数据注水?Polymarket百亿估值的真正逻辑是…

比推Publicado em 2025-11-12Última atualização em 2025-11-12

最近,哥伦比亚大学的一篇研究论文让风口上的“预测市场(Prediction Markets)”陷入争议。

image.png

研究团队分析了区块链平台 Polymarket 的两年历史数据,发现约 25% 的成交量可能属于清洗交易(wash trading)——即同一主体在自己账户之间对倒买卖,制造虚假活跃度。在某些热点事件周,比如美国大选或体育总决赛,这一比例甚至飙升至 60%。

image.png

image.png

研究虽未正式同行评审,但足以撕开预测市场狂热表象的一角。因为过去半年,这个赛道的热度几乎肉眼可见监管放宽、巨头加持、资本狂飙、政治加码——预测市场,正在成为2025年最受瞩目的金融新物种

从“边缘博彩”到“金融新物种”

预测市场的玩法并不复杂:你可以押注“特朗普能否赢得选举”“美联储是否会降息”“下一届诺奖得主是哪国人”等事件,平台根据交易双方的价格形成“市场概率”,被视作“群体智慧”的体现。

2025年,这种“用钱投票”的方式迎来了三重爆发契机:

监管松绑
今年5月,美国商品期货交易委员会(CFTC)撤回了对 Kalshi 的诉讼,正式承认预测合约可以在“特定框架下合法交易”。
9月,CFTC 又向 Polymarket 发出“不执法函”(No-Action Letter),允许其重新开放美国市场。

这意味着预测市场从“灰色地带”走向“监管可视”,为资本介入扫清最大障碍。

资本+政治下注
紧接着,资金蜂拥而至:

8月,Polymarket 获得特朗普长子 Donald Trump Jr. 参股的 1789 Capital 投资;

image.png

然后,继9月纽约证券交易所母公司ICE以20亿美元投资将Polymarket估值推高至80亿美元,以及10月竞争对手Kalshi在a16z与红杉资本领投下估值达到50亿美元后,市场热情仍在急剧升温。

据彭博社最新消息,Polymarket正寻求以120亿至150亿美元的更高估值进行新一轮融资,而Kalshi的估值据信也已突破100亿美元大关。

在这场资本狂飙的背后,政治力量的深度介入不容忽视。

特朗普政府上台后营造的“市场友好型”监管氛围,为预测市场的爆发铺平了道路。CFTC的态度转变、ICE的巨额投资,均被市场解读为明确的政策信号。

更值得注意的是特朗普家族的亲身参与:Donald Trump Jr.不仅通过1789 Capital投资Polymarket,还出任Kalshi的顾问职务;

image.png

ICE首席执行官Jeff Sprecher—他也是美国前小企业管理局局长Kelly Loeffler的丈夫–亲自主导了对Polymarket的投资;

而特朗普旗下社交平台Truth Social也宣布将推出自有加密预测平台“Truth Predict”。

资本、政策与家族影响力的多重合力,正推动预测市场从边缘试验走向主流金融舞台。

巨头助推主流化
10月,谷歌(Google)宣布将在 Google Finance 搜索结果中整合 Polymarket 和 Kalshi 的实时预测数据,比如,当用户搜索“2028年总统是谁”或“美联储降息概率”时,结果下方会出现预测市场的实时数据图。

image.png这意味着,预测市场第一次被“嵌入”到全球最大的信息入口中,成为公共信息流的一部分。

谷歌没有披露与两家公司的具体合作模式,但对于市场来说,这一动作堪称“主流化里程碑”:预测市场从“币圈玩家的对赌工具”,变成了普通用户可见、媒体可引用的数据产品。

结果很明显:10 月 Polymarket 的交易量创下历史新高,单月交易额超过30亿美元,用户数较9月份增长了93.7%。

image.png

哥大质疑的“假成交”,究竟有多严重?

回到哥大论文的研究数据:Polymarket 在 2024-2025 年间有大约四分之一的成交存在可疑模式:账户之间频繁对倒、成交时间间隔极短、几乎不持仓结算。这些特征与过去 NFT 市场的“刷量”极为相似。

报告作者推测,预测市场 wash trading 动机主要有三种:
① 为争夺未来代币空投或激励积分;
② 为制造市场热度吸引新用户;
③ 个别做市商通过“假成交”稳定价格区间。

换句话说,有人可能为了刷活跃度、抢积分、获得未来代币奖励,而反复在市场里“假下单”。这在加密领域并不陌生:从NFT到DeFi,几乎每一轮创新都伴随“刷数据”行为,但即便如此,预测市场的“水分”并非行业最高。对比来看:

未受监管的比特币交易所早期“假量”曾高达 70% 以上(来自 Bitwise 2019 年的报告)。

image.png

在 NFT 市场,市场火热时期的 washing trading 比例也在 20%—50% 之间。

image.png

相较之下,Polymarket 的平均 25% 属于“中等偏高”水平。再加上 Kalshi 的合规性更强、KYC 严格,整个行业的“真实度”其实已远超早期的币圈,因此,从行业维度看,预测市场的“注水并非灾难性问题。

此外,针对哥大研究的结论,业内也出现了不同的声音。

AWS 前工程师 yassinelanda.eth 在审阅该论文后提出了几点反驳

他认为,这项研究在方法上存在局限——其结论建立在单一的链上数据模型之上,而像 Polymarket 这样的平台实际拥有更复杂的信号系统来识别真实用户并公平分发奖励。此外,研究的结论对分析时设置的参数非常敏感,其揭示的问题严重性可能并不稳定。

他进一步指出了预测市场的一个关键特性:在这个领域,有价值的信号远比原始交易量重要。单纯的“左手倒右手”刷量循环无法产生真正的利润(PNL)。如今,先进的链上监控和推荐系统已经能够有效区分有信息的真实交易流与来自做市商、机器人及自成交的市场噪音,并在推荐和奖励中降低后者的权重。

在他看来,评判一个预测市场的核心标准不应是“总交易量”这个易掺水的表面数据,而应是:

  • 预测准确度:市场的结果准不准。

  • 校准度:预测概率与实际发生频率是否匹配。

  • 买卖价差与市场深度:市场流动性好不好,交易成本高不高。

  • 新闻事件时的滑点:在有新消息时,价格能否迅速、平稳地反应,而不是剧烈波动。

这些关于市场质量和信息效率的指标,才是衡量预测市场价值的真正核心。

赌性回潮:当“下注”成为时代情绪

正如芝加哥大学社会学者莉迪亚·格兰特所观察到的:“预测市场在某种意义上延续了美国的信仰体系—它让人们在巨大的不确定性中,依然能通过‘下注’这个动作,获得一种虚幻的掌控感。”

这句话,精准捕捉了当下美国的社会脉搏。面对高通胀、政治撕裂与阶层固化的现实,一种“赌徒心态”正悄然成为普遍的情绪出口,从体育博彩到加密货币,再到如今的预测市场,越来越多的美国人开始将命运交给概率,把焦虑释放在盘口之间。

而当华尔街巨头也投身其中,这一趋势更获得了资本与制度的双重认证。 ICE等机构的巨额投资表明,主流金融界正将预测市场视为下一代“事件驱动型”风险定价的基础设施,而不再仅仅是边缘的博彩游戏。

正如 SynFutures 首席执行官 Rachel Lin 所指出的:“预测市场的真正价值,在于能量化传统金融无法定价的东西,比如政策决策、技术突破和地缘政治风险。”

与此同时,Polymarket 推出 POLY 代币等动作为生态发展注入了新燃料。研究机构 Delphi Digital 认为,未来融合多市场数据与AI分析的预测“终端”,很可能开启一个类似meme币热潮的新交易赛道。

当然,挑战依然存在。 美国监管机构仍在为“衍生品”还是“赌博”的定义争论不休,这层未散去的政策阴云,仍是预测市场走向完全主流化的最后一道关卡。

但资本、技术与社会情绪合流的大势已不可逆。人们以为自己在预测未来,却没意识到,这场全民下注的热潮,本身就成了这个时代最真实的写照。

作者:Bootly


Twitter:https://twitter.com/BitpushNewsCN

比推 TG 交流群:https://t.me/BitPushCommunity

比推 TG 订阅: https://t.me/bitpush

说明: 比推所有文章只代表作者观点,不构成投资建议

Leituras Relacionadas

Claude Accused of Becoming Dumber by the Entire Internet, Anthropic Steps In to Reveal: It’s Not the Model That’s Tricking You

When users complained that Claude was "getting dumber," the root cause wasn't the AI model itself. In an official blog post, Anthropic clarified the critical difference between two key settings in Claude Code: Model and Effort. Model refers to the core "brain"—the fixed, trained weights of a specific AI (like Sonnet, Opus, or Fable). Changing the Model addresses *capability* ("can it do this?"), but its knowledge is static post-training. Effort, however, controls the AI's *approach and thoroughness* for a specific task. A higher Effort level instructs Claude to read more files, run tests, perform verification, and complete multi-step reasoning before responding, significantly increasing its "work output" for that job. Conversely, low Effort leads to quicker, less thorough replies. This distinction explains the March 2024 uproar where users experienced a sudden drop in Claude's performance. The cause was not a model change but Anthropic quietly lowering the *default* Effort setting from "high" to "medium" to reduce latency, which was later reverted. The key insight is that a smaller, capable model (like Sonnet) on high Effort can often outperform a larger, more powerful model (like Opus) on low Effort for many tasks. The article provides a practical troubleshooting framework: if Claude makes an error, first check the context and instructions. If it seems to skip necessary steps or validations, increase Effort. If it diligently attempts the task but fails conceptually or makes consistent factual errors despite good context, then consider switching to a more capable Model. The takeaway is a shift in focus: effective AI programming is less about always choosing the "strongest" model and more about intelligently *orchestrating* models and effort levels—acting like a project manager to assign the right "brain" with the right level of diligence for each job, optimizing both results and cost.

marsbitHá 28m

Claude Accused of Becoming Dumber by the Entire Internet, Anthropic Steps In to Reveal: It’s Not the Model That’s Tricking You

marsbitHá 28m

Will the Ethereum Foundation Evolve into a 'Mascot'? Diversified Organizations Are Fragmenting Its Functions

The Ethereum Foundation (EF) is undergoing significant internal turmoil and functional erosion. Following its largest-ever layoff of 54 staff (20% of its workforce) and a major organizational restructuring announced in June, its Protocol Support Team has been officially dissolved. This comes alongside the high-profile resignation of key figures like co-executive director Xiaowei Wang, bringing senior departures this year to at least eight. Criticism of EF's rigid structure, opaque decision-making, and perceived lack of a clear value narrative for ETH has intensified within the community. The layoffs have catalyzed the emergence of independent, non-profit organizations like Ethlabs and Ethereum Institutional, founded by former EF researchers and members. These entities are now taking on core functions such as protocol research/development and institutional adoption, effectively fragmenting the EF's traditional leadership role. Concurrently, EF's security team is adapting to technological change, deploying specialized AI agents to audit Ethereum's codebase, which successfully discovered a critical vulnerability (CVE-2026-34219). While EF states AI complements rather than replaces researchers, it signals a potential future shift in its operational model. Faced with these challenges—internal restructuring, talent drain, the rise of competing organizations, and AI integration—the Ethereum Foundation appears to be stepping back from a central commanding role. Analysts and community observers speculate it may increasingly transition towards a symbolic "ecosystem mascot" function, while decentralized initiatives drive Ethereum's future growth and institutional adoption.

marsbitHá 54m

Will the Ethereum Foundation Evolve into a 'Mascot'? Diversified Organizations Are Fragmenting Its Functions

marsbitHá 54m

Nearly a Hundred Players Rush into Embodied Data: With 4.47 Billion Yuan in Financing in One Year, Who Can Really Make Money by 'Selling Data'?

The domestic embodied AI data industry has attracted nearly 100 players, with 70 focused on data collection and 27 on data infrastructure. In the past year, 15 independent embodied data service providers raised approximately 4.47 billion yuan. Despite this growth, the sector remains early-stage, fragmented, and faces significant challenges. Data collection methods are diverse, categorized into four main routes: teleoperation of real robots, human demonstration without a robot (using motion capture, exoskeletons, etc.), simulation synthesis, and distillation from internet videos. Most companies (43%) adopt hybrid approaches, combining multiple routes, as no single method can meet all training needs. Teleoperation alone is pursued by 31% of players, often by state-owned platforms and robot companies, while newer firms favor asset-light, no-hardware human demonstration. Independent data service providers now form the largest player group (40%), indicating the emergence of a distinct industry segment rather than just a subsidiary function for robot makers. Two-thirds of all players are "embodied-native" startups, while one-third are companies that pivoted from fields like AI data annotation, which are more prevalent in the data infrastructure layer. Current annual industry capacity is estimated at 1.6-1.8 million hours plus 70-80 million data points, with a short-term goal to increase this 15-20 fold within 1-3 years. Data collection factories are spread across 20 provinces in China, concentrated in the Yangtze River Delta, Beijing-Tianjin-Hebei, and Pearl River Delta regions. Financially, the 4.47 billion yuan raised in the past year pales compared to the 43.8 billion yuan raised by the broader embodied intelligence sector in just the first half of 2026, highlighting that data remains a less "sexy" bet for investors. The 15 funded independent providers show clear stratification: a top tier led by a unicorn (Lightwheel Intelligence, 3.1 billion yuan), a middle tier of 11 firms raising tens to hundreds of millions, and an early-stage tier of 3 companies. Sixty-nine investment institutions have participated, but none have made concentrated bets, reflecting uncertainty about viable business models. Over half of these funded companies are less than a year old, most are at pre-A or A rounds, and profitability remains largely unproven. In summary, the embodied data industry has become an independent track creating jobs and local economic activity. However, it is still nascent, with unformed consensus, unsolved problems, and unproven business models. The coming 1-2 years will be a critical validation window to see if companies can build sustainable, profitable businesses purely by "selling data."

marsbitHá 3h

Nearly a Hundred Players Rush into Embodied Data: With 4.47 Billion Yuan in Financing in One Year, Who Can Really Make Money by 'Selling Data'?

marsbitHá 3h

Trading

Spot
活动图片