#TradeyAI Closed Beta is over. Public Beta is open. 🏆
Here's how our champion won 1,000,000
$COS straight from their experience in the Closed Beta Live Trading competition. ⬇️
-
How I won 1,000,000
$COS in the TradeyAI trading competition:
I thought I was testing a Trading Agent.
Turns out, over 3 weeks, what I was really testing was three things: Reading the ruleset. Execution discipline. And just how bad humans are at sitting on their hands.
I ended up winning the
@TradeyAI closed beta.
But the honest picture before that win:
- 8 entries and exits.
- $8 burned in fees.
- Ranking dropped from 2nd to 4th.
- Almost took myself out of the game entirely.
That's when it hit me:
In a flat market, constantly entering and exiting is slow suicide.
--- The journey ---
I saw the
@TradeyAI closed beta competition and signed up. Roughly April 20 to May 11 — 3 weeks.
First week I was in Hong Kong. Started conservative, nothing aggressive. Then I started trying everything, looking for alpha.
Week 1 — Agent #1
Chased US equity narratives —
$NVDA /
$TSLA perps and related plays. Performance was flat. Couldn't catch good entry/exit timing. Too tied to the broader market.
Week 2 — Agent #2
Started testing altcoins —
$HYPE /
$SOL /
$BNB. Realized fees are the real final boss. Positions that looked promising got their profits ground away.
Week 3 — Competition week
Went back to
$BTC and a disciplined approach. Going into the last day, I was still sitting in 2nd place.
That's when I finally read the structure clearly:
5-person format. Winner takes all. 2nd place and below gets nothing.
In that kind of setup, "staying safe and not blowing up" earns you nothing.
A steady 2nd place finish is the same result as getting liquidated.
So on the last day, I made a decision that broke every rule of healthy trading:
- 30x BTC long
- 100% NAV
- No stop-loss
- Take-profit at $15 only
Because all I needed was to finish ahead of 2nd place after fees. That's it.
Execution didn't go as clean as planned.
After opening, BTC went basically sideways for hours. Market like a dead pond.
I got the itch.
Started wanting to "optimize" the Guide.
Have the Agent break into partial fills.
Reduce slippage.
Make it smarter somehow.
Claude stopped me from making impulsive moves at least 3 times during our strategy discussions.
Then I started opening and closing. Opening and closing.
8 total entries and exits.
$8 gone for nothing.
Ranking fell to 4th.
The strategy wasn't the problem. The execution frequency was. No volatility, do absolutely nothing.
Eventually I stripped the Guide down to 3 hard rules:
- One job only: close when $15 is hit
- Ignore every other signal
- Nothing else is allowed
I was betting on a big move sometime overnight. The risk/reward felt right — risk $100 to win $1,000, or go to zero. Fine. Let's go.
Set the instruction. Went to sleep.
04:30 AM — BTC flash crashes.
NAV dropped from $110 to $92.
If I'd been awake managing it myself, I would've panic-closed 100% of the time and finished 5th.
But I was dead asleep.
And the Agent was locked by its rules.
It couldn't make random moves. It could only wait for the TP.
06:00 AM — BTC rebounded to $82,460.
TP triggered. Agent locked in the profit automatically.
Woke up to:
14.25%. First place.
---
3 things I took away from these 3 weeks:
1. Understand the game first, then decide your risk.
I spent the first 2 weeks playing "healthy, disciplined trading." But that closed beta ruleset didn't reward healthy discipline at all.
In a winner-takes-all game, the first question has to be:
What is this game actually rewarding?
2. No volatility means every entry is just paying fees.
I made that mistake 8 times. Lost $8. Dropped to 4th.
Not a strategy issue. Pure impulse. I iterated from v1 to v8 purely because I was too antsy to do nothing.
3. Build a system that executes correctly even while you're asleep.
That's where an AI Agent actually proves its value.
It didn't start second-guessing itself when NAV dropped to 92. It didn't scramble to change strategy when the ranking fell to 4th. It just executed the rules written into the Guide.
The catch: your rules need to be clear enough to hold.
Thanks to
@contentosio for running this — took home the prize, and honestly learned more about how to communicate strategy to an AI Agent than I expected.
---
Worth noting: the new Public Beta ruleset has already closed the door on my all-in approach.
@TradeyAI redesigned it around 5 Master Awards:
Master of Precision: Risk control
Master of Discipline: Capital efficiency and consistency
Master of Opportunity: Capturing market moves
Master of Resilience: Recovering from drawdowns
A whole different level.
It no longer just rewards whoever is willing to go all-in on the final day.
It rewards people who can genuinely write good Agent Guides, design sound capital management, and let AI execute stably.
My closed beta winning strategy is probably best kept as a cautionary tale from here on lol.
---
Public Beta is now open:
- Registration deadline: May 18
- Agent setup deadline: May 19
- Live trading competition: May 20 – June 2
Requirements:
- Hyperliquid account with ≥ 100 USDC
- Bind API key for live trading
Real prizes. Real trades. Check the official announcement for the registration link.
#TradeyAI #Contentos #COS #TradingAgent
交易賽拿第一,擼了一百萬顆
$COS 的心得:
原本以為自己是在測 Trading Agent。
後來才發現,這 3 週真正測的是三件事:
賽制判斷、執行紀律、以及人類到底有多容易手癢。
最後拿了
@TradeyAI 封測冠軍。
但冠軍前的真實狀態是:
進出 8 次。
磨損 $8。
排名從第 2 掉到第 4。
差一點直接把自己玩沒。
也是那時候我才悟到:
在沒波動的市場裡,頻繁進出就是慢性自殺。
---歷程---
這次看到
@TradeyAI 有封測活動就去參加。
時間大概從 4/20 開始,到 5/11 結算,總共 3 週。
前面一週我人在香港,一開始用保守策略,沒有特別 aggressive。
中間我試了一堆方法,想找 Alpha。
第一週:Agent 1 號機
主要找美股相關標的,例如
$NVDA /
$TSLA 合約跟敘事。
結果績效平平,可能剛好沒抓到好的進出點,整體還是跟大盤連動太高。
第二週:Agent 2 號機
開始測小幣,像
$HYPE /
$SOL / $bnb。
結果發現手續費才是魔王。
看起來有機會,實際打下去,獲利很容易被磨掉。
第三週比賽:回歸
$BTC 紀律型策略。
結果到最後一天,我還是只有第 2 名。
這時候我才真正看懂這個賽制:
5 人賽。
Winner takes all。
第 2 名以後都沒有獎金。
在這種結構下,「保守不爆倉」本身不會被獎勵。
你穩穩地輸,跟爆倉出局,結果一樣。
所以最後一天,我做了一個違反所有健康交易紀律的決定:
30x BTC long
100% NAV
不停損
只設止盈 $15
因為只要扣完手續費後贏過第 2 名,我就拿冠軍。
問題是,執行沒那麼順。
開倉後幾個小時,BTC 幾乎橫盤,市場像死水一樣。
我開始手癢。想說100u隨便衝!
想「優化」Guide。
想讓 Agent 拆單分批。
想降低滑價。
想讓它更聰明一點。
中間跟claude溝通策略它阻止我至少3次,不要我衝動開單
然後我開始開了關、關了開。
一共進出 8 次。
白白磨掉 $8。
排名掉到第 4。
策略再好,執行頻率錯了,一樣會輸。
中間沒波動,就什麼都不要做。
最後我把 Guide 改成 3 條死規則:
只做一件事:賺到 $15 就關倉
其他訊號全部忽略
其他什麼都不准做
賭凌晨到早上會有大波動,算了盈虧比不錯,100u賭1000u,或者100u歸零,衝吧!
下完指令就去睡了。
凌晨 04:30,BTC 急跌。
NAV 從 110 掉到 92。
如果我那時候醒著,自己主觀操盤,我大概 100% 會恐慌平倉,然後直接變第 5 名。
但我睡死了。
Agent 也被規則鎖死了。
它不能亂動。
它只能等 TP。
06:00,BTC 反彈到 $82,460。
TP 觸發。
Agent 自動鎖利。
醒來看結果:
14.25%。
拿冠軍。
------
這 3 週,學到 3 件事。
第一,先看懂賽制,再決定風險。
我前 2 週一直在做「健康紀律交易」。
但那個封測賽制根本不獎勵健康紀律。
Winner takes all 的局,要先問的是:
這個遊戲到底在獎勵什麼?
第二,沒波動的時候,進出就是磨損。
我自己犯了 8 次。
賠了 $8。
排名掉到第 4。
不是策略問題,單純手癢。
中間策略我從v1迭代到v8 都是手癢想衝
第三,設計一個你睡著也會照規則執行的系統。
這才是 AI Agent 最有價值的地方。
它不會因為 NAV 突然掉到 92 就開始腦補。
它不會因為排名掉到第 4 就亂改策略。
它只會照你寫進 Guide 的規則執行。
前提是,你的規則要夠清楚。
感謝
@contentosio 辦這個活動讓拿了獎金,也學到一些經驗!尤其是我跟ai對話怎麼下prompt的歷程。
---
有趣的是,新一輪 Public Beta 賽制已經改了。
@TradeyAI 直接把我這種 ALL-IN 梭哈打法明文禁止。
新賽制變成 5 個 Master Awards:
Master of Precision:風控
Master of Discipline:資金效率與紀律
Master of Autonomy:AI 自主執行
Master of Opportunity:捕捉機會
Master of Resilience:回撤後翻盤
這個賽制層次就高很多。
它不再只獎勵最後一天敢梭哈的人。
它開始獎勵真正會寫 Agent Guide、會設計資金管理、會讓 AI 穩定執行的人。
我那套封測冠軍打法,以後大概只能當反面教材了 lol
⸻
下一輪 Public Beta 已開放報名。
官方資訊:
報名截止:5/18
Agent 設定截止:5/19
實盤競賽:5/20 - 6/2
需要 Hyperliquid 帳戶 ≥ 100 USDC
需要綁 API Key 實盤打
Master Awards 總獎池加上策略貢獻獎、社群分享獎,獎勵是真的,交易也是真的。
報名連結可以看官方公告。
#TradeyAI #Contentos #COS #TradingAgent