How Token Snipers and Bundled Wallets Work — and How Anti-Sniper Fair Launches Fight Back

本文为原文前 6,000 字符的节选翻译，完整内容请查看原文。

If you’ve ever bought a new meme coin “early” and then found that a few wallets already held a big share of the supply, you’ve met snipers and bundlers. That outcome isn’t luck. It follows from how open launchpads work and how transactions get ordered on-chain. This post walks through the mechanics and then looks at what an anti-sniper fair launch does about them.

如果你曾经“趁早”买入一种新的模因币，却发现少数几个钱包已经持有大量供应，那么你已经遇到了狙击手（snipers）和捆绑者（bundlers）。这种结果并非运气使然，而是由开放式启动平台（launchpads）的运作方式以及链上交易的排序机制所决定的。本文将剖析其运作机制，并探讨“反狙击公平启动”如何应对这些问题。

The setup: a permissionless launch. Most open launchpads work the same way. Anyone can call a “create token” instruction, and the token starts trading right away against a bonding curve or a freshly seeded liquidity pool. The price rises as supply is bought. The earliest buyers therefore get the most tokens per unit of SOL or ETH, and the earliest buyers are usually whoever can see the creation and react fastest. That’s the core problem. When creation and trading open at the same moment, speed decides the starting distribution, and bots are faster than people.

启动设置：无需许可的发布。大多数开放式启动平台的运作方式如出一辙。任何人都可以调用“创建代币”指令，代币随即开始根据联合曲线（bonding curve）或新注入的流动性池进行交易。随着供应被买入，价格会随之上涨。因此，最早的买家能以单位 SOL 或 ETH 获得最多的代币，而这些最早的买家通常是那些能最先发现创建并做出反应的人。这就是核心问题所在。当创建和交易在同一时刻开启时，速度决定了初始分配，而机器人的速度远超人类。

How sniping works on Solana. Solana has no public mempool in the Ethereum sense. Snipers don’t usually see your transaction before it lands. Instead they: Stream on-chain events in real time. Bots subscribe to validator or RPC data streams (account and log subscriptions, or lower-latency feeds) and watch for the launchpad program’s “create” instruction. Fire a pre-built buy right away. The transaction is assembled ahead of time with only the new mint address left to fill in, then sent with a high priority fee so it lands in the same slot or the next one. Use bundles to land atomically. Through block-engine bundles (for example, Jito on Solana), a sender can submit an ordered group of transactions that all execute together, in order, or not at all, along with a tip for the block producer. The result: by the time a token appears in a human’s app feed, the first buys have already happened.

Solana 上的狙击运作方式。Solana 没有以太坊意义上的公共内存池（mempool）。狙击手通常无法在你的交易落地前看到它。相反，他们会：实时流式传输链上事件。机器人订阅验证者或 RPC 数据流（账户和日志订阅，或低延迟馈送），并监控启动程序发出的“创建”指令。立即触发预先构建的买入交易。交易被提前组装好，仅留出新的铸造地址待填，随后以高优先费发送，确保其落在同一区块槽或下一个区块槽中。使用捆绑包（bundles）实现原子化落地。通过区块引擎捆绑包（例如 Solana 上的 Jito），发送者可以提交一组有序的交易，这些交易要么按顺序全部执行，要么全部不执行，并附带给区块生产者的小费。结果是：当代币出现在人类用户的应用信息流中时，首批买入交易已经完成。

How sniping works on Base. Base is an OP Stack L2 run by a single sequencer, with no public mempool to front-run in the classic way. Sniping there usually means watching for the pool creation or the “trading enabled” event and landing a buy right after it, using priority fees and low-latency connections. Some token contracts also have owner-controlled trading switches. Bots watch for the transaction that flips that switch and try to be first in line once it’s on.

Base 上的狙击运作方式。Base 是由单一排序器运行的 OP Stack 二层网络（L2），没有可供传统方式抢先交易的公共内存池。在那里的狙击通常意味着监控流动性池的创建或“交易开启”事件，并利用优先费和低延迟连接在事件发生后立即买入。一些代币合约还设有由所有者控制的交易开关。机器人会监控触发该开关的交易，并试图在开关开启后第一时间排队买入。

Bundled wallets: the dev snipes their own launch. Sniping by outsiders is one problem. Bundling is often worse, because the creator is the one doing it. Here’s the pattern: The creator funds many fresh wallets ahead of time, often from one source or a short chain of hops. In a single bundle, they create the token and buy from all of those wallets in the same atomic execution. On a block explorer, the holder list looks spread out: lots of different wallets each holding a modest amount. In reality one party controls a large share of the supply from the very first slot. Later, those wallets can sell gradually, which is a slow “soft rug,” or all at once. Because the buys landed in the same bundle as the creation, no outside buyer could have gotten in before them.

捆绑钱包：开发者狙击自己的发布。外部人员的狙击是一个问题，但捆绑行为往往更糟糕，因为这是由创建者本人操作的。其模式如下：创建者提前为许多新钱包注资，资金通常来自同一个源头或短链跳转。在同一个捆绑包中，他们创建代币并利用所有这些钱包在同一次原子执行中进行买入。在区块浏览器上，持币列表看起来很分散：许多不同的钱包各持有少量代币。实际上，一方从第一个区块槽开始就控制了大部分供应。随后，这些钱包可以逐渐抛售（即缓慢的“软地毯”骗局），也可以一次性全部抛售。由于买入交易与创建交易在同一个捆绑包中落地，没有任何外部买家能在他们之前买入。

How to spot it yourself. Check the first slot/block. Open the token’s earliest transactions on an explorer (Solscan, Basescan). Several buys in the same slot as creation is a red flag. Trace funding. Click into early buyer wallets. If many were funded by the same address shortly before launch, treat them as one holder. Look for identical behavior. Similar buy sizes, wallets created the same day, and coordinated sells all suggest a single operator. Use clustering tools. Holder-map tools (Bubblemaps is a well-known example) show links between wallets that a plain holder list hides.

如何自行识别。检查第一个区块槽/区块。在浏览器（Solscan, Basescan）上打开代币的最早交易。如果在创建的同一区块槽内出现多次买入，这是一个危险信号。追踪资金来源。点击进入早期买家的钱包。如果许多钱包在发布前不久由同一个地址注资，应将其视为同一持有人。寻找相同的行为模式。相似的买入规模、同一天创建的钱包以及协同抛售，都暗示着单一操作者的存在。使用聚类工具。持币映射工具（Bubblemaps 是一个著名的例子）可以显示普通持币列表所隐藏的钱包之间的关联。

What an anti-sniper fair launch actually does. “Anti-sniper” isn’t one feature. It’s a set of design choices that cut down the speed advantage. Common approaches: 1. Separate creation from trading. If trading opens a known amount of time after creation, and everyone can see that ahead of time, there’s nothing to gain from detecting the creation first. The race doesn’t go away, but it moves to a moment everyone can prepare for. 2. Early-window per-wallet caps. For the first N slots or blocks, each wallet can buy only up to a set limit. This blunts single-wallet snipes. On its own, though, it pushes attackers toward splitting their buys across many wallets, which is why caps need the next item too. 3. Bundle and sniper detection. The launch logic or an off-chain screening layer can flag wallets that were funded from the same source, buys bundled with the creation transaction, or known sniper addresses. Flagged activity can be blocked, delayed, or limited. The trade-off is false positives, plus an ongoing back-and-forth as attackers adapt. 4. Batch or uniform-price opening. Instead of first-come-first-served, all orders placed in an opening window clear at the same price. When ordering inside the window doesn’t matter, priority fees and bundles stop paying off. 5. Creator restrictions. The creator’s own wallets can be barred from buying in the opening window, or required to put their allocation under vesting, so a dev can’t snipe their own launch. None of these is a cure-all. Determined attackers will try to work around each one, and a well-designed launch layers several together. Still, each of them targets a specific, known mechanism, which is very different from a “fair launch” label with nothing behind it.

反狙击公平启动的实际作用。“反狙击”并非单一功能，而是一系列旨在削弱速度优势的设计选择。常见方法包括：1. 将创建与交易分离。如果交易在创建后的一段已知时间内开启，且所有人都能提前获知，那么抢先发现创建行为就毫无意义。竞争并未消失，但转移到了每个人都能准备好的时刻。2. 早期窗口的单钱包限额。在最初的 N 个区块槽或区块内，每个钱包只能买入设定的限额。这削弱了单钱包狙击。但仅靠此举会促使攻击者将买入分散到多个钱包中，因此限额需要配合下一项措施。3. 捆绑包与狙击手检测。启动逻辑或链下筛选层可以标记由同一来源注资的钱包、与创建交易捆绑的买入，或已知的狙击手地址。被标记的活动可以被拦截、延迟或限制。其代价是误报，以及随着攻击者不断适应而产生的持续博弈。4. 批量或统一价格开启。不再采用先到先得原则，而是将开启窗口内的所有订单按同一价格成交。当窗口内的排序不再重要时，优先费和捆绑包就失去了作用。5. 创建者限制。禁止创建者自己的钱包在开启窗口内买入，或要求其分配额度进行锁定（vesting），从而防止开发者狙击自己的发布。这些措施都不是万能的。坚定的攻击者会试图绕过每一项，而设计良好的发布会叠加使用多种措施。尽管如此，每一项措施都针对特定的已知机制，这与毫无实质内容的“公平启动”标签有着本质区别。

Why this matters for normal traders. If a launch has no anti-sniper design, assume the first slot was contested and check the holder distribution before you trade. If a launch claims to be anti-sniper, verify it: look at the first-slot buyers and the funding clusters yourself. On-chain data doesn’t depend on marketing. This is the problem we’re working on at MemeSwap: fixing what’s broken with launchpads like pump.fun. Whatever platform you use, though, the checks above are the ones that count. Disclosure: I work on MemeSwap; its Safe Launchpad is coming soon and not live yet. This article is educational.

这对普通交易者为何重要。如果一个发布项目没有反狙击设计，请假设第一个区块槽已被争夺，并在交易前检查持币分布。如果一个项目声称是反狙击的，请自行验证：查看第一个区块槽的买家和资金聚类。链上数据不依赖于营销。这正是我们在 MemeSwap 致力于解决的问题：修复像 pump.fun 这类启动平台存在的问题。无论你使用什么平台，上述检查才是关键。披露：我在 MemeSwap 工作；其安全启动平台（Safe Launchpad）即将推出，目前尚未上线。本文仅供教育参考。