the fee-layer first so the small-pool tail does not bleed out while the cap is biting. The "fee-layer first → stake-cap → k recalibration" sequencing the solution-evaluation conclusion names is the only order that does not move the productive landscape through a regressive intermediate state.

Appendix C — Origin and references

C.1. Identity card

C.2. Origin and context

Authorship and moment. Written in April 2022 by Michael Liesenfelt and co-authors, updated in May 2025. The authors' central observation: under $a_0 = 0.3$, the yield gap between zero-pledge and fully-pledged pools is ~22 % — "statistically unnoticed" by delegators and dominated by epoch-to-epoch variance. The CIP converts pledge from a nudge (a small yield advantage that delegators could notice) into a constraint (a hard cap on reward-eligible stake that makes pledge load-bearing).

Scope. CIP-0050 modifies the reward-distribution formula itself (the σ' clipping rule), not the fee-layer split. It is therefore a pools-distribution-layer instrument, not a fee-layer instrument. A pool with clipped σ' produces less reward in total; the split between operator and delegators remains whatever the fee-layer parameters (minPoolCost, poolRate) dictate.

Relation to other CIPs in the evaluation bundle. - Fee-layer CIPs (CIP-0023, CIP-0082) act on a different layer (post-per-pool-allocation split). CIP-0050 composes cleanly with either on the mechanical axis — they act on sequential stages of the reward pipeline — but the two sets target different V2 milestones. - CIP-0037 is the other stake-cap candidate in this evaluation. Same target (§3.2 pledge signal + §3.4 concentration) via a different primitive (smooth saturation curve vs hard cap). A coherent V2 package picks one stake-cap instrument; stacking both is not canonical. - k lever is the transversal parameter. CIP-0050 text argues $L$ converts a k-raise from a concentration risk into a decentralisation lever (stake-cap CIPs foreclose MPO fleet expansion; new slots therefore go to new operators rather than existing fleets). Standalone k-analysis: cip-0082 §B.3 standalone k-lever deep dive.

C.3. References

• Induced problems addressed (and not): μ01 — Closing the Consensus Incentive Gap (pledge paradox) — partially, via the σ′ clip ; μ02 — Guarantee operator viability — regressed (the cap deepens the viability gap, see §A.10) ; μ04 — SPO supply side — partially, via fragmentation cost.
• V2 Roadmap milestones potentially affected: V2 Roadmap M1 — Repair pledge, sustain the small SPO base (the four-move repair this CIP would interact with) ; V2 Roadmap §2.5 — Research axis on concentration (the entity-level work CIP-0050's pledge-leverage cap engages indirectly).
• Numerical baselines: Appendix A.3 of this file (pledge-ratio-to-clip mapping at L=100); pools-distribution §4.1.3 (nine-tier taxonomy); operator-delegator §4.3.3 (custodial / retail decomposition).
• Diagnostic findings cited in this file: POL.O2.F1, POL.O2.F2, POL.O5.F1, POL.O5.F2, POL.O5.F3, POL.O6.F2 (pools-distribution); OPE.O6.F4, OPE.O7.F1 (operator-delegator).
• Companion CIP: cip-0037.md — alternative stake-cap primitive (smooth curve vs hard cap). Same-layer pairing is not canonical.
• Fee-layer companions (§3.1 instruments): ../operator-delegator/cip-0023.md, ../operator-delegator/cip-0082.md. CIP-0050's precondition for constructive deployment is a fee-layer viability instrument active first.
• Transversal lever: cip-0082 §B.3 standalone k-lever deep dive — the standalone k-raise analysis; CIP-0050 is the stake-cap companion that converts a k-raise from a concentration risk into a decentralisation lever (CIP's own argument).
• Canonical source: CIP-0050 official page at cips.cardano.org/cip/CIP-0050; source PRs #242 and #1042.