FREESEE IT
A · MechanicsReference
Glossary
English ⇄ 中文
2026-08-21 · 85 entries · with the bases I had to learn the hard way
Nearly every Chinese-language source in this field is a direct translation of English terms,
and the same word routinely gets three renderings. This table is the reference I write against.
It differs from other glossaries in one way: the last column is not a definition,
it is what tripped me up — where it is marked ⚠,
the literal reading is wrong.
How to usesearch anything you do not recognise. Arriving from
You are buying a verdict, sections A and B are enough
to read any board; C onward only matters once you are actually trading.
AMarkets & contracts16
| English | 中文 | In one line | My basis |
|---|
| prediction market | 预测市场 | A market where prices express the probability of an event | — |
| outcome token | 结果代币 | Pays $1 if its outcome wins, $0 otherwise | — |
| CLOB | 中央限价订单簿 | How orders match — not an AMM pool | Polymarket is a CLOB, not an automated pool |
| maker / taker | 做市方 / 吃单方 | Posting liquidity vs consuming it | makers pay no fee; the whole fee sits on the taker |
| split | 拆分 | USDC → one full set of complementary tokens | — |
| merge | 合并 | A full set → USDC back | the activity feed carries a MERGE row — no inference needed |
| redeem | 赎回 | After settlement, convert the winner to USDC | do not trust a redeemable flag — see the phantom-empty trap |
| complementary set | 互补集 | All mutually exclusive outcomes of one event | — |
| NegRisk | 无风险聚合 | The complementary-set mechanism for multi-outcome events | — |
| resolution | 结算 | Who rules on the outcome, and on what | the source differs per category — read it every time |
| optimistic oracle (UMA) | 乐观预言机 | Assume the submission is true; challengeable within a window | — |
| up/down market | 上下盘 | Does a crypto price rise or fall within a fixed window | — |
| tenor | 周期 | How long a board runs (5m / 15m / 1h / 4h) | tenor decides everything: pair rates 47% / 71% / 87% / 95% |
| window | 窗 | One complete cycle of an up/down board | I weight statistics per window, equally |
| strike | 行权价 | The reference price for up vs down | — |
rewardsMinSize | 挂单最小额 | Minimum size to qualify for rewards | metadata — it does not mean the board actually pays |
BFees & subsidies9
| English | 中文 | In one line | My basis |
|---|
| taker fee | 吃单费 | What the taker pays | Measured: fee = 0.07·p(1−p)·shares; heaviest at 50¢ |
| theta (θ) | 费率系数 | The 0.07 in that formula | Measured by reconciliation, not documented |
| maker rebate | 做市返还 | 20% of the taker's fee returns to the filled maker | ≈ +0.25¢/share, only on fills — unrelated to how long you quoted |
| taker rebate ladder | 吃单阶梯返利 | Six tiers by 30-day rolling weighted volume | Bronze 3% → Obsidian 50%; taker volume only, maker volume does not count |
| weighted volume (wV) | 加权量 | The ladder's unit of account | stake ×(1−entry price)× category weight; Crypto 2.3, Geopolitics 0 |
| LP rewards | 流动性奖励 | Scored on quote quality, paid from a daily pool | clobRewards:[] does not mean no rewards — check the sampling endpoint |
| referral | 邀请返利 | 10% of the referee's net taker fee, for 30 days | current terms date from 2026-05-28; most guides still quote 30%/180 days |
| net fee | 净费 | Venue receipts = taker fee − the referee's own ladder rebate | Referral pays on net, not gross |
| effective uplift | 等效增厚 | Rebates restated as points off each trade | Both maxed ≈ 0.17–0.18pp against a fee line of 1.75pp |
CBook & microstructure14
| English | 中文 | In one line | My basis |
|---|
| spread | 价差 | Best ask minus best bid | — |
| mid | 中价 | The midpoint of the spread | A price at which nothing trades |
| depth | 深度 | Resting size at each level | State how many levels, and one side or both |
| queue position | 排队位置 | Your place among orders at the same price | simulate conservatively: assume all displayed size sits ahead of you |
| adverse selection | 逆向选择 | The fills you get are the ones you did not want | The real cost of market making — racing the rebate |
| toxicity | 毒性 | How much a flow hurts the maker | Measured via short-window price drift after the fill |
| markout | 成交后漂移 | Price drift shortly after your own fill, excluding it | Tells you whether that fill was good or bad |
| walk the book | 吃穿订单簿 | Fill through successive levels to reach a size | Backtests must use real levels — a flat price gives a false answer |
| notional | 名义额 | The money size of one order | CLOB minimum is 5 shares → $5 is the only price-independent fixed stake |
| clip | 单笔规模 | Size of a single order | — |
| no-fill | 未成交 | You wanted the fill and did not get it | Report the rate separately — never let them vanish from the sample |
| pair rate | 配对率 | Share of two-sided fills that form mergeable sets | truncated replay overstates it — measured 1.3× to 2.1× |
| naked leg | 裸腿 | The unmatched side | a losing naked leg leaves no trace in the activity feed |
| inventory / exposure | 库存 / 敞口 | What you are holding and which way it leans | — |
DSports & odds15
| English | 中文 | In one line | My basis |
|---|
| vig / juice / overround | 水位 / 抽水 | The bookmaker's margin baked into the odds | The amount by which implied probabilities exceed 100% |
| devig | 去水 | Strip the margin to recover fair probabilities | — |
| the four devig methods | 去水四法 | multiplicative / additive / power / Shin | they diverge noticeably at long odds — exactly where spreads appear |
| fair probability | 公允概率 | The post-devig probability | Methods differ most at long odds; test a finding under all four |
| sharp book | 锐利庄家 | Prices accurately, takes size, treated as consensus | — |
| soft book | 软庄家 | Retail-facing, slower to move | — |
| closing line | 收盘线 | The last price before kickoff | The industry's accepted benchmark |
| CLV | 胜过收盘线 | How much better your entry was than the close | The standard skill metric — converges far faster than P&L |
| American / decimal / HK odds | 美式 / 小数 / 香港赔率 | Three ways of writing the same thing | — |
| moneyline | 独赢 | Straight winner | — |
| spread / handicap | 让分 | Winner with a head start | — |
| total / over-under | 大小球 | Combined score above or below a line | — |
| 3-way | 三路盘 | Win / draw / loss — typical in football | A three-outcome complementary set |
| pregame / in-play | 赛前 / 滚球 | Before the start vs during | the real winners' money is almost entirely pregame |
| odds API | 赔率 API | A feed of bookmaker odds | mainstream odds APIs have zero esports coverage |
EResearch method18
| English | 中文 | In one line | My basis |
|---|
| preregistration | 预注册 | Freeze the pass/fail criteria before looking | Timestamp it; afterwards you may append, never revise |
| gate | 判决门 | A pass condition written down in advance | Any one failing is a failure; the bar is never lowered |
| single read | 单次读数 | Read once at expiry; no peeking for a decision | Stops you from stopping at the prettiest number |
| kill rule | 杀死规则 | The condition under which the line stops | — |
| pre-claimed failure modes | 预先认领的死法 | Write down how it will die, before it runs | Claim one from the list afterwards; inventing new ones is not allowed |
| exploratory scan | 探索扫描 | Hypothesis generation — disclose the whole grid | Scan output is never a conclusion by itself |
| promotion rule | 晋级规则 | What a scan finding must pass before it gets money | Stops cherry-picking cells |
| shadow / paper trade | 影子盘 | Run it without money first | — |
| forward test | 前向验证 | Validate on data generated after the fact | — |
| out-of-sample | 样本外 | Not part of what the model was built on | — |
| null / baseline | 空模型 / 基准 | A control that uses no strategy at all | a profit that cannot beat the null is not an edge |
| replay / backtest | 回放 | Re-run on historical data | State the direction of the bias — conservative or optimistic |
| measuring stick | 量尺 | The ground truth you judge wins and losses against | rulers break: oracle TWAP vs exchange close mislabelled 10.1% of the whole sample, with direction |
| settle label | 结算标签 | Labels taken from actual market settlement | House rule: score against settlement; the other basis is a companion read only |
| excess (pp) | 超额 | mean(win − entry price), points per share | — |
| pp | 百分点 | An absolute difference in probability | Kept strictly separate from % |
| adversarial verification | 对抗验证 | Assign someone to overturn your own finding | Default: if it cannot be shown to hold, treat it as overturned |
| epoch | 纪元 | A sample segment between configuration changes | never pool across epochs |
FEngineering & data9
| English | 中文 | In one line | My basis |
|---|
| relayer | 中继器 | A service that submits chain transactions for you | — |
| EOA | 外部账户 | An ordinary key-controlled address | — |
| allowance | 授权额度 | How much of your token a contract may move | — |
| disperse | 批量派发 | Batched reward transfers | About 400 per transaction |
| tape | 逐笔成交流 | Every fill, in order | the data API caps at 5,500 rows |
| snapshot | 快照 | A point-in-time slice of book or positions | — |
| hard stop / circuit breaker | 硬停 / 熔断 | Trip and halt | poisoned input causes a false halt — validate before you stop |
| shakedown | 预检期 | An opening period spent checking the instrument works | Do not collect a full sample and only then discover it was wrong |
| commit-reveal | 承诺-揭示 | Publish a hash now, the contents later | Proves what you thought at the time without disclosing it |
GThe "don't confuse these" list
The four pairs I get wrong most often when writing.
| Easily confused | The difference |
|---|
| The four “rebates” | Maker rebate (only on fills) · taker ladder (by taker volume) · LP rewards (on quote quality) · referral (on recruits) — four rulebooks, four pools, four payout times |
| Two clocks | The settling price (on-chain TWAP) ≠ the price you watch (exchange spot). Confusing them is a broken ruler |
| Two “profits” | Trading P&L ≠ net income. ⚠ the leaderboard's /profit field excludes every subsidy — subsidies are transfers, not trades |
| pp vs % | “hit rate from 50% to 52%” is +2pp, not +2%. Every edge on this site is in pp |
BasisWhy the last column is the point
Glossaries normally stop at translation and definition. But in this business
almost nothing is lost on definitions — it is lost on bases:
what you thought a number meant differs slightly from what it means, and everything computed from it is wrong.
| Word | Literal reading | Actually |
|---|
| rebate | money the venue gives back |
four unrelated mechanisms with different rules, pools and payout times →
the four rebates |
| profit | what someone made |
trading P&L ≠ net income. Subsidies are transfers and never enter that field |
| ruler | the price that decides wins |
possibly not the price you watch →
10.1% measured |
And one unglamorous item that keeps causing damage: pp versus %.
"Hit rate from 50% to 52%" is +2pp, not +2% (that would be 51%).
Mix the two symbols once and the magnitudes in an entire document become unreadable.
Every edge here is stated in pp.
RulesHow this table is maintained
Single source of truth my own writing follows this table; where prose and table disagree, the prose changes.
Add before you write a term not in the table gets added first, then used. That is how renderings stay stable across entries.
Untranslated tables, code, numbers and formulas stay as they are.
Where the ⚠ column comes from entirely from my own live trading and reconciliation, not copied from documentation.
Wherever it is checkable, the method is given in the corresponding entry.
What is not here this table covers public mechanics only. Internal names,
parameters and features from my own strategies are not in it and will not be added.
NextWhere to go