Could SLP Have Been Saved?

August 8, 2026 · A research synthesis · ~50 sources, four research passes, cross-checked against this repo's on-chain data

The question, and the short answer: Axie's SLP token lost 99.88% of its value and never recovered. This document synthesizes every serious published analysis — pre-collapse warnings, real-time crisis coverage, retrospectives, academic work, the mechanism-design literature, comparative cases, and Sky Mavis's own financial record — to answer one question: was it saveable? The evidence-weighted answer is uncomfortable: it was delayable, but not saveable as designed. Every intervention that was available in 2021 shifts the collapse's timing; the only mechanisms that remove the collapse also remove the thing that made SLP famous — the uncapped, cash-out-able "earn." Every game economy that has survived decades refuses exactly that property. But you should not take my word for it: the workbench lets you run every rescue in this document yourself, with the funding constraints attached.

Three schools exist in the literature, and you'll want to know which one you land in by the end:

Who knew what, when

The most striking fact in the corpus is chronological: this was one of the best-predicted economic failures in gaming history. The chart below plots every major published position on the real SLP price curve at its publication date. Hover or tap the dots. Notice what was in print before the peak — and when the money arrived versus when the warnings did.

The warning timeline — sources plotted on the SLP price curve (log)

bull case published warning Sky Mavis admission retrospective
Hover a dot to read what was said, by whom, with a link.

Read the sequence: the burn design was praised as adequate in July 2021 (Outlier Ventures). The first "this is a Ponzi on top of a legit game" was published August 11, 2021 — before the DAU peak (The Gouverneur). a16z's $152M Series B landed October 6, at almost exactly the user peak. Naavik's full quantified teardown ("fundamentally unsustainable," mint/burn ~2.5 and rising, sink structurally decaying, with specific fixes) ran November 14. Sky Mavis publicly agreed January 4, 2022 ("this inflation is not sustainable") and by February 8 was writing "we risk total and permanent economic collapse." The diagnosis preceded the death by half a year; the money preceded the diagnosis by five weeks.

The measured record — play with the data

Everything below is the repo's merged dataset: daily price (CoinGecko + Yahoo deep history) and on-chain Ronin mint/burn/active-addresses (Dune, from the April 2021 migration). Drag across any chart to select a window and the stats update; the preset chips jump to the episodes that matter. Event markers are the dated interventions from events.yaml.

SLP price, USD

Minted vs. burned, M SLP/day (7d avg)

minted burned

Active SLP addresses, thousands (7d avg)

Drag on any chart to zoom into a window. Mint counts claims (batched up to 14 days), so single-day spikes are cash-out events, not earn-rate changes — the 1.2B day is Katana's launch.

The numbers that settle the published disputes

The published record disagrees on the key emission figures — Naavik estimated 100–200M SLP/day minted (Nov 2021), CoinDesk reported "over 250M/day" (Feb 2022), and Sky Mavis's own Season 20 percentages imply ~320M/day. The on-chain record settles it:

ClaimPublished figureMeasured (this repo, 30-day averages)
Daily mint, Nov 2021100–200M (Naavik) 277M/day — the canonical warning undercounted
Daily mint vs burn, pre-Season 20"~4x more minted than burned" (Sky Mavis) 250M vs 45M — 5.5:1, worse than admitted
Season 20 cut size~55–56% claimed 250M → 106M/day = −58% — as advertised
Peak USD value of daily emissionnot published anywhere $21.7M/day (30d avg, Nov 26, 2021)
Net emission turns structurally positive— May 12, 2021 — 11 days after the price peak
Price −90% from peak— Dec 10, 2021 — 2 months before Season 20, 3 before the hack

Why it died: the consensus mechanism

Across ~20 substantive post-mortems, a stable five-step account emerges — and each step is confirmed by the measured data above:

  1. Uncapped, activity-linked emission. SLP minted as a linear function of play-hours with no supply cap — 250–280M/day at peak against 45–100M burned. Only ~5% of players ever bred (Reason). Sky Mavis: creation grew "over 160x" in 12 months.
  2. A sink that self-destructs. Breeding — ~90% of protocol revenue and effectively the only sink — exists to make Axies for new players. 8.4M Axies by Oct 2021, no Axie burn mechanism: the sink structurally weakened as the economy grew. Machinations' Gheza: "a guaranteed means of creating inflation… a positive feedback loop."
  3. A seller class by design. 60–65% of players were scholars recruited on an income promise — rational pure sellers. The ETH Zürich on-chain thesis adds the twist the pundits missed: the top 1% of wallets captured two-thirds of all real-world earnings.
  4. Price floor = marginal labor cost. Friedman's equilibrium argument: a grindable currency's price must fall to the labor cost of the world's cheapest grinder. Scholar earnings crossed below the Philippine minimum wage (~$7/day) in November 2021 — precisely killing the acquisition message.
  5. Unbounded exit. "There was no bottleneck for how quickly people could leave" (Will Robinson). The flywheel ran in reverse: $0.36 → $0.01 in seven months. The hack, arriving in March 2022, was an accelerant on an economy already down ~95%.
"The 'earn' in Play-to-Earn turned out to be dependent on an endless supply of new capital, a reflexive system where new players provided exit liquidity for existing ones." — Jeff "Jihoz" Zirlin, Axie co-founder, 2026 — the founder's own final diagnosis

Could the revenue have funded a rescue?

The strongest "it was saveable" argument is financial: Axie generated real money — $1.3B in the year to Feb 2022, a $364M peak month (Aug 2021), a $152M Series B, a treasury holding billions in token value. Could that have funded stablecoin wages, or a buyback floor? The assembled record:

PeriodProtocol revenue / daySLP emission value / dayCoverage
Apr–Jun 2021$22K → $410K SOLID small (SLP $0.30+, DAU <600K) GAP≫1 briefly
Jul 2021~$6.3M SOLID $3–8M EST~1x
Aug 2021 (revenue peak)~$11.7M SOLID $5–15M EST~1x, the best it ever got
Nov 2021 (emission-value peak)~$5–7M EST $21.7M measured SOLID outflow 3–4x revenue
Feb 2022 (pre-cut)~$0.5–1M EST ~$2.5M SOLIDunderwater
2024 (for scale)~$11K ($4M/yr) SOLID net-deflationary—

Three facts kill the naive version of the rescue:

Two claims circulating in secondary coverage found no support: the "$1M SLP burn events" (no evidence; treat as false) and any published 2021 proposal for stablecoin-denominated scholar wages (genuine gap — nobody wrote it down at the time; the workbench runs the counterfactual nobody wrote).

Every distinct "it could have been saved by X" claim found in the wild, merged with the mechanism-design literature's assessment, ranked by evidence strength × counterfactual plausibility for mid-2021. Each row can be run live in the workbench.

#MechanismIn print byEvidenceVerdict for mid-2021 AxieTry it
1Sink the Axies (release/burn mechanics, soulbound competitive Axies) Castronova & Lehdonvirta 2014 (theory); Naavik Nov 2021 (Axie-specific designs) Medium-high — 20 yrs of EVE practice; tempered by OSRS causal study (sinks misfire) Best single lever, fully in Sky Mavis's control, never shipped in time. Attacks the population ratchet SLP cuts never touched. Sinks ×2–3
2Dynamic emission (mint ≤ k × trailing burn) Samani 2018; shipped in Helium pre-2021 Medium — prevents the 160x supply blowout mechanically; no demonstrated P2E save Genuinely available. Bounds supply, not demand; cutting wages mid-boom was politically brutal. Dynamic mint cap
3Cut earlier and harder Naavik Nov 2021 (implicitly); enacted Feb 2022 Measured: real cuts produced pops with 2-day half-lives; our model gives peak-timing +120 days of solvency The cheapest counterfactual: same policy, 6 months earlier, ~triple the economy — still not a save. Takeover day + faucet cut
4Revenue-funded buybacks/burns Sky Mavis's own Jan 2022 dev journal Works iff funded by real external revenue (framework consensus); Olympus/Klima/Terra map the failure branch; StepN's $122.5M saved the capped token, not the earn token The one honest demand-side lever they could demonstrably fund at peak. Deployed in 2024 at $60K scale. Buyback rate
5Onboarding gates / breeding throttles Practiced Sept–Oct 2021 (Delphi-advised fees, MMR floor) Low-medium; never modeled ex ante; internally conflicted (breeding is also the sink) Extends runway, trades peak for plateau. They pulled these levers — late and reactively. Onboarding cap
6Earnings vesting/lockups DeFi Kingdoms shipped it Aug 2021 Weak — DFK's locked JEWEL still fell >90%; overhang gets priced immediately Timing tool; regressive against the scholars who needed daily wages. Vesting
7Stablecoin-denominated rewards Nobody in 2021 (verified gap); industry standard by 2023–26 Removes the reflexive loop entirely (our model: no collapse); real balance-sheet criticism The right answer, unavailable in practice: the math only closed for one month, and it converts the studio into a wage payer. The 2026 games (USDC prize budgets on Ronin itself) are this lesson, learned. Stablecoin share
8Non-tradeable/expiring reward currency Lisa JY Tan (retrospective on Origins' Moonshards) Origins shipped it in 2022 and the supply stabilized Works — by quietly deleting "earn." Stabilization by euthanasia of the value proposition. ≈ stablecoin share at 100% with $0 wage
9One-token redesign / cap SLP Velocity literature: Ehrsam 2016, Buterin 2017, Samani 2017–18 High as diagnosis — the theory predicting this exact failure was 4–5 years old and widely read Required not building it this way in 2019. Not a 2021 option: the architecture was load-bearing for the scholarship economy. —
10Pre-launch simulation cadCAD open-sourced 2019; Machinations practicing commercially The loop was discoverable in silico before a single token shipped (Machinations ran the sims later) The cheapest counterfactual of all — "what happens when DAU growth hits zero" was a one-week 2021 exercise. This entire repo is that exercise, run late. The simulator is the demo

The comparative record

Whether SLP was saveable is also an empirical question: other economies faced the same physics. The record is stark — every economy that survived refuses the one property that defined SLP (an uncapped, cash-out-able reward currency at the center of the core loop), and the one economy that ran SLP's design with better instrumentation died on the same schedule.

CaseSurvival mechanismDid SLP have it?
StepN (GST), 2022 — the natural experiment Per-user energy caps, weeks-fast iteration, $122.5M quarterly profit, buyback-burns — and the earn token still died; value was triaged into the capped token (GMT) Partially — proving the interventions themselves are insufficient for an uncapped earn token
EVE Online (ISK), 20+ yrs Closed loop (no legal cash-out), destruction-as-sink at colossal scale, staff economists since 2007, monthly public faucet/sink reports No, on all four counts
WoW Token / OSRS Bonds One-way valve: real money flows in, never out; RMT demand converted into a currency sink No — SLP was the two-way valve these systems exist to prevent
Second Life (L$), ~L$250/USD for 20 yrs Issuer-operated exchange; supply managed daily against demand for stability, not appreciation No — free-floating CEX price; Axie's version (the Stability Fund) arrived Jan 2024, seeded $60K
Diablo 3 RMAH (closed 2014) Amputated the real-money layer without killing the game — "it undermines Diablo's core game play" No — for scholars, the money layer was the game; removing earn removed the players (2022 proved it)
Gaia Online (the control-group death) None: sold currency for cash in 2013 → 16,500% inflation → users never returned SLP's uncapped faucet ≈ a gold generator; like Gaia, trust never returned at scale
Axie itself, post-2022 Emission cuts of 56% → 100%, crafting burns, 44B hard cap, stability fund → deflationary 2023–25 Eventually — proving the tools worked mechanically, once "earning" was abolished. Stability achieved by ending the experiment.
Pixels, 2024–25 (the rerun) 1.03M DAU → 283K; response: −84% inflation cut, inflationary sub-token deleted Same playbook, same result: supply stabilized; users and price didn't return
2026 stablecoin wave (Rumble Arcade on Ronin, etc.) USDC prize budgets; token price removed as a systemic variable; operator carries the cost No — and it's arguably what SLP's actual use-case (income) required all along

Three verdicts

1. On the mechanism: structural, predicted, and admitted. The endogenous account wins on the data. Net emission went structurally positive 11 days after the price peak; the price was −90% before Season 20 and the hack; the founder's own 2026 words — "new players provided exit liquidity for existing ones" — are the warning cases published in August 2021, conceded. The exogenous-shock school fails the timeline test.
2. On the rescue: delayable, not saveable as designed. The best-evidenced 2021 package — peak-timed emission cuts + Axie sinks + mint-burn coupling + fee-funded burns, all of it in print or in Sky Mavis's own journals by January 2022 — extends the economy by months and saves real people real money (a −55% cut at the peak roughly triples player-days in our model). Nothing in that package survives growth saturation, because none of it creates demand that isn't a bet on more players. StepN is the controlled experiment: better data, faster hands, more money, same grave.
3. On the design that would have survived: it existed, and it wasn't SLP. Closed loops (EVE, WoW), issuer-managed pegs (Second Life), or operator-funded stable rewards (the 2026 standard, now shipping on Axie's own chain) all survive the same stress. Each requires either closing the cash-out door or moving price risk from token holders to the operator — i.e., renouncing the growth engine that made Axie a $3B company. That is why "could it have been saved?" has an honest two-word answer: saved for whom? The token was the marketing. The marketing was the liability.

Questions for your own verdict

The workbench attaches funding constraints to every lever in this document. Some challenges to structure your own conclusions — the notes box there exports your answers as your own document:

1. Find any lever combination that keeps token-denominated rewards and survives past day 500 with a positive treasury and no unfunded reward days. Does one exist? What did you have to give up?
2. Take over at day 71 (the peak) with the full Naavik package: cut to 45%, sinks ×2, mint cap on. How many player-days do you save vs. history? Now move the takeover to day 150, then 220. Where does the rescue stop mattering — and is that before or after Naavik published?
3. Run 100% stablecoin rewards from day 71 at $2/day. Watch the unfunded-days diagnostic. Now cut the wage until it funds cleanly. What wage is actually sustainable — and would anyone have shown up for it?
4. The onboarding cap trades peak height for plateau length. Is there a cap that produces more total player-days than the uncapped boom? What does that imply about how fast Axie "should" have grown?
5. Give the buyback lever 100% of revenue. Note that the collapse still comes — then explain, in your own notes, who your buybacks actually paid.

Source library

PRE-COLLAPSE (2021)

CRISIS ERA (2022)

RETROSPECTIVES & ACADEMIC

MECHANISM-DESIGN LITERATURE

COMPARATIVE CASES & FINANCIAL RECORD