一树🌴.𝟎𝐱𝐔

2025 All in 预测市场 | 打磨产品，偶尔开发，暂居美东 | Cofounder @0xUclub🇭🇰 | Prev @Alibaba ByteDance @Johnshopkins

[一树🌴.𝟎𝐱𝐔 @issue_wong]

🚀我把Predict-Raven开源了，这是一个可以在预测市场上自主交易的Agent框架。 在一个月的时间里他已经取得了实盘+8%的收益。我身边一小撮对AI trading和Polymarket感兴趣的极客和交易员朋友，已经开始部署和测试它

Predict-Raven有点像一只盘旋在上空持续寻找机会的小乌鸦，它会定时扫描数千个市场，持续地搜集证据，评估事件真实发生概率，再对比市场隐含的赔率。当它发现足够大的偏差时，就会自动下单、风控、止损，并记录下来自己的决策思路和依据

我做这个项目的动机源自主观性较强的几点判断：
1. AI已经拥有了和人类相等甚至超过人类的推理能力
2. 交易，在本质上可能和其他推理任务并没有不同，AI缺少的，可能只是更好的交互环境和上下文
3. 预测市场在交易里又尤为特殊，他们交易的，本质上是主观预估概率vs市场隐含概率的gap。如果你能比市场更准确地预测未来，你就能持续地获得收益。而预测本身，可能可以拆成两个相对独立的能力：信息搜集和推理。ai在一系列高难度benchmark的进步表明，ai在agentic search和reasoing上终会战胜人类，或者，已经战胜

人类的优势是直觉和通过社交来获得隐含信息。再退几步说，哪怕AI的推理能力仍然稍弱于人类，也能通过同时扫描复数个市场和7x24的待机，来等候定价错误的市场出现实现盈利，战胜市场某种程度上也是胜出人类的baseline。就像在德扑中：牌技高低只是盈利能力的一部分，更重要的是“找局”能力，能不能稳定地找上有漏洞的对手。以盈利为目的时，永远要在自己有edge的场子玩。

预测市场，尤其是科技和政治市场，是一个还没被充分开发的蓝海，当做市商还在畏惧库存管理和逆向选择时。agent真的有可能跑出一些优势；同时我也相信：开源协作会比闭门造车更快跑出东西，对于 Predict-Raven 来讲，还非常有趣的一点是，它也有类似龙虾 soul.md 的架构，允许用户按照自己的偏好和设定去交易。

举个例子，我经过测试并打磨的一个用例是告诉 Agent：市场存在 long-shot bias，即市场会高估小概率事件发生的概率。同时，我让它限定在流动性非常差的政治和科技市场，因为我相信这些地方被不正确定价的概率更高，竞争也不激烈。大家可以很自由地把自己的交易原则代入到小乌鸦里，并通过一些参数增加强约束

哪怕这已经是一个经过很多轮迭代的项目，我也想向大家坦白这个项目还有非常多可以改进的地方。比如真实的测试资金只有500u，而预期的策略容量可能在15万美金以上。目前的一些最优先的需求还包括：更高频的交易，持久化运行和云端部署。我觉得现在也是一个合适的时机把它发布向公众，通过building in public收获非常珍贵的反馈，来调整后续的开发方向

我会挑选五位转发并私信我的朋友，可以简单介绍下自己想参与到Predict-Raven的原因，我将手把手地帮你们搭建属于自己的预测市场agent.

repo，实盘地址和持仓、决策分析网站如下

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@KKyleChen 佛的脚真太难绷了

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@EudemoniaCC 早日康复 下次再来！

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@love_u_4ever @defioasis 现在已经有了，来看这个看板 Poly combo 数据: https://t.co/dCM1Xmxjuo

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@tianyi 孔某人的低维认知 这个公众号里也有很多关于harness对比的好内容 https://t.co/GXxCyZXNJk

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@__o_o__O_o_O_O_ 牛啊 我同样让fable跑一个九人桌的训练推荐，和你这一比简直就是个渣子

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@sarahshine007 I concurred that.
Strongest model should aim to beat baseline by 5-10% but not "predict it right every time". Still you might be able to gain profit in place like polymarket and kalshi. e.g. you buying a market with implied possibility of 20% but your model says 35%.

[一树🌴.𝟎𝐱𝐔 @issue_wong]

Predict-Raven is an open-source forecasting agent framework.
We just shipped a sprint update for the World Cup: champion, group stage results and knockouts
Yes — it predicts 🇪🇸 Spain to win it all, with reasoning traces and sources fully public. DM me for free trial access 🎆

[一树🌴.𝟎𝐱𝐔 @issue_wong]

Victor, I agree with you in some sense. There are indeed many classical ML method to predict sport outcome. In predict-raven, we use elo rating with bayesian updates to predict. https://t.co/4U3ygOUYnE

In a sense, this website give us a base elo rating. while LLM use news (player status or team style) and other factors to update elo.

I also wanna highlight that coding agent can run any ML with sufficient data. Future forecasting harness will be a hierarchy that LLM call ML model as methods.

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@CS_Unemployment lmao, that's funny.

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@0xdopo a little bit, surprise to find out AI thinking this way :)

[一树🌴.𝟎𝐱𝐔 @issue_wong]

https://t.co/D0BeG5ET5A

[一树🌴.𝟎𝐱𝐔 @issue_wong]

check out our website for detailed forecasts
https://t.co/VLCHEsuM2D

[一树🌴.𝟎𝐱𝐔 @issue_wong]

今天token消耗应该是差不多干到1B了 claude + codex + manus. 用上dynamic work flow后甚至压力不算太大.

[一树🌴.𝟎𝐱𝐔 @issue_wong]

我当前的TPD即token per day大概是120M

[Brian Huang @brianryhuang]

There's a few instances of scientific heroics I've seen from seniors on Gemini where, given some issue in the stack (say, model behavior or training dynamics), they find and prove out a fix that seems quite distant from the original bug but works well and is elegant

I think that's a hallmark of research skill at the very top: the space of solutions for some modeling issue is often very vast, and making 🚀 progress relies on some ability -- hard-fought intuition over many years; System 2 transformed to System 1, I imagine -- to both know the full vastness of the search space, and then bisect with speed and high signal at each step to arrive at an answer

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@patpcj like ur work. I will try to get inspiration and integrate feature from this fantastic work into my opensourced forecasting agent project, to provide better search

[Marc Andreessen 🇺🇸 @pmarca]

There are no casual meetings with VC’s.

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@tianyi 确实蛮好读的. 作者关于上下文工程的部分其实还挺有insights的，关于办公室政治的文风很戏谑就是了hh

[一树🌴.𝟎𝐱𝐔 @issue_wong]

@KevinZbtc 这么好的文章，怎么浏览量低了😭

[秋田散人 Mr.Akita @lnkybtc]

对于所有人为制造的流动性割裂（譬如 ID, license, entry visa, bank account, HTTP auth，VPN），商人从不反感，只会选择比别人多走几步路，加以利用。

像 GFW 这种天然的 latency 仍然是在座各位华文母语者最主流最长远的利润来源。

[一树🌴.𝟎𝐱𝐔 @issue_wong]

i18n狂喜，真希望这种技术栈普及所有内容平台，让天下没有难做的多语言

[郭宇 guoyu.eth @turingou]

很好奇推特的自动翻译需要匹配多少语言对，Grok 在这件事上一天的推理成本需要多少钱？