5D Library › Books › 5DL-B-059
预测市场微观结构、布雷格曼信息投影与跨盘口无套利边界
对数市场评分规则、冷热门校准偏差、资金锁仓机会成本与组合定价算法
Prediction Market Microstructure, Bregman Information Projection and Cross-Book Arbitrage Bounds: Logarithmic Market Scoring Rules, Favorite-Longshot Calibration Bias, Lock-Up Opportunity Cost and Frank-Wolfe Combinatorial Pricing
DOI10.5281/zenodo.24090166摘要
预测市场的价格常被读作概率:价格 0.20,事件就有 20% 的可能。本书把这个读法拆成三件事——信念如何聚合成价格、做市机制如何为信息付钱、资本被锁住时价格还能偏离多远——逐一写成可运行的模型,并计算看上去的套利还剩多少能执行。本书不是投资、交易或博彩建议;全部数据为程序生成的合成数据。
第一章用 Manski 与 Wolfers-Zitzewitz 的模型说明价格是信念分布的分位数或加权均值:平均只信 0.100 的人群,价格是 0.167;对数效用下价格恰为财富加权均值 0.0989,风险态度与信念相关时可偏离到 0.1444;只要信念分散、价格有界,价格 0.05 到 0.10 的合约平均价 0.084,真实概率只有 0.024——冷热门偏差不需要任何人爱好风险。第二章核对 Hanson 对数市场评分规则的凸对偶:价格是 Softmax,交易者的期望收益恰为流动性参数乘以 KL 散度(布雷格曼投影),做市商最坏亏损恰为参数乘以结果数的对数,两个结果时 69.31;补贴与价格冲击之积趋于常数 34.66。第三章建立资金锁仓的无套利带:真实概率 0.99、资金成本 14.5% 时,半年期合约的下沿是 0.9252,一年期是 0.8646,这解释了几乎确定的合约为何停在 0.92 到 0.96,也让冷热门偏差在长期合约中存活。
第四章计算互斥结果盘口的可执行套利:卖价之和 0.96 的五结果盘口,最好的情形年化 +7.0%,按 0.07·P(1−P) 收费时任何规模都亏钱;陈旧报价制造的套利逐腿执行后平均亏损。第五章把逻辑相关的合约放进凸多面体,手写 Frank-Wolfe 投影求出在所有状态下都赚钱的套利证书;普通变体走 400 步仍未收敛,成对变体精确收敛;两两逻辑规则只抓住 92.9% 的不一致,执行后最好的年化回报只有 +4.6%。第六章说明偏差小到难以确认:1,000 份独立合约的检验把握度只有 0.575,事件内相关系数 0.3 时 1,000 份合约只相当于 139 份,灵活的修正使预测变差。第七章把《管子·国蓄》的「智者有什倍人之功」与《史记》白圭的「人弃我取,人取我与」读作信息租与持有成本,并明确标为诠释。第八章给出四步核查与一项预注册的锁仓检验(600 个事件,把握度 0.950)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。
关键词:预测市场;市场微观结构;冷热门偏差;校准;风险厌恶;对数市场评分规则;自动做市商;适当评分规则;Softmax;KL 散度;布雷格曼散度;信息投影;凸对偶;资金成本;套利的极限;荷兰赌;交易成本;腿部风险;组合市场;边际多面体;Frank-Wolfe 算法;布赖尔分数;聚类标准误;预注册
Abstract
Prediction-market prices are often read as probabilities: a price of 0.20 means a 20 % chance. This book splits that reading into three questions — how beliefs aggregate into a price, how the market mechanism pays for information, and how far a price can stray when capital is locked up — writes each as a runnable model, and computes how much of an apparent arbitrage can actually be executed. It is not investment, trading or betting advice; all data are synthetic, generated by its programs.
Chapter 1 uses the models of Manski and of Wolfers and Zitzewitz to show that a price is a quantile or weighted mean of the belief distribution: a population whose mean belief is 0.100 trades at 0.167; under log utility the price is exactly the wealth-weighted mean, 0.0989, and when risk attitude goes with belief it can move to 0.1444; with dispersed beliefs and bounded prices alone, contracts priced from 0.05 to 0.10 average 0.084 against a true probability of only 0.024 — the favourite-longshot bias needs no one who loves risk. Chapter 2 checks the convex duality of Hanson's logarithmic market scoring rule: prices are a softmax, a trader's expected gain is exactly the liquidity parameter times a KL divergence (a Bregman projection), and the market maker's worst-case loss is exactly the parameter times the log of the number of outcomes, 69.31 with two; subsidy times price impact tends to a constant, 34.66. Chapter 3 builds the no-arbitrage band of capital lock-up: with a true probability of 0.99 and a 14.5 % cost of capital its lower edge is 0.9252 at six months and 0.8646 at one year, which explains why near-certain contracts sit at 0.92 to 0.96 and lets the favourite-longshot bias survive in long-dated contracts.
Chapter 4 computes the executable arbitrage in a book of exclusive outcomes: for a five-outcome book whose asks sum to 0.96, the best case is +7.0 % a year, and with a fee of 0.07·P(1−P) every size loses; arbitrage created by stale quotes loses money when executed leg by leg. Chapter 5 puts logically related contracts into a convex polytope and uses a hand-written Frank-Wolfe projection to find a certificate that gains in every state; the plain variant is unconverged after 400 steps while the pairwise variant converges exactly; pairwise logical rules catch only 92.9 % of incoherent quotes, and the best executed return is only +4.6 % a year. Chapter 6 shows that the bias is hard to confirm: with 1,000 independent contracts a test has a power of only 0.575, with a within-event correlation of 0.3 1,000 contracts are worth 139, and a flexible correction makes forecasts worse. Chapter 7 reads the *Guanzi*'s "the clever gain tenfold" and the *Shiji*'s Bai Gui, "what others discard I take, what others take I let go", as information rent and the cost of carry, marked as interpretation. Chapter 8 gives a four-step check and a pre-registered test of the lock-up discount (600 events, power 0.950). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses.
Keywords: prediction markets; market microstructure; favourite-longshot bias; calibration; risk aversion; logarithmic market scoring rule; LMSR; automated market maker; proper scoring rules; softmax; KL divergence; Bregman divergence; information projection; convex duality; cost of capital; limits of arbitrage; Dutch book; transaction costs; leg risk; combinatorial markets; marginal polytope; Frank-Wolfe algorithm; Brier score; clustered standard errors; pre-registration
目录 · Contents
- 序:价格、概率与资本English 中文Preface: Prices, Probabilities and Capital
- 第一章 价格不是概率:信念分布、风险态度与冷热门偏差English 中文Chapter 1. A Price Is Not a Probability: Belief Dispersion, Risk Attitudes and the Favourite-Longshot Bias
- 第二章 对数市场评分规则:Softmax、KL 散度与补贴上限English 中文Chapter 2. The Logarithmic Market Scoring Rule: Softmax, KL Divergence and the Subsidy Cap
- 第三章 资金锁仓:为什么几乎确定的合约停在 0.92 到 0.96English 中文Chapter 3. Capital Lock-Up: Why Near-Certain Contracts Sit at 0.92 to 0.96
- 第四章 互斥结果的盘口:价格之和不等于 1 时的可执行套利English 中文Chapter 4. Books of Exclusive Outcomes: Executable Arbitrage When Prices Do Not Sum to 1
- 第五章 逻辑相关的事件:凸多面体、Frank-Wolfe 投影与套利证书English 中文Chapter 5. Logically Related Events: Convex Polytopes, Frank-Wolfe Projection and Arbitrage Certificates
- 第六章 从价格读概率:校准检验、相关结算与修正的代价English 中文Chapter 6. Reading Probabilities from Prices: Calibration Tests, Correlated Resolutions and the Cost of Correction
- 第七章 「智者有什倍人之功」与「乐观时变」:信息租与持有成本English 中文Chapter 7. "The Wise Gain Tenfold" and "Watching the Turn of the Times": Information Rent and the Cost of Carry
- 第八章 四步核查与一项预注册的锁仓检验English 中文Chapter 8. A Four-Step Check and a Pre-Registered Test of the Lock-Up Discount
How to cite · 引用
Fan, Y. (2026). Prediction Market Microstructure, Bregman Information Projection and Cross-Book Arbitrage Bounds: Logarithmic Market Scoring Rules, Favorite-Longshot Calibration Bias, Lock-Up Opportunity Cost and Frank-Wolfe Combinatorial Pricing [预测市场微观结构、布雷格曼信息投影与跨盘口无套利边界] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090166
BibTeX
@book{fan2026prediction,
title = {{Prediction Market Microstructure, Bregman Information Projection and Cross-Book Arbitrage Bounds}: Logarithmic Market Scoring Rules, Favorite-Longshot Calibration Bias, Lock-Up Opportunity Cost and Frank-Wolfe Combinatorial Pricing},
author = {Fan, Yafu},
year = {2026},
publisher = {5D Library},
edition = {Version 1.0.0},
doi = {10.5281/zenodo.24090166},
url = {https://doi.org/10.5281/zenodo.24090166},
note = {Chinese title: 预测市场微观结构、布雷格曼信息投影与跨盘口无套利边界}
}
| DOI | https://doi.org/10.5281/zenodo.24090166 |
|---|---|
| DOI(全部版本 all versions) | https://doi.org/10.5281/zenodo.24090165 |
| ORCID | https://orcid.org/0009-0004-3519-1203 |
| 5D Library ID | 5DL-B-059 |
| 版本 Version | 1.0.0 · 中英双语版 Bilingual edition |
| 出版日期 Date | 2026-11-07 |
| 网址 Online | https://5dlibrary.com/books/prediction-market-microstructure-bregman-projection-arbitrage/ |
| SHA-256 · prediction-market-microstructure-bregman-projection-arbitrage_bilingual_v1.0.0.pdf | 9522804eea101d9021e82e613564745f13bdf49651047c9670abe48e799b73f0 |
| SHA-256 · prediction-market-microstructure-bregman-projection-arbitrage_en_v1.0.0.pdf | d0d94de8504007373bff1dd9620b795ff74bc7dd462956db802685adf3687d92 |
| SHA-256 · prediction-market-microstructure-bregman-projection-arbitrage_zh_v1.0.0.pdf | eaf438bf2f02158d8898df55d31021fa12b6af516b4e21b6752ef10160fd492c |
| Licence · 许可 | Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0). https://creativecommons.org/licenses/by-nc-nd/4.0/ |
Disclosure. The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and therefore has a commercial interest in the subject of this book. The book promotes no course, card deck, software, exchange, prediction-market platform, signal service or fund, provides no link to or recommendation of any trading platform, and accepts no sponsorship from any financial firm or platform. It is not investment, trading or betting advice. The test in Chapter 8 uses public archive data only, is to be run and analysed by people independent of any trading decision, and its code and protocol are published openly for anyone to check.