Joined May 2026
- Tweets 63
- Following 18
- Followers 21
- Likes 6
4 Photos and videos
Pinned Tweet
Jun 14
Just open-sourced my Polymarket trading bot.
Most "trading bot" repos promise returns and hide the data.
This one does the opposite: the README leads with "no proven
edge."
Here's why I'm sharing it anyway ๐งต
github.com/Polybotmarket/polโฆ
11
11
249
Jun 14
Just open-sourced my Polymarket trading bot.
Most "trading bot" repos promise returns and hide the data.
This one does the opposite: the README leads with "no proven
edge."
Here's why I'm sharing it anyway ๐งต
github.com/Polybotmarket/polโฆ
11
11
249
Jun 14
๐ง What's next?
The current release is intentionally limited and runs in dry-run mode only.
Early users get the opportunity to test the system, provide feedback, and help shape the next phase before public deployment.
4
1
30
Jun 14
๐ Coming soon: โ Live trading mode โ Additional proprietary strategies โ Advanced portfolio management โ Expanded risk engine โ More market coverage
Once the project is tokenized, a significant portion of the live bot infrastructure and currently unreleased strategies will be opened to the community.
Early testers won't just watch the development process โ they'll be part of it.
Stay early. Stay ahead. โก๏ธ
1
26
Jun 14
The Polymarket Bot is ready for its next phase.
I'm exploring the launch of a community token to fund development, expand features, improve infrastructure, and reward early supporters.
The mission remains the same: build the best AI-powered assistant for prediction markets.
More details soon.
๐
2
53
Jun 12
Commitment post:
My Polymarket bot is now locked โ no more parameter changes,
no more strategy pivots. It runs as-is until 100 trades.
At 100, I publish the full numbers: win rate, PnL, R:R,
category breakdown. Profitable or not.
Currently at 56. See you at 100.
1
50
Jun 10
What Claude (Fable 5) actually contributed to my Polymarket bot:
โ 1,400 lines of Python I couldn't have written
โ Caught a validator bug that silently killed 167 markets/scan
โ Talked me out of going live on 40 trades of noise
โ Told me honestly when the data said "no edge"
The code was the easy part. The pushback was the value.
1
60
Jun 10
Honest disclosure: my Polymarket bot is built with Claude
(Fable 5) as engineering partner.
AI wrote the code. I made the decisions. 40 trades in,
no proven edge โ but the build process itself taught me
more than the results did.
Building in public means sharing that too.
105
Jun 7
Update on the adaptive sizing bot:
The learner already picked up signals after 36 trades:
โ Book depth: 0.218 correlation (strongest)
โ EV: 0.204 (strong)
โ Momentum: -0.065 (hurts!)
โ Sentiment: -0.043 (hurts!)
So I removed weak categories (crypto, pop culture) and
tightened depth filter โ fully data-driven, no guesswork.
60
Jun 7
Just shipped the biggest engineering upgrade to my prediction
market bot:
Adaptive position sizing.
Same strategy, same signals โ but now the bot decides HOW MUCH
to risk based on its own confidence.
Here's how it works ๐งต
5
4
145
Jun 7
What this looks like in practice:
โ Weak signal, calm market: $1.50 (minimum)
โ Average setup, normal vol: $2.50
โ Strong signal, volatile: $5.00
โ High conviction, fat tail: $6.50
Position size now mirrors actual edge โ not gut feeling.
1
41
Jun 7
Why this matters:
Last week one position 3x'd (No 0.24 โ 0.998). $9.47 on a $3
position. With adaptive sizing, that same trade would have been
$5 โ $15 .
Capital deployment > strategy refinement at this stage.
Still in dry-run. Testing for 100 trades before conclusions.
28
Jun 3
Lesson learned the hard way today on my prediction market bot:
Opened 4 similar political positions. News dropped. All 4 hit
SL within hours. -6.49 USDC day (dry-run, thankfully).
Added max-per-category constraint. Now no more than 2 positions
in any single category.
Diversification > clever strategy.
36
Jun 1
First substantial TP on my prediction market bot:
Entry: No @ 0.24
Exit: No @ 0.998
PnL: 9.47 USDC (3.15x on a 3 USDC position)
This single trade carried this week's portfolio.
That's how asymmetric payoff works.
Win rate โ profitability when tails matter.
2
63
May 31
Just made the biggest strategy change on my prediction market bot.
Switched from "patient swing trades" to scalping.
TP from 2.0x โ 1.3x. SL from 0.5x โ 0.75x. Max hold: 12 hours.
The reason wasn't backtest results. It was something simpler. ๐งต
1
33
May 31
Realized: in research mode, cycle time matters more than per-trade edge.
Tiny win/loss with fast feedback > big win/loss with slow feedback.
You can't optimize what you don't observe.
And you can't observe what hasn't closed yet.
1
22
May 31
New parameters:
โ TP 1.30x (small wins)
โ SL 0.75x (small losses)
โ Trailing 6% (lock in fast)
โ Max hold 12 hours (forced exit)
Lower per-trade R:R, but 10x more data points per day.
The strategy isn't just about edge. It's about feedback speed.
1
16