Holder rewards: what works, and the structure we would build
We tested the idea that a stock pair or a "get paid for holding" mechanic makes a launch an obvious trade. Half of it holds up. Below: the evidence, a review of the first design we proposed, and the structure we now think is best.
Short answer
- A reward traders can see in their wallet is the strongest hook we found. 7 of 12 runners had one, and it hands callers their line: a running "paid to holders so far" dollar total.
- A stock pair on its own is not an edge. On Robinhood Chain, stock-paired launches ran less often than ETH-paired ones.
- The reward raises the odds of a small run, not a big one. Every reward token that touched $1M that week is now down 89% to 99%.
- Being first in a new payout asset is what won, not the mechanic. Copies of the winning asset did worst.
- The payout is marketing, not yield. It is huge at launch size and tiny at runner size. What held the survivors was a second leg (Binance listings), not the dividend.
- So the edge is a structure, not one mechanism: first-in-category payout asset, a sell-heavy tax that ends, a distributor that makes sellers pay holders, a live receipt page, and a second leg booked before launch. Details below.
Your hypothesis, tested
"There needs to be an interesting pair, maybe a stock pair, or some way users get paid by holding. Traders now look for direct rewards instead of empty promises... How can we do better, so it becomes an obvious trade?"
RightTraders want direct, visible rewards, not promises
The 2026 winners paid now, automatically, with no claim step: tokenized SpaceX to MARSCOIN holders, ZEC to ZCAT holders, dollars to X accounts for PAID. Deferred promises failed: pump.fun went 365 days without its promised airdrop, then switched to buybacks.
Right, and it is the main effectIt gives callers fuel
What callers posted were running dollar totals, never yields: "has already distributed 2,113 $SPCXB ($245,943) to holders" at 29 hours; "basically prints $zec"; "the QQQB dividends alone paid for a hotpot".
WrongA stock pair on its own is the edge
On Robinhood Chain, tokens launched against a tokenized stock reached $1M less often than tokens launched against ETH (1.03% vs 1.84%, 5,540 vs 4,736 tokens). Pure pairs with no payout gave the two worst collapses in our runner set: FATCOIN -98.5%, MICRODUCK -93%.
PartlyGetting paid by holding makes it an obvious trade
It makes the first week an obvious trade, not the token. MARSCOIN paid about 70% of its own market cap to holders on day 1. At runner size the same mechanic pays 0.005% to 0.3% a day against 5% to 16% daily price swings.
RightIt is crowded; we need to do better
Every launchpad now offers it (Flap, StonkFun, Four.meme Tax Mode, pump.fun Custom Pairs). The edge in each winner was being first in a new payout asset (SpaceX, QQQ, ZEC, dollars to X accounts) on the venue getting attention that month, plus a second leg that held the price after the yield faded.
Part A
What exists (2025 to Sep 2026)
On the venues that produced runners (Flap on BNB, StonkFun on Solana) the pair and the reward are one machine: the pool is quoted in an exotic asset, a trade tax is taken, sold into the pool, and the proceeds go to holders in the quote asset.
1. Pair against a tokenized stock, no payout
How: pool quoted in NVDA, LLY, SpaceX and so on; each buy forces a buy of the stock token. Free. Terminals print the stock next to the ticker, so the frame ("a meme that is also long SpaceX") sells itself.
What kills it: thin exotic pools, volume moves to WETH pools within a day, no second leg.
Best / worst: CASHCAT shows the pair is not even needed (WETH pair, $176.5M now). FATCOIN (LLY) $44.0M to $0.58M; MICRODUCK (NVDA) $50.6M to $3.5M.
2. Tax paid to holders in the quote asset (Flap, StonkFun, Four.meme Tax Mode)
How: 1% to 5% buy/sell tax, sold into the pool, proceeds pushed to holders above a minimum balance. Traders fund it; the platform takes a cut first.
What kills it: the tax is a permanent sell program (ZCAT's tax sold 61% of launch supply into its own pool in three weeks, RAYCAT 91%); the payout collapses with volume; on BNB the tax only binds on the main pool, so volume routes around it.
Best: MARSCOIN, 3% tax, about $4.7M of SPCXB paid, peak $265.5M, now $152.5M. 牛来, 1%, about $1.9M of QQQB paid, peak $163.3M, now $109.6M.
Worst: Stonks (a QQQB copy) $19.3M to $0.79M; RAYCAT $15.5M to $1.3M; KNOTS $49.8M to $7.1M.
3. Creator-fee and trading-fee sharing (pump.fun, Bags, Meteora, Believe, Flaunch, Clanker, Zora, UsePaid)
How: 0.05% to 1% of each swap goes to the creator, sometimes split onward. Pitch: "the creator earns forever, so keeps pushing".
What kills it: volume fades (Believe $419 a week against $32.7M all-time; Heaven and Flaunch $0 in 30 days), and payout rails a third party can switch off.
Best / worst: PAID went $0.04M to $52.1M in about 11 days, then fell 76% when X Money paused its payouts. LAUNCHCOIN (Believe) -99.98%. Zora volume -99.8% after Base ended creator rewards.
4. Buyback-and-burn from fees or revenue
How: a share of revenue buys the token and burns or reserves it. Pitch: "revenue goes into the chart".
What kills it: insiders sell into the buyback; for a single meme it is circular (it only buys when people already trade).
Best / worst: STONK buys about 0.28% of its market cap a day and is up 951% in 30 days, the best ratio in the set. PUMP bought $463.5M and is still -35% from its high; CLANKER -90%.
5. Staking, locks, points and airdrop farming
For memes, locks are a supply tool, not yield (BUN locked 71.4% at launch and stayed roughly flat). Points get farmed and dumped at the token event; no 2026 meme runner used them.
What held and what killed, across all families
| Held | Killed |
|---|---|
| Payout in an asset people want and had not been offered before (SpaceX, QQQ, ZEC) | Volume fading: StonkFun daily fees halved in a week |
| Fully on-chain, automatic, no claim step, authority revoked | A third party that can switch the payout off (PAID) |
| A second leg: Binance Wallet, Alpha, futures, spot | The tax acting as a sell program; volume routing around the taxed pool (MARSCOIN's untaxed pools now carry 95% of its DEX volume) |
| Snipers collecting the most reward, because they own the most supply at the lowest cost | |
| Platform risk: StonkFun sent at least $1.41M of reward money to its own wallets |
The one formula that explains everything
At a 3% tax, holders get 1% a day only if a third of the market cap trades every day through a taxed route. That happens at launch size and almost never after.
| Case | Paid per day, % of mcap | Typical daily move | Days of payout to cover one day's move |
|---|---|---|---|
| MARSCOIN, day 1 (~$0.35M mcap) | ~70% | n/a | under 1 |
| ZCAT, hot week | 0.53% | 16.4% | ~30 |
| ZCAT, today | 0.15% | 12.5 to 16.4% | 80 to 110 |
| MARSCOIN, today | 0.005 to 0.12% | 9.5 to 11.9% | 100 to 2,000 |
| STONK platform buyback | ~0.28% | 11.3% | ~40 |
Sources: contract reads, CoinGecko Pro, DexScreener, Stonk Board. A buyer of ZCAT at its peak has received about 4.5% of cost in ZEC while the token fell about 70% (estimate).
Reading: a holder reward is a launch-week marketing engine with an on-chain receipt, plus a tax that filters out flippers. It is not a price floor. No token in the set earned back its price move through rewards.
Part B
Does it work? The control group
The runner study only looked at winners. Here are two same-period launch universes where we can see who had the mechanism and who did not, and count how many ran.
BNB Chain, 8 to 10 Sep: Flap tax-and-dividend tokens vs plain launches
| Group | Tokens | Touched $1M | Closed $5M+ | Now |
|---|---|---|---|---|
| Dividend paid in a tokenized stock | 78 | 1 | 0 | GME20: $1.26M intraday, now $0.01M |
| Dividend paid in WBNB | 34 | 1 | 0 | うさぎ: $3.8M close, now $0.41M |
| Dividend paid in another token | 41 | 2 | 0 | VS, UPONLY: ~$1.8M, now under $0.1M |
| All tax + dividend | 153 | 4 (2.6%) | 0 | all four down 89 to 99% |
| Flap standard, no tax (same venue) | 506 | 0 | 0 | |
| four.meme (random 400) and other BNB (206) | 606 | 0 | 0 | median four.meme peak ~$4K |
| All plain launches | 1,154 | 0 | 0 |
6,468 tokens from our scanner log; tax vs no-tax read from each token's bytecode and taxRate() on-chain. Peaks are volume-filtered daily closes and highs, because raw highs were full of fake prints. Grade A.
Robinhood Chain, 4 Aug to 5 Sep: stock pair vs no stock pair
| Launch quote | Tokens | $1M+ | $20M+ | Organic $1M+ (opened under $1M) |
|---|---|---|---|---|
| Tokenized stock | 5,540 | 1.03% | 0.11% | 0.79% |
| ETH | 4,736 | 1.84% | 0.23% | 0.68% |
| USDG stablecoin | 1,200 | 3.17% | 0.67% | 1.75% |
| Another meme or platform token | 1,946 | 0.77% | 0.05% | 0.62% |
13,567-token census from 240,309 launches, quote asset resolved on-chain. Robinhood venues did not pay holders in the stock, so this tests pairing only. Grade A.
What the data supports
- An on-chain holder reward goes with more small runs than no reward (2.6% vs 0% to $1M, same venue and window; Fisher p about 0.003, estimate).
- A stock pair without a reward does not.
- Copying the current winner's payout asset is the weakest version.
What it does not support
- That rewards cause $20M runs. No group produced one that week. The chance for a reward token was under 1 in 150 (estimate). Our $20M reward runners were each first in their asset and each had an outside catalyst.
- Solana had no full control census. One 39-hour window is one market regime; a second window in October would test whether 2.6% holds.
Part C, review
Review of the first design we proposed
The research worker's pick was Streak Rewards: the token's fee stream pays holders weighted by how long they have held without selling (1x on day 1, 2x at day 3, 4x at day 7), paid by a keeper, published with Merkle proofs on a Live Receipt page. The receipt page is right. The streak mechanic has five problems.
1. It pays the wrong people most
Longest streak means earliest entry, and the earliest entries are snipers, bundlers and insiders who also hold the most supply at the lowest cost. The buyer who actually makes a run, the one coming in at $3M on day 4, gets a quarter of the weight of the sniper. The trade gets less obvious exactly when we need new buyers.
2. It adds a claim step
Merkle proofs mean users claim. Every 2026 winner pushed payouts into wallets automatically; the airdrop landing is the screenshot. A claim button is also the phishing pattern traders have learned to avoid.
3. The sell deterrent is invisible
Selling resets a multiplier on future payouts. Nobody feels that. People feel losing money they can see.
4. It reshuffles a shrinking pie
The core finding is that payout decays with volume. Streak weighting changes who gets the pie, not its size, so it does not fix the decay.
5. It treats the mechanism as the edge
The data says a mechanism alone gives under 1 in 150 odds of $20M. What held every survivor was a second leg. Our real asset for that is the market-making partner and the listing path, and the v1 design does not use it.
Part C, recommendation
Best assumed structure: the launch stack
No single mechanism gets a token to $20M. Our edge is a stack of five layers where each one covers the failure the data found in the layer before. The reward's job is to win week one and keep the token alive until the second leg lands.
First-in-category payout asset
Choose a payout asset no live token owns yet, on the venue getting attention that month. We track which assets are claimed and how the first token in each did, from the same scanner that produced the control group. This is the one choice that separated winners from the 78 copies, and it is research we already do.
Sell-heavy tax that ends
Low buy tax, high sell tax, stepping down to zero. Example (to test, not measured): buy 1% / sell 5% for 72 hours, then buy 0% / sell 3% to day 10, then 0% from day 14. Flap V3 already supports separate buy and sell rates and an expiry. The pitch: free to buy, costly to dump, and it ends. It removes the "permanent sell program" objection that hurt ZCAT and RAYCAT, and turns the tax end date into a scheduled event.
Keep-or-Forfeit distributor
Every holder accrues rewards pro rata to their balance, so a buyer on day 4 earns at the same rate as a buyer on day 1. Each accrual vests after 72 hours and is then pushed to the wallet automatically, no claim. Sell or move tokens out and your unvested rewards are forfeited back to the pot for everyone who stayed. Wallets that bought in the launch blocks or through bundles accrue into the pot instead of for themselves.
- Why it beats streaks: fair to late buyers; the sell deterrent is a dollar figure you can see ("you lose $340 if you sell now"); it hits snipers who dump; it keeps push-to-wallet.
- The caller line: "Sellers have handed holders $312K so far." Paper hands pay diamond hands, as a live counter.
- Side effect: the pending balance works as a soft lock on float in week one, without locking anything.
- Trust: in v1, split the tax: half through the launchpad's own dividend (trustless, the base everyone already recognises), half through our distributor from a published wallet with every flow on the receipt page. Move to 100% once the distributor is a verified contract. A big "marketing wallet" is a red flag to traders, so we never ask them to trust one blindly.
Live Receipt page
Paid so far in dollars, forfeited by sellers, largest payout, and a per-wallet lookup ("earned $X, pending $Y, vests in 2 days"). Shows whether tax and fee authority are revoked, plus one honest yield line. Auto-made share cards at milestones ($100K paid, 1,000 wallets paid). This is the screenshot every winner's callers posted, handed to them daily.
A second leg booked before launch
Plan the post-peak event before day one and time it for day 7 to 14, when the payout has decayed and the tax is stepping down: market-maker depth, a Binance Wallet / Alpha-style listing application in the category they are scanning, and the tax end date as its own moment. Every collapser in our set lacked this; every survivor had it. This is where the MM relationship is worth more than any contract.
What we left out, and why
| Idea | Decision |
|---|---|
| Stock pair with no payout | Out. The data says it does not help. |
| Backed token with a redemption floor | Later. The most obvious trade of all, but backing is small next to market cap early, it needs a custom vault, an audit and 5 to 8 weeks, and it looks most like a fund. |
| Milestone pot (part of the tax unlocks to holders at market-cap targets) | Later. A strong caller line ("$400K unlocks at $20M"), but it creates a price target people can manipulate. Test once the core stack has run once. |
| House token that earns from all our launches | Out. It makes the agency the issuer. |
| Points and airdrop farming | Out. The "empty promise" traders avoid. |
Honest limits of this structure
- Keep-or-Forfeit is our design and untested. We can backtest it before any client: replay MARSCOIN's and ZCAT's first week of holder history and compute what the forfeit pot and per-wallet pending balances would have been.
- Fewer sellers means less taxed volume, so less to pay. That is the intended trade: in week one a tighter float is worth more than a bigger payout.
- Anyone who moves tokens to an exchange or a new wallet forfeits. Accepted; the page says so up front.
- Solana payouts in a new token cost account rent for each new holder (about 0.002 SOL each, so roughly 10 SOL per 5,000 holders, estimate). BNB has no such cost, one reason to start there.
- Legal, one line: paying holders from fees can be read as a security or fund; counsel checks the payout design before the first client launch, without blocking the build.
Build order
| When (estimate) | What | Proves |
|---|---|---|
| Week 1 | Live Receipt page, read-only, running on MARSCOIN, 牛来 and ZCAT today. Category tracker (which payout assets are taken, how the first token did). | We can show a prospect a live sample in the readiness report with no client yet. |
| Week 1, parallel | Keep-or-Forfeit backtest on MARSCOIN and ZCAT holder history. | Whether the forfeit pot is big enough to be a story. If it is tiny, we drop it before building. |
| Weeks 2 to 3 | Distributor on BNB via Flap's marketing share: keeper swaps into the payout asset, pushes vested rewards, publishes every flow. | End-to-end on a small test token of our own before a client. |
| First client | Full stack: category pick, tax schedule, distributor, receipt page, second leg booked with the MM partner. | The real test. October control window runs alongside. |
Default: BNB first (Flap supports the tax shape, the control data is from there, no per-holder rent). The backtest decides whether Keep-or-Forfeit stays in the stack.