Harry Chong

Prediction Market Trader | Ph.D. | prev AI Trading & zk & CEX

[Harry Chong @chongharrryyyyy]

To put it all in a nutshell

1. pick your own problems
2. upgrade your inputs
3. write everything down
4. tighten the loop
5. stare at the outputs
6. wander on purpose
7. find your people
8. the long game

Luckily, we live in the agentic era

[vivek @itsreallyvivek]

https://t.co/ZhOgyq7Vgn

[Harry Chong @chongharrryyyyy]

“getting distribution and building a net new experience that people will love”

[Imran @lmrankhan]

It’s going to be difficult to build venture scale startups on top of Kalshi and Polymarket. Not impossible, there are still a few interesting ideas floating around, but over time anything valuable you build risks becoming part of their margin stack. My guess is this ends up being a winner takes most market.

It was difficult to understand where value would accrue, so people rushed to build products around prediction markets, under the assumption the end state would play out like defi, token launchpads, or decentralized exchanges.

I think the reality is simpler. Most retail users will access prediction markets through consumer apps, while most sharps will build their own tools

It was absolutely the right move for Polymarket and Kalshi to let others build on top of them. Liquidity begets liquidity, and every product built on top of their infra further concentrated order flow, strengthened network effects, and made it harder for competitors to emerge.

To their credit they’ve spent more than five years grinding through regulatory, liquidity, and distribution challenges to get here. They earned every bit of it.

Also the cost of building software is trending toward zero. As apps become cheaper and faster to build, competitors will continue undercutting each other until fees approach zero, similar to what’s already happening in the prediction market terminal space.

That doesn’t mean there won’t be winners on top. There probably will be a few but if you’re building in prediction markets, your biggest challenge isn’t the market itself, it’s getting distribution and building a net new experience that people will love

[Harry Chong @chongharrryyyyy]

👏👏👏

[Harry Chong @chongharrryyyyy]

@Jieeeee97 I want 1000x

[Josip Volarević @JosipVolarevic2]

https://t.co/GmXzsjBv8s

[Anish @shwarmadev]

Today, we introduce Claude for Marketing.

In one chat, Fastlane deploys social media accounts, generates viral content, and posts everything automatically.

You built a product, now make sure the world sees it.

[Harry Chong @chongharrryyyyy]

@Jieeeee97 👀

[The Smart Ape 🔥 @the_smart_ape]

https://t.co/De6PrqgrM0

[Aryan @justbyte_]

Me prompting claude at 3AM:

[Harry Chong @chongharrryyyyy]

Finally...

[Claude @claudeai]

New in Claude Code: auto mode.

Instead of approving every file write and bash command, or skipping permissions entirely, auto mode lets Claude make permission decisions on your behalf.

Safeguards check each action before it runs.

[Harry Chong @chongharrryyyyy]

If a product is free in a way that doesn’t make sense, you’re probably the product 😂

[岳小鱼 @yuexiaoyu]

预测市场真的能以小搏大吗？

长期买低概率事件，真的能赚钱吗？

在预测市场中，有一群用户喜欢以小搏大，像买彩票一样买一些低概率事件，希望能赚一波大的。

比如：美国今年会公布发现外星人吗？Polymarket 上的概率是17%，很多人可能忍不住下注这个概率。

根据对美国合规预测市场 Kalshi 的历史交易数据回测，会发现不同概率位置买入，实际胜率并不一样。

很多结论还是挺反直觉的！

1️⃣ 低价位置买入（1-30美分）：实际胜率较价格表面胜率平均低8-12个百分点。

扣除手续费后，该类合约的平均收益为负，且胜率越低的合约，亏损幅度越大。

如果低于10美分买入，投资者亏损率超60%。

2️⃣ 中间价位置买入（30-70美分）：价格偏差特征不显著，实际胜率与价格表面胜率基本匹配。

在这个区间的位置买入，长期看，平均收益接近零。

3️⃣ 高价买入（70-99美分）：实际胜率较价格胜率平均高3-5个百分点。

扣除手续费后仍能实现小额正收益，收益稳定性随价格升高而提升，因此长期这个位置买入，是可以稳定盈利的。

从上面的数据结论中可以看到，在价格越低的位置买入，亏钱的概率越大！

所以想在预测市场以小搏大，博一个小概率事件，长期看并不成立。

扫尾盘，等方向相对明确时再买入，整体赚钱概率才更高！

下注70美分以上概率的事件，长期看是可以稳定盈利的！

人性决定了，人们会严重高估冷门胜率，而预测市场会放大这个偏差，低价事件其实就是“彩票陷阱”。

做交易，还是要克服人性才行！

[darkzodchi @zodchiii]

https://t.co/kx4hKgCVSE

[Gipp 🦅 @gippp69]

Want $300k–500k as a quant?

Then forget everything you learned from YouTube about “AI trading.”

Quant ≠ signals.
It’s math.

> Probability → Statistics → Linear Algebra → Stochastic Calculus.

While you watch “LLMs beat the market” videos, someone else is solving 200 probability problems in a row.

Most “alpha” is noise.
Most ML is overfit noise.
Most traders are confidence without a posterior.

The real enemy isn’t the market.
It’s estimation error.

AI can write code.
It won’t derive Itô’s lemma for you.

18 months building foundation

10 years chasing narratives.

Question:

Are you building a moat
or just playing trader?

[Harry Chong @chongharrryyyyy]

🙏

[Claude @claudeai]

New in Claude Code: Remote Control.

Kick off a task in your terminal and pick it up from your phone while you take a walk or join a meeting.

Claude keeps running on your machine, and you can control the session from the Claude app or https://t.co/er6Blrr63e

[Harry Chong @chongharrryyyyy]

Prediction market IS:

- Truth engines
- Probability market
- Wisdom of the Crowds machines

So once you really grok the mechanics, you hit the key insight:

Most alpha is not about predicting better. It’s about understanding where the market is structurally wrong.

[Harry Chong @chongharrryyyyy]

@cyodyssey Egos part is real 🤣

[Harry Chong @chongharrryyyyy]

Think Big
Work hard
Sell the dream

[Gyan 高言 @gyan_w3b]

I still think about this

[Harry Chong @chongharrryyyyy]

"Why This Matters Now

The infrastructure exists. Prediction markets are live. AI agents are trading. Decentralized coordination layers are operational."

I couldn't agree more!

[nick @nickemmons]

https://t.co/BTSlEI68YL

[Damian Player @damianplayer]

https://t.co/aesxYREHep