Where crypto returns actually come from (research, September 2026)

Question researched: what produces "quick gains" in crypto, and what does the data say about who captures them.

Short answer: nothing that is both quick and positive expected value is available to a retail participant trading on public information. Every strategy splits into one of two buckets:

Positive expected value, slow, boring: staking, stablecoin lending, market-neutral basis trades, skill-based earning (security research). Returns of roughly 3-12% per year, with real but bounded risks.
High variance, negative expected value for the median participant, fast: memecoins, leveraged perpetuals, sniping bots, most airdrop farming. The winners are a few percent of participants who have infrastructure, information, or capital that a home machine does not.

Anyone promising fast, guaranteed, or "AI-powered" returns is running a scam. That category is covered at the end.

Companion note: gaining-crypto-assets.md lists the developer bounties and hackathons that were actually open as of 2 September 2026, with eligibility and required resources, plus regulator fraud data.

Scorecard

Strategy	Typical return	Who actually wins	Speed	Verdict
Memecoin / launchpad trading	Median participant loses; 96% of Pump.fun wallets lost or made < $500 in March 2026	Bundlers, insiders, sub-50ms bots	Fast	Lottery ticket
Leveraged perps	65-80% of retail accounts net negative over 12 months	Market makers, disciplined minority (5-15%)	Fast	Negative EV without an edge
Sniping / MEV bots from a home machine	Most bot transactions fail to slippage, MEV, sandwiching	Co-located nodes, Jito bundles, private feeds	Fast	You are the liquidity
Airdrop farming	Zero after 6 months is a normal outcome; pros still profit	Operators with infra, capital, patience	Slow (months)	Not worth it for most
Volatile-pair LP	Only 28% of Uniswap LPs positive over 4 years; avg IL -3.8% per position	JIT liquidity (Wintermute, SCP, jaredfromsubway)	Slow	Negative EV passively
Stable-pair LP	Small positive, low single digits	Nearly everyone, barely	Slow	Fine, low yield
Basis / funding arbitrage	Mid-single to low-double digits net APR in calm markets; 20-40% in squeezes; ~0 or negative Feb-Jul 2026	Anyone with capital on a perp venue and risk controls	Slow	Best risk-adjusted "trading" yield
Staking ETH / SOL	ETH 2.2-3.8%; SOL 5-7% nominal	Everyone, proportional to capital	Slow	Baseline, low risk
Stablecoin lending	3-5% on Aave/Compound; spikes on utilization	Everyone	Slow	Baseline, low risk
Ethena sUSDe	4-12% variable, ~3.7% in early 2026	Holders while funding is positive	Slow	High risk vs cash-backed stables
Security research (bug bounties, audit contests)	Median payout $2k, median critical $20k, tail to $16M	Engineers who put in the hours	Weeks to first payout	The one skill-based path

1. Memecoins and launchpads

On-chain data from a Dune dashboard covering ~1.4M wallets that traded Pump.fun tokens in March 2026:

50.6% of wallets lost money.
45.4% made a profit under $500.
Combined: ~96% of participants lost or made under $500.
671,376 wallets lost under $500; 9,160 lost $1k-$10k.
Two wallets realized more than $1M. Two others lost $500k-$1M.

Token quality explains most of it. Solidus Labs found 98.6% of Pump.fun tokens collapsed to under $1,000 of liquidity; of 7M+ tokens with at least five trades, about 97,000 kept meaningful liquidity.

A study of 2,380 Solana tokens that reached a $250k market cap found 43.4% ended as rugs, 11.6% dumped, 19.2% went sideways, and 25.9% doubled at some point. Testing rule-based entry/exit strategies on public price data, the best result was +0.3% per trade on 195 tokens, which the authors called statistically inconclusive. Their conclusion: the real edges (sub-50ms RPC, Jito bundles, copy-trading known wallets, same-block sniping) exist but are unavailable from a normal machine on public data.

An estimated 70-80% of Pump.fun launches use bundled buys where the developer acquires 20-50% of supply in the deployment block via Jito bundles, then sells into retail. The average successful rug nets 50-200 SOL.

April 2026 dashboards showed profitable wallets rising to ~73%. Analysts attribute this to survivorship: unprofitable retail left, not to the game getting easier.

2. Leveraged perpetual futures

65-80% of retail perp accounts are net negative over rolling 12-month windows (Barber-Odean-style studies, CFTC/FCA retail disclosures, and exchange data).
Aggregated 2026 exchange data: 5-15% of retail traders net profitable over 12 months; futures-only traders at the low end (5-9%).
One survey: 84% of retail crypto traders lose money in their first year; 58% lose nearly all of it.
The Brazilian day-trader study: 1.1% of traders active 300+ days earned more than minimum wage.
October 2025: ~$19B of leveraged positions wiped in roughly a day. February 2026: $3.2B liquidated in 24 hours on a geopolitical headline. Traders with the right directional view still lost because leverage did not survive the path.
100x leverage is commonly available. A 1% move erases 100x margin; a 10% move erases 10x. 5-10% daily swings are routine.

Perps are a zero-sum game against market makers and funding. Without a measurable edge, the expected value after fees, funding, and slippage is negative, and the variance guarantees eventual ruin at high leverage.

3. Sniping, MEV, trading bots

The professional Solana memecoin subculture runs on low-latency Geyser feeds, Jito bundles, and co-located nodes with research budgets behind them. From a home machine on public RPC, the majority of bot transactions fail to slippage, MEV, and sandwiching. Any bot sold on Telegram or marketed with "AI" is either a scam or is selling you the privilege of being exit liquidity.

4. Airdrop farming

Sybil farms still capture ~40% of airdrop pools in 2026 (down from ~50% at the 2022-2023 peak), but detection now runs at >90% accuracy via wallet clustering, funding-source tracing, and behavioral ML. Entire farms get zeroed.
One entity claimed >60% of the aPriori airdrop with 14,000 wallets; the collateral damage was legitimate wallets flagged by aggressive clustering.
Consensus from farming communities: "for most, not worth it; for professionals with infrastructure, capital, and time, still profitable with high risk." Farming six months and receiving nothing is a normal outcome.
What still works: 2-5 wallets with 6+ months of organic history, real capital, real fees paid, on protocols you would use anyway (Hyperliquid, Polymarket, and Base are the commonly cited 2026 farms). Bear markets have historically paid farmers better (Arbitrum 2023, Hyperliquid HYPE) because fewer wallets split the pool.

This is a slow optionality play, not a quick gain.

5. Liquidity provision

Uniswap v3 ETH-USDC 5bps pool: LPs incurred cumulative markout losses exceeding $30M from August 2021 to August 2025.
Across LPs who provided $1 or more over four years, only 28% had positive cumulative PnL; ~10% exceeded $10; ~5% exceeded $100. Mean and median PnL are negative.
Top performers are JIT (just-in-time) providers who inject liquidity in the same block as a swap and remove it after: jaredfromsubway.eth, SCP, Wintermute.
Average realized impermanent loss is -3.8% per position versus holding; 49.5% of positions had negative returns; over half landed between -1% and +1%.
Narrow concentrated ranges amplify IL and go out of range in volatile markets; unmanaged positions end up one-sided in the depreciating asset.
Stable pairs (DAI/USDC, 3pool) show nearly all LPs slightly positive, at 3-7% APY.

Passive LP in volatile pairs is a negative-EV donation to arbitrageurs (LVR).

6. Basis and funding-rate arbitrage

The one trading strategy with a structural, non-directional source of return: long spot, short an equal notional of the perpetual, collect funding paid by longs.

Calm markets on BTC/ETH: mid-single to low-double-digit net APR after fees.
Squeeze regimes: 20-40% on majors, higher on mid-caps, typically lasting 1-4 weeks before arbitrageurs compress it.
February through July 2026: BTC funding was compressed or negative (Glassnode) as the market sold off from above $120k. The trade returned roughly nothing for half a year.
As of late August 2026: BTC funding ~0.01% per 8h, annualizing to high single digits. Market makers are back in the trade.
Fee math: a Bybit round trip costs ~0.22% of notional on the perp leg; held for 30 days over ~90 funding settlements, net APR sits close to gross.

Risks: funding flips negative (exit rule: two consecutive negative periods), short-leg liquidation on a gap up if under-collateralized, basis divergence during squeezes, and exchange counterparty risk (FTX). Requires capital held on a CEX or perp DEX.

7. Staking and lending (the baseline)

Live rates observed in early September 2026:

Venue	Asset	APY
Lido	ETH	2.22%
Rocket Pool	ETH	2.18%
ETH solo staking (32 ETH)	ETH	3.1-3.8% incl. MEV tips
Jito	SOL	5.06%
Marinade	SOL	6.05%
SOL native staking	SOL	5.6-7.0%
Aave v3	USDC	3.2-5.2% (short spikes to 12%)
Aave v3	USDT	3.37%
Compound v3	USDC	4.79%
Sky (Maker)	USDS	3.52% (SSR ~4.75%)
Morpho Blue	USDC	4.1-6.8%
Ethena	sUSDe	~3.7% early 2026, ~9.4% late April 2026

Notes:

SOL staking yield is partly inflation; the dollar return depends on SOL price.
Staking risk: slashing, validator downtime, liquid-staking smart-contract risk. Lending risk: smart-contract exploits, liquidity crunches, stablecoin depeg.
Ethena sUSDe is not a cash-backed stablecoin. Its yield is the basis trade in Section 6 wrapped in a token. Reserve fund ~~$61M against ~$5.6B supply (~~1.1% of TVL). No third-party attestation, no direct retail redemption, collateral concentrated on a few exchanges, brief 50-100 bps depegs have occurred. Rated HIGH risk relative to USDC/USDP by independent reviewers.

8. Skill-based earning: security research

This is the one path where being a good engineer is the edge, and payouts are in crypto or USDC.

Immunefi: $110M+ paid to date; 45,000+ researchers; 650+ programs. H1 2026: $13.45M across 837 reports, median $2,000, mean $16,000. Across 593 long-running programs, median confirmed critical is $20,000, mean $114,355.
Sherlock: hosts the largest single bounty ($16M, Usual). Stake-to-submit model ($250 USDC per report, refunded if valid), 52% hit rate on impactful submissions.
Cantina: Coinbase $5M bounty. Uniswap v4: $15.5M critical.
Code4rena wound down in May 2026; Immunefi absorbed its programs. Audit contests now run on Immunefi, Sherlock, Cantina, Spearbit.
Audit competitions are time-boxed with fixed pools (Base Azul upgrade: $250k, April-May 2026) and are explicitly open to all experience levels. Rewards are share-weighted: a solo critical earns ~12x the shares of a solo medium, so the incentive is depth over volume.

Realistic path: Foundry + Solidity or Anchor + Rust, work through public vulnerable-contract exercises, read past contest reports, enter contests for the feedback loop (bugs validated within 24h), then move to standing bounties. First payouts typically take weeks of focused work, not days. The distribution is fat-tailed, and the median is modest, but unlike every strategy above the expected value is positive and it improves with practice.

9. The "quick gains" scam funnel

Searching for fast crypto returns puts you in the target audience for the largest fraud category in finance.

Pig butchering / synthetic brokerages: contact via dating apps, LinkedIn, a "wrong number" text, or a Telegram group; weeks of rapport; a trading platform with fabricated profits; withdrawals blocked behind "tax" or "release" fees. In 2026 the platforms and the relationship management are AI-generated at industrial scale. Median victim makes 4-7 escalating payments over 4-8 weeks. Global losses exceeded $75B from 2020 to 2024; average payment size grew 253% year-over-year into 2025.
The SEC's NanoBit case (June 2026, $5.5M judgment): the platform never executed a single real trade.
MLM "trading" schemes promising to double funds in 40-45 days (FQL, 2026).
CertiK documented $370M in scam-related crypto losses in January 2026 alone.

Rules that filter nearly all of it: no one with a real edge sells signals or bots to strangers; any counterparty that controls your withdrawal is not an exchange; a displayed balance is not money until it is in a wallet you hold keys to; and "guaranteed" plus "crypto" in the same sentence is a scam.

10. Expected value at $10,000 for 12 months

Illustrative, using the ranges above:

Allocation	Expected outcome	Distribution
Stablecoin lending @ 4%	+$400	Tight; tail risk is protocol exploit
SOL staking @ 6%	+$600 in SOL	Dollar value tracks SOL price
Basis trade @ 8-12%	+$800 to +$1,200	Near zero in negative-funding regimes; tail risk is venue failure
Perps, 10x, no edge	Negative	~70% chance of net loss; meaningful chance of total loss
Memecoins	Negative	~50% chance of loss; ~96% chance of < $500 gain; ~0.0001% chance of $1M
Bug bounties (time, not capital)	Positive, skill-dependent	Median $2k per valid report; fat tail

The math for "quick" is simple: a strategy that could plausibly double $10k in a month has a payoff distribution where the median outcome is a large loss. Strategies with a positive median compound at 3-12% per year.

Sources

Memecoins and launchpads

Leveraged trading

Airdrops

Liquidity provision

Basis / funding arbitrage

Staking, lending, synthetic dollars

Security research

Scams