A Polymarket player with about a 96% win rate on the main battleground: I split BipBop’s 1,632 trades into separate positions

This account on the leaderboard: BipBop

1,632 settled positions, main battleground win rate about 96%, positive for multiple consecutive months.
Most people’s first reaction is: copy him.
We actually backtested that layer that looks “like him” — turning “buy 0.85–0.95 high-probability markets, and get to settlement” into a mechanical strategy, backtesting 5,378 markets (high-probability bucket compared group): win rate 88.9%, net profit −2.5%.
Win rate is an illusion. A high win rate can be a losing strategy. The money he makes is not the money you think he’s making.
——
What I split
Pull the full history of public executed trades: 2,414 trades, match the 1,632 settled positions to the official settlement, then cut a matrix by event type × entry price range.
Total settled: cost about $300k, profit about +$58k, ROI about +19%.
No copy-trade path is provided. Only the aggregated results, and a way of reading you can use later to interpret other people.
——
Main conclusion: profit isn’t in 0.85+
The main battleground is mention/say-type events.
(Figure 1: win-rate illusion control)
(Figure 2: main battleground across four price bands)
There are only three lines of how to read it:
1. Don’t be fooled by the highest win rate. 0.85+ looks best, but it’s not a money bag.
2. Don’t be fooled by the highest ROI. <0.4 is strong, but the sample size is too small to be the main engine.
3. Find the absolute profit anchor first, then explain why it holds.
In the main battleground, absolute profit is about +$28.6K for 0.6–0.85; for 0.85+ it’s only about +$10.9K. The middle segment with fewer trades actually moved away most of the profit.
——
Why this happens
These markets often look like this: “Will someone say a certain word in some event / earnings call.”
The most intuitive hypothesis is: historical word-frequency base rates vs market-pricing by feel — the middle gets marked cheap.
I matched a batch of identifiable main-battleground positions against public transcripts: using historical word frequency to identify the earning middle segments he took, the identification rate is only about one tenth. The hypothesis doesn’t hold.
Then I aligned the execution timestamps to the event windows: a large portion of trades happen on the day of the event, even within the live-stream window. Opening locks the topic, mid-session follows the word, counting markets’ live numbers, then the word is spoken and settlement gets chased with lag — all of those traces are there.
It looks more like real-time in-the-moment information, not a statistics table computed in advance.
The map is still useful: where the money lands in each “cell,” and whether it goes mainstream for ordinary people — those don’t change.
What changes is the reading method — don’t misread middle-segment profit as “if someone has a word-frequency table, you can copy.”
This also explains the opening backtest: blindly buying 0.85–0.95, win rate can be very high, yet the expectation can still be negative; and his edge isn’t in that precomputed table.
——
Two observations you can use immediately
1) Empty-cash snapshots can mislead
These players redeem once they win. Their positions page usually sits with only tens of dollars of junk orders — I captured about $38. Don’t judge life or death by position value; look at the PnL curve.
Historical peak about +$53.8K; still about +$3.5K in the last 30 days; monthly executed trades rise from about 25 in April to about 220 in June.
No-cash is the norm, not a sign of leaving.
2) Specialization boundaries are worth more than win rate
(Figure 3: specialization vs going mainstream)
Once you leave the main category: n=53, win rate about 0.49, ROI −17.2%.
The overall table gets prettied up by the main category. If you don’t look at the boundary, you’ll overestimate the transferability of “copying an address.”
——
How to look at high-win-rate addresses from now on
Ask four questions first:
1. Which price band does the money land in?
2. When leaving the specialization category, what’s left?
3. Is empty cash due to redemption, or a real stop?
4. Are trades stacked on the event day / within the live window, or are they spread in advance?
Answering these four questions is more useful than collecting a single 96% address.
——
Go look at these markets on Polymarket / look at BipBop yourself, using this entry (invite link):

To get the full multi-category map + mechanism backtest details, on Leo Insider ($5/month):

For data research sharing only and does not constitute investment advice. No copy-trade path is provided.