盘口上的博弈The Game on the Order Book

序0数据粒度，即策略边界Data granularity is the boundary of your strategy

同一个市场，不同的人看到的是完全不同的东西。看日线的人看到趋势，看分钟线的人看到波段，看盘口逐笔的人看到的是每一双手在做什么。你手里数据的粒度，直接圈定了你有资格玩哪一层的游戏——这就是为什么「自建连续盘口流」在草根量化史上是一个分水岭事件：它不是让人变聪明，而是让人看见了别人看不见的行为层。The same market shows completely different things to different people. Whoever watches daily bars sees trends, whoever watches minute bars sees swings, and whoever watches the order book tick by tick sees what every pair of hands is doing. The granularity of the data in your hands directly fences in which level of the game you are qualified to play—which is why “building your own continuous order-book feed” was a watershed event in the history of grassroots quant: it did not make anyone smarter, it made them able to see a layer of behavior nobody else could see.

注意阶梯里的结构：数据越细，信号越强，但保鲜期越短。日线信号能活几周，盘口失衡信号只能活几秒到几分钟。信号越短命，对速度和成本的要求越苛刻——这正是草根与机构分野的地方。所以这份教学的立场自始至终是：把微观结构当显微镜用（看清市场在发生什么，改进你的进出场），而不是当赛车用（拼谁快）。Notice the structure inside the ladder: the finer the data, the stronger the signal, but the shorter its shelf life. A daily-bar signal can live for weeks; an order-book imbalance signal lives only seconds to minutes. The shorter-lived the signal, the harsher its demands on speed and cost—and that is precisely where grassroots and institutions part ways. So the stance of this tutorial, from beginning to end, is this: use microstructure as a microscope (to see clearly what the market is doing, and to improve your entries and exits), not as a race car (to compete on who is faster).

壹I读懂盘口：市场的表情Reading the Order Book: The Market's Expression

盘口就是限价订单簿（order book）：所有还没成交的买单和卖单，按价格排好队，摆在桌面上。页首那个红绿相间的小框就是它的五档形态。读懂它只需要记住一个三分法——The order book is simply the limit order book (order book): every buy order and sell order that has not yet been filled, queued up by price and laid out on the table. That small red-and-green box at the top of the page is its five-level form. To read it you only need to remember one three-way split—

挂单是意图，成交是行动，撤单是变卦。普通行情软件只给你看意图的总量（五档数字）；逐笔数据让你同时看到三者的每一次变化。全部微观结构分析，本质上都是在这三种事件流里找规律。A resting order is intent, a trade is action, a cancel is a change of mind. Ordinary quote software only shows you the total of the intent (the five-level numbers); order-by-order data lets you see every single change in all three at once. All microstructure analysis is, at bottom, looking for patterns inside these three event streams.

一个先进的小概念：微观价格A slightly advanced little concept: the micro-price

中间价（买一卖一的平均）对谁都一样，但它浪费了深度信息。更聪明的估计是微观价格：把买卖一档的挂单量当作砝码，谁重，「真实价格」就被压向谁的对面——买盘厚，价格被顶向卖一；卖盘厚，价格被压向买一。The mid-price (the average of the best bid and the best ask) is the same for everyone, but it throws away the depth information. A smarter estimate is the micro-price: treat the resting volume at the best bid and the best ask as weights on a scale—whichever side is heavier pushes the “true price” toward the opposite side—a thick bid pushes the price up toward the best ask; a thick ask presses the price down toward the best bid.

这一个小公式就是全部盘口分析的雏形：订单簿的形状里，藏着下一步价格的倾向。第貳章把它展开成完整的信号族。This one small formula is the embryo of all order-book analysis: the shape of the order book hides the price's next-step tendency. Chapter II unfolds it into a complete family of signals.

貳II盘口失衡：桥往人多的一边压Order Book Imbalance: The Bridge Tilts Toward the Crowded Side

想象一座独木桥，两端各排着一队人要过桥。哪边队伍长，桥面就往哪边倾——不需要知道任何一个人为什么过桥，仅凭两队人数的不对称，就能猜出桥的受力方向。盘口失衡策略就是这件事的市场版：买卖两侧挂单量的不对称，是未来几秒到几分钟价格方向的统计性前兆。Imagine a single-plank bridge, with a queue of people waiting to cross at each end. Whichever queue is longer, the bridge tilts that way—you do not need to know why any single person is crossing; from the asymmetry between the two queues alone you can guess which way the bridge is being loaded. The order book imbalance strategy is the market version of exactly this: the asymmetry between the resting volume on the buy side and the sell side is a statistical precursor of the price direction over the next few seconds to few minutes.

三个可以动手算的信号Three signals you can compute yourself

① 队列失衡 QI —— 最简单的一个① Queue imbalance, QI — the simplest one

这不是玄学，是被反复验证过的统计规律：对「厚队列」的股票（价差恒等于一个最小报价单位、每档排队很长的品种，A 股里典型如低价大盘银行股），下一跳方向对 ρ 的依赖是单调的。原因也朴素——薄的那一侧先被吃穿。This is not mysticism, it is a statistical regularity that has been verified over and over: for “thick queue” stocks (names whose spread is permanently equal to one minimum tick and whose queue at each level is very long—in A-shares, typically the low-priced large-cap bank stocks), the dependence of the next tick's direction on ρ is monotonic. The reason is plain too—the thin side gets eaten through first.

② 订单流失衡 OFI —— 看变化而不是存量② Order flow imbalance, OFI — look at the change, not the stock

QI 看的是某一瞬间的「存量对比」，OFI 看的是一段时间里的「增减对比」。口语化定义：QI looks at the “comparison of stocks” at one instant; OFI looks at the “comparison of additions and subtractions” over a stretch of time. In colloquial terms:

直觉：价格不是被「成交量」推动的，而是被供需的净变化推动的。一笔大买单挂上来（需求增加）和一笔大卖单撤走（供给减少），对价格的意义是同号的——OFI 把它们记在同一本账上。这也是为什么必须有逐笔委托流才能算它：快照只给你存量，逐笔才给你每一次增减。The intuition: price is not driven by “volume traded”, it is driven by the net change in supply and demand. A large buy order being posted (demand increases) and a large sell order being cancelled (supply decreases) mean the same thing, with the same sign, as far as price is concerned—OFI books them in the same ledger. This is also why you must have the order-by-order submission stream to compute it: snapshots only give you the stock, only order-by-order data gives you every addition and subtraction.

③ 主动成交净额 —— 只信真金白银③ Net aggressive volume — trust only hard cash

挂单可以作秀，成交不能。当 QI 与主动成交净额同向共振时（买盘厚、且真的有人在持续扫货），信号最扎实；当二者背离（买盘看着厚、成交却持续砸向买一），要高度怀疑厚买盘是布景。Resting orders can put on a show, trades cannot. When QI and net aggressive volume resonate in the same direction (the bid is thick, and someone really is sweeping continuously), the signal is at its most solid; when the two diverge (the bid looks thick, yet trades keep slamming into the best bid), you should strongly suspect that the thick bid is stage scenery.

为什么它有效——三个来源Why it works—three sources

• 机械来源：薄的一侧先被消耗殆尽，价格自然向那一侧移动。哪怕全市场都是随机下单的傻子，这条也成立。Mechanical source: the thin side is used up first, so the price naturally moves toward that side. Even if the whole market were fools placing orders at random, this would still hold.
• 信息来源：知情资金进场会在订单流里留下脚印——持续的同向主动成交、耐心的分批挂单。失衡是脚印的统计像。Information source: informed money entering leaves footprints in the order flow—sustained aggressive trades in one direction, patient orders posted in slices. Imbalance is the statistical image of those footprints.
• 库存来源：做市性质的资金持仓偏了就会歪着报价来回归中性。读懂它们的歪法，等于读懂了短期的供需修复方向。Inventory source: money doing market making will skew its quotes to get back to neutral once its position has drifted. Reading how they skew is the same as reading the short-term direction in which supply and demand will be repaired.

大单行为学：真单、假单与诱单The behavior of large orders: real orders, fake orders and spoofing orders

逐笔数据最诱人的用途，是解剖「大单」。四条实战判别法：The most tempting use of order-by-order data is dissecting the “large order”. Four practical tests:

• 看它敢不敢挨打。真实防守单会留在原地承接成交；布景单在价格逼近时悄悄后移或整笔撤掉。逐笔流里「大单后移」是教科书级的撤退前兆。See whether it dares to take a hit. A genuine defensive order stays where it is and absorbs the trades; a scenery order quietly steps back, or is pulled in full, as the price approaches. In the order-by-order stream, “the large order stepping back” is a textbook precursor of retreat.
• 看撤单率。同一价位反复地挂了撤、撤了挂，成交寥寥——姿态大于诚意。挂单量大而撤单率低，才是真队列。Look at the cancel rate. Posted then cancelled, cancelled then posted again at the same price over and over, with barely any trades—posture over sincerity. Large resting volume together with a low cancel rate is what makes a real queue.
• 看位置。挂在关键心理位（整数关口、涨停价、昨日收盘价）的大单，表演成分的先验概率更高，因为观众最多的地方最值得演戏。Look at the location. A large order posted at a key psychological level (a round number, the limit-up price, yesterday's close) has a higher prior probability of being performance, because the place with the biggest audience is the place most worth acting in.
• 只用成交做终审。任何挂单叙事，最终以主动成交流为准。扫单 > 挂单，行动 > 姿态，永远如此。Let trades alone be the final judge. Any narrative built on resting orders is settled in the end by the aggressive trade flow. Sweeps > resting orders, action > posture, always.

冷水：成本账，以及为什么机构不在股票上玩这个Cold water: the cost arithmetic, and why institutions do not play this game in stocks

失衡信号的单次期望收益极小——通常只有千分之零点几（几个 bp），靠高频次累积。于是成本决定生死。A 股股票一次买卖往返的摩擦：佣金约万 1～万 3（双边）＋ 印花税万 5（卖出单边）＋ 过户费约万 0.1（双边），合计约 0.08%～0.12%。而信号单次期望往往只有 0.02%～0.05%。结论刺眼：在股票上做纯粹的高频失衡交易，数学上是给印花税打工。The expected return of an imbalance signal on any single occasion is tiny—usually only a few tenths of one thousandth (a few bp)—and it has to accumulate through sheer frequency. So costs decide life and death. The friction on one round trip in an A-share stock: commission of about 0.01%–0.03% (both sides) + stamp duty of 0.05% (sell side only) + transfer fee of about 0.001% (both sides), totalling roughly 0.08%–0.12%. And the expected value of a single signal is often only 0.02%–0.05%. The conclusion stings: doing pure high-frequency imbalance trading in stocks is, mathematically, working for the stamp duty.

所以机构的真实做法是绕开这笔税：去 T+0 且免印花税的品种（可转债、ETF、股指期货、期权）做，或者用「底仓＋日内回转」的方式在股票上变相 T+0，并谈到极低的佣金。数据和信号是一样的，战场是挑过的。So what institutions actually do is route around this tax: they do it in instruments that are T+0 and exempt from stamp duty (convertible bonds, ETFs, index futures, options), or they use a “core position + intraday round trip” approach to get de facto T+0 in stocks, and they negotiate commissions down to almost nothing. The data and the signals are the same; the battlefield has been picked.

草根的三条现实路径Three realistic paths for the grassroots trader

• 路径一：可转债。T+0、免印花税、最小变动 0.001 元、佣金可谈到十万分之几——散户「手炒盘口」的主战场几乎全在这里。同一套 QI/OFI 逻辑，换到摩擦低一个数量级的品种上，才有生存空间。（注意转债有 ±20% 涨跌幅与临停等规则，且波动剧烈。）Path one: convertible bonds. T+0, exempt from stamp duty, a minimum tick of 0.001 yuan, and commissions negotiable down to a few parts in a hundred thousand—almost the entire main battlefield of retail traders who “trade the order book by hand” is here. The very same QI/OFI logic only has room to survive once it is moved onto an instrument whose friction is an order of magnitude lower. (Note that convertible bonds have rules such as a ±20% price limit and intraday trading halts, and that they are violently volatile.)
• 路径二：底仓 T+0。长期持有一只熟悉的股票作底仓，日内用盘口信号高抛低吸做回转：先卖后买或先买后卖，收盘时仓位回到原样。把「方向判断」降级成「节奏改进」，风险敞口小得多。Path two: core-position T+0. Hold one stock you know well as a long-term core position, and use order book signals intraday to sell high and buy back low: sell first then buy, or buy first then sell, with the position back to where it started by the close. This demotes “calling the direction” into “improving the timing”, and the risk exposure is far smaller.
• 路径三：把失衡当「入场时机层」。你反正要买的票（基于任何慢逻辑：基本面、趋势、事件），下单前看一眼盘口：失衡顺风时用市价，失衡逆风时挂单等。每次省下几个 bp 的滑点，一年累积下来是真钱——这是微观结构对普通人性价比最高的用法，也是「用细数据做慢决策」的字面落地。Path three: treat imbalance as an “entry timing layer”. For a stock you were going to buy anyway (on any slow logic: fundamentals, trend, events), glance at the order book before you send the order: when the imbalance is a tailwind, use a market order; when it is a headwind, post a limit order and wait. Saving a few bp of slippage each time adds up to real money over a year—this is the best value for effort use of microstructure for an ordinary person, and it is the literal implementation of “using fine data to make slow decisions”.

參III动量与涨停：被制度整形的人性Momentum and the Limit-Up: Human Nature Reshaped by the Rules

动量的意思是「涨了的还会接着涨一会儿」。它在全世界都存在，但在 A 股格外规整，原因是三样制度把散户行为「整了形」：个人投资者长期贡献大约六成上下的成交（口径不一，量级如此）、T+1（今天买的明天才能卖）、涨跌停板（主板 ±10%，创业板/科创板 ±20%，北交所 ±30%，ST ±5%）。行为金融里所有的追涨、锚定、羊群，在这三面墙里被压出了可辨认的形状。Momentum means “what has already gone up will keep going up for a while.” It exists everywhere in the world, but in A-shares it is unusually regular, because three institutional rules have “reshaped” retail behavior: individual investors have long contributed roughly sixty percent of turnover (definitions vary; the order of magnitude is right), T+1 (what you buy today can only be sold tomorrow), and price limits (main board ±10%, ChiNext/STAR Market ±20%, Beijing Stock Exchange ±30%, ST stocks ±5%). All the chasing, anchoring and herding of behavioral finance get pressed into recognizable shapes inside these three walls.

日内动量的三种典型形态Three typical shapes of intraday momentum

• 竞价定调。集合竞价高开且放量的强势股，开盘后延续概率显著更高——隔夜积压的买入需求（昨天想买没买到的人）在开盘集中释放。竞价是全天的第一份体检报告。The auction sets the tone.A strong stock that opens high on heavy volume in the call auction has a significantly higher probability of continuing after the open——the overnight backlog of buying demand (the people who wanted to buy yesterday and did not get filled) is released all at once at the open. The auction is the first health check of the whole day.
• 分时均线攻防。日内分时图上的均价线（约等于当日 VWAP，成交量加权均价）是多空的心理分界：强势股回踩均线不破、缩量回、放量起，是日内趋势健康的标准形态；跌破均线且反抽无力，动量宣告结束。The fight over the intraday average line.The average-price line on the intraday chart (roughly the day's VWAP, the volume-weighted average price) is the psychological dividing line between bulls and bears: a strong stock that pulls back to the line without breaking it, retreats on shrinking volume and turns up on expanding volume — that is the standard shape of a healthy intraday trend; once the line is broken and the bounce back is feeble, the momentum is declared over.
• 二波与回封。早盘冲高回落后，若在均线上方重新放量走出第二波，往往比第一波更可信——第一波里想卖的人已经卖过一轮了。The second wave and the re-seal.After an early-session spike and pullback, if a second wave forms above the average line on renewed volume, it is often more credible than the first——the people who wanted to sell already sold one round during the first wave.

涨停板显微镜：封单的「化学成分」The limit-up board under a microscope: the “chemical composition” of the sealing order queue

涨停是 A 股独有的博弈奇观：价格被钉死在 +10%，供需的较量全部转移到涨停价上的那一队买单（封单）里。看总量只是入门，逐笔数据让你看到封单的成分与动态——这正是连续盘口流最锋利的用武之地。The limit-up is a game spectacle unique to A-shares: the price is nailed at +10%, and the entire contest between supply and demand moves into that one queue of buy orders sitting at the limit price (the sealing order queue). Looking at the total size is only the entry level; tick-by-tick data lets you see the composition and the dynamics of that queue——and this is exactly where a continuous order-book feed is at its sharpest.

• 封单强度三指标：封单金额 / 流通市值（经验上百分之几才算厚）；封成比＝封单金额 / 当日成交额（封单比全天成交还大，说明想买的远多于已成交的）；首次封板时间（早盘板 > 午后板 > 尾盘板，越早封住说明分歧越小）。Three metrics for the strength of the sealing order queue: sealing order value / free-float market cap (empirically it takes a few percent to count as thick); seal-to-turnover ratio ＝ sealing order value / the day's turnover (a queue larger than the whole day's turnover means far more people want to buy than have already been filled); the time of the first seal (a morning seal > an afternoon seal > a late-session seal; the earlier it seals, the smaller the disagreement).
• 板的质量光谱：一字板（开盘即封从未打开）＞ 秒板（开盘后迅速封死）＞ 换手板（反复打开又封回，筹码充分交换）＞ 烂板（尾盘勉强封住、封单稀薄）。质量越高，次日惯性越强，但你也越买不进——排队太长。The quality spectrum of the board: the one-line board (sealed at the open and never opened again) ＞ the instant seal (sealed shut right after the open) ＞ the high-turnover limit-up (opening and re-sealing repeatedly, with chips fully exchanged) ＞ a failed limit-up (barely sealed near the close, with a thin queue). The higher the quality, the stronger the next-day inertia, but the less able you are to get in——the queue is too long.
• 盘口层的先行指标：封单里大单的占比与位置、撤单速率、撑板大单是否悄悄后移。总量还没变，成分已经在变——比看总量的人早半拍发现撤退，是这套数据在打板场景里的全部意义。Leading indicators at the order-book layer: the share and the position of the large orders inside the queue, the cancellation rate, and whether the big orders propping up the board are quietly moving to the back. The total size has not changed yet, but the composition already has——spotting the retreat half a beat earlier than the people who only watch the total is the entire point of this data in a limit-up chasing scenario.

次日剧本与情绪周期The next-day script and the sentiment cycle

这里必须说破打板策略的本质：由于 T+1，今天打的板今天卖不掉——打板是一笔用日内信息下注隔夜的交易。全部收益兑现在次日竞价与早盘：高开放量则兑现或持有，平开看承接，低开执行止损纪律，没有例外。而次日的溢价并非常数，它随全市场的「情绪周期」涨落：启动期首板容错率高，高潮期连板股鸡犬升天，退潮期最高连板股断板（俗称「核按钮」）后，同类股会被连坐性抛售。同一套打法在周期的不同段位，胜率天差地别——很多人以为自己在选股，其实是在被周期选。Here the essence of limit-up chasing has to be said out loud: because of T+1, a board you chase today cannot be sold today——limit-up chasing is a trade that uses intraday information to bet on the overnight. The entire return is realized in the next day's call auction and early session: if it opens high on volume, take the profit or hold; if it opens flat, watch the follow-through; if it opens low, execute the stop-loss discipline, with no exceptions. And the next-day premium is not a constant; it rises and falls with the “sentiment cycle” of the whole market: in the launch phase a first board has a high tolerance for error, in the climax phase the consecutive-board stocks all rise together, and in the ebb phase, once the stock with the most consecutive boards breaks its board (colloquially “the nuclear button”), stocks of the same kind get sold off by association. The very same playbook has wildly different win rates at different points of the cycle——many people think they are picking stocks, when in fact they are being picked by the cycle.

肆IV时间窗口：一天里的潮汐表Time Windows: The Tide Table of a Single Day

交易日不是均匀的。信息、资金和规则在固定的钟点集结，形成结构性的「潮汐」：有的窗口专属价格发现，有的窗口专属机构调仓，有的窗口是规则本身制造的。事件驱动策略的第一课，就是把这张潮汐表刻进直觉。点击下面的时间轴逐段查看。A trading day is not uniform. Information, money and rules gather at fixed hours and form structural “tides”: some windows belong to price discovery, some belong to institutional rebalancing, and some are manufactured by the rules themselves. The first lesson of event-driven strategy is to carve this tide table into your intuition. Click the timeline below to go through it segment by segment.

竞价解剖：十分钟里的两个世界The auction dissected: two worlds inside ten minutes

开盘集合竞价 9:15–9:25 被一条规则切成性质完全不同的两段：9:15–9:20 可以撤单，9:20–9:25 不可以。于是前五分钟是舞台——试盘单、诱导单大量出没，虚拟匹配价被画得千姿百态；后五分钟才是法庭——不能反悔的申报开始显影真实需求。看竞价只需盯三样：匹配价的走向、匹配量的增减、以及 9:20 那一瞬间匹配量有没有塌方（前五分钟的繁荣是不是撤单撤出来的）。这是识别「竞价诱多」成本最低的一招。The opening call auction, 9:15–9:25, is cut by a single rule into two segments of completely different character: from 9:15–9:20 orders can be cancelled, from 9:20–9:25 they cannot. So the first five minutes are a stage——probing orders and decoy orders swarm everywhere, and the indicative match price gets drawn into every imaginable shape; only the last five minutes are a courtroom——declarations that cannot be taken back begin to develop the image of real demand. Watching the auction only takes three things: the direction of the match price, the increase or decrease of the matched volume, and whether the matched volume collapses at that instant of 9:20 (whether the prosperity of the first five minutes was simply cancelled away). This is the cheapest single move for identifying an “auction bull trap.”

尾盘：被规则制造的确定性流量The close: deterministic flow manufactured by the rules

14:30 后是机构的调仓时段；14:57–15:00 沪深两市均以收盘集合竞价定收盘价（沪市 2018 年起改制）。这三分钟在特定日子会出现可预测的巨量：指数调仓生效日，跟踪指数的被动基金必须在收盘价附近完成建仓/清仓——需求的时间和方向是提前公告的，只有价格未知。这类「收盘竞价博弈」是全球通行的机构游戏，散户至少应该知道：调仓日的尾盘异动不是消息，是日历。另外，日常的尾盘急拉急砸（改善净值曲线、做技术图形），次日以均值回归收场的概率偏高——尾盘异动，多看少追。After 14:30 comes the institutions' rebalancing window; from 14:57–15:00 both the Shanghai and Shenzhen markets set the closing price through a closing call auction (Shanghai switched to this system in 2018). On certain days these three minutes produce predictable enormous volume: on the effective date of an index rebalance, the passive funds tracking the index must complete their building/liquidating near the closing price——the timing and the direction of that demand are announced in advance, and only the price is unknown. This kind of “closing auction game” is an institutional game played worldwide, and retail traders should at least know this: an unusual move at the close on a rebalancing day is not news, it is the calendar. On top of that, the everyday sharp ramps and dumps at the close (to improve the net-asset-value curve, to paint the technical chart) have a fairly high probability of ending in mean reversion the next day——when the close moves oddly, watch a lot and chase little.

隔夜之谜：A 股最反直觉的统计事实The overnight puzzle: the most counterintuitive statistical fact in A-shares

多篇学术研究记录了同一个现象：A 股的平均隔夜收益长期为负、日内收益为正——与美股（隔夜为正）恰好相反，主流解释指向 T+1 制度与散户结构：隔夜持仓的风险无法日内对冲，需要折价补偿。对日内交易者的含义很直白：在 A 股，持仓过夜不是默认选项，而是一个需要单独理由的决定。也因此，第參章的打板（主动选择隔夜）必须由足够强的封单证据来支付这笔「隔夜税」。Several academic studies have recorded the same phenomenon: the average overnight return in A-shares has long been negative, while the intraday return is positive——exactly the opposite of US equities (where the overnight return is positive), and the mainstream explanation points to the T+1 system and the retail investor structure: the risk of an overnight position cannot be hedged intraday, so it has to be compensated with a discount. The implication for the intraday trader is blunt: in A-shares, holding overnight is not the default option but a decision that requires its own separate reason. And that is precisely why the limit-up chasing of Chapter III (deliberately choosing to go overnight) has to pay this “overnight tax” with strong enough evidence from the sealing order queue.

伍V迁移 Polymarket：换了物种的市场Migrating to Polymarket: A Market of a Different Species

先给总判词：策略不可搬运，方法论可以。微观结构研究的本体，是「读懂交易机制在价格上留下的指纹」；换一个市场等于换一套机制，指纹要重新采集。Polymarket 与 A 股的差异不是「另一个股市」，而是另一个物种——但恰恰因为你理解了 A 股的机制如何塑造策略，你才有能力快速看清那边的机制会塑造出什么。The verdict first: strategies do not transfer; methodology does. What microstructure research is really about is reading the fingerprints that a trading mechanism leaves on price; change the market and you change the mechanism, so the fingerprints have to be collected all over again. The difference between Polymarket and the A-share market is not that it is "another stock market" — it is another species. But precisely because you understand how A-share mechanics shape strategy, you are equipped to see quickly what the mechanics over there will shape.

物种对照表A Species Comparison

三类策略的迁移判决Migration Verdicts on the Three Strategy Families

A 股根本不存在的四件原生武器Four Native Weapons That Simply Do Not Exist in A-Shares

• 互补套利。二元市场里 YES + NO 理应恒等于 1 美元；多结果市场里所有结果的价格之和理应等于 1。当挂单簿两边加起来小于 1（可同时买入锁定无风险差价）或多结果之和大于 1（可全部卖出），就是结构性套利。缺口通常极小且转瞬即逝，但它存在——这是 A 股里没有对应物的「机制白送钱」。Complementary arbitrage. In a binary market YES + NO ought to be identically 1 dollar; in a multi-outcome market the prices of all outcomes ought to sum to 1. When the two sides of the order book add up to less than 1 (buy both at once and lock in a risk-free spread), or the multi-outcome sum exceeds 1 (sell them all), that is structural arbitrage. The gaps are usually tiny and vanish in a blink, but they exist — this is "free money from the mechanism", something with no counterpart in A-shares.
• 跨所套利。同一事件常在 Polymarket 与 Kalshi 同时挂牌，两边报价的分歧即机会。需要两边账户、资金与合规条件，摩擦不小，但逻辑上是最干净的价差交易。Cross-venue arbitrage. The same event is often listed on Polymarket and Kalshi at the same time, and any divergence between the two sets of quotes is an opportunity. It requires accounts, capital and compliance standing on both sides, and the friction is not small, but logically it is the cleanest spread trade of all.
• 聪明钱跟踪。A 股里你永远不知道对手盘是谁——机构持仓一季度才披露一次。这里每个钱包的每笔交易与持仓全部公开：可以给任何地址算历史胜率与校准分，实时看高分钱包在建什么仓。「跟单聪明钱」在 A 股是物理不可能，在这里只是个数据工程问题。这也是为什么围绕它做数据产品，本身就是一门生意。Smart-money tracking. In A-shares you never know who is on the other side of your trade — institutional holdings are disclosed only once a quarter. Here every trade and every position of every wallet is fully public: you can compute a historical hit rate and a calibration score for any address, and watch in real time what the high-scoring wallets are building. "Copy-trading smart money" is physically impossible in A-shares; here it is merely a data-engineering problem. Which is also why building a data product around it is a business in its own right.
• 做市＋流动性返佣。平台把吃单费的一部分每日返给挂单方。对会写程序的人，这是一份「底薪＋绩效」的官方工作：底薪是返佣，绩效是价差，风险是被知情流扫穿——于是第貳章的全部失衡功夫，在这里变成了做市商的防身术。Market making ＋ liquidity rebates. The platform returns part of the taker fee to the maker every day. For anyone who can write code this is an official job paying "a base salary ＋ performance pay": the base salary is the rebate, the performance pay is the spread, and the risk is being run through by informed flow — so all the imbalance craft of Chapter II turns, over here, into a market maker's self-defence technique.

三盆冷水Three Buckets of Cold Water

• 容量。头部市场（大选级）有深度，长尾市场几千到几万美元就能打穿簿。这决定了那边的策略天花板是「零花钱到小生意」量级，不是资管量级——对个人是优点（机构看不上，竞争者少），对野心是缺点。Capacity. The top markets (election-scale) have depth; in long-tail markets a few thousand to a few tens of thousands of dollars is enough to punch straight through the book. That fixes the ceiling of strategies over there at the "pocket money to small business" scale, not the asset-management scale — an advantage for an individual (institutions cannot be bothered, so there are few competitors) and a drawback for ambition.
• 对手性质。体育与新闻类市场里，靠更快数据源抢跳的延迟机器人已把速度游戏卷到毫秒级；事件市场天然存在真正的知情者。你的每一张吃单，都要先问一句：为什么这个价格轮得到我？The nature of the counterparty. In sports and news markets, latency bots that race the jump on faster data feeds have already ground the speed game down to the millisecond; and event markets inherently contain genuinely informed traders. Before every single order you take, ask one question first: why did this price get left for me?
• 合规与平台风险。平台条款明列受限司法辖区；中国大陆居民参与境外此类平台的法律定性与资金通道风险需自行评估，本文不构成任何参与建议。此外预言机裁定争议、智能合约风险、稳定币风险都真实存在。Compliance and platform risk. The platform's terms explicitly list restricted jurisdictions; residents of mainland China must assess for themselves the legal characterisation of taking part in such offshore platforms and the risks of the funding channel, and this article does not constitute any recommendation to participate. Beyond that, oracle resolution disputes, smart contract risk and stablecoin risk are all real.

陸VI外行路线图：复现，然后才谈钱A Layperson's Roadmap: Reproduce First, Talk Money Later

理解一个策略的唯一标准，是能亲手复现它的统计规律。以下四步全部可以零风险完成，任何一步复现不出来，就停在那一步，别往下走——更别掏钱包。The only test of whether you understand a strategy is whether you can reproduce its statistical regularities with your own hands. All four steps below can be completed at zero risk; wherever a step fails to reproduce, stop there and go no further — and certainly do not reach for your wallet.

• 第一步：复现 QI 的单调性。选一只厚队列的低价大盘股，用免费的快照行情录一到两周数据，统计不同 ρ 分段下「下一次价格变动向上」的频率。如果你亲眼看到频率随 ρ 单调爬升，第貳章就从知识变成了你的经验。Step one: reproduce the monotonicity of QI. Pick a low-priced large-cap with thick queues, record one to two weeks of data from a free snapshot feed, and tabulate the frequency of “the next price change is upward” across different ρ buckets. If you watch that frequency climb monotonically with ρ with your own eyes, Chapter II stops being knowledge and becomes your own experience.
• 第二步：纸面解剖十个涨停板。不下单。连续跟踪十只涨停股的封单金额、封成比、封板时间与次日竞价表现，用实验二的框架逐一记录。十个样本足够让你看清「封单化学成分」与次日溢价的关系，也足够让你看见几次炸板的残酷。Step two: dissect ten limit-up boards on paper. Place no orders. Track ten limit-up stocks in a row — the amount of the sealing order, the seal-to-turnover ratio, the time the board was sealed, and how each one behaved in the call auction the following morning — and log every one of them with the framework from Experiment 2. Ten samples are enough to show you how the “chemical composition of the sealing order” relates to the next day’s premium, and enough to let you witness a few brutal moments when the board cracked open.
• 第三步：录一周 Polymarket 的簿。用官方公开的 WebSocket 接口录下若干市场的订单簿与成交流（这一步同时是数据工程练习），复算失衡指标，标注每次新闻跳变前后的簿形变化。你会亲眼看到「薄簿的失衡有多吵」与「信息如何碾压微观结构」。Step three: record a week of Polymarket’s book. Use the official public WebSocket interface to record the order books and trade flow of several markets (this step doubles as a data-engineering exercise), recompute the imbalance metric, and annotate how the shape of the book changes before and after each news jump. You will see for yourself “how noisy imbalance is on a thin book” and “how information steamrolls microstructure”.
• 第四步：只有统计在你自己手里复现了，才谈真钱。而且从第貳章「路径三」这种最温和的用法开始：给你本来就要做的交易省几个 bp。任何跳过前三步直接下注的冲动，都是这份教学的失败。Step four: only once the statistics have reproduced in your own hands do you talk about real money. And even then, begin with the mildest use of all, “Path Three” from Chapter II: saving a few bp on trades you were going to make anyway. Any urge to skip the first three steps and go straight to betting is a failure of this lesson.