Three engines. One pool. Every win buys $WIRE.
A trading pool on Hyperliquid, seeded by the dev and grown by creator fees. WIRE buys market flushes, farms funding gaps between venues, and snipes exchange listings. A share of every winning trade buys $WIRE back.
- Flush trades won
- 69% of 89
- Funding months positive
- 12 / 12
- Listing snipes won
- 84% of 19
Paper run. Live data, no real money.
All three engines on live market data, sized from a 100 SOL pool. Highlighted rows are trades taken since the paper run started. Earlier rows are the same rules over the last 30 days.
The bot itself.
The run above rebuilds the rules from public data. This is the bot on the Tokyo server, trading the same rules on paper against live order books.
The trading engines. Each one tested out of sample.
Each rule was picked on the first half of a year of data and then run unchanged on the second half. Costs and slippage are included. These three held up.
-
Flush bounce
When BTC drops 1% in five minutes, WIRE buys any coin down 5% or more in the same window and sells into the bounce. A 5% target at 3x returns +15% on margin.
Flush bounceHyperliquid · 3x69%of 89 trades wonAverage trade+0.81%Worst day at 3x−31.0%Exit+5% · −10% · 60 minSize20% · 3x · first 5 a day -
Funding spread
Short the perp where funding is high, long the same coin where it is low. Price moves cancel. The difference is paid out every hour.
Funding spreadTwo venues · 2x12 / 12months positiveYearly yield at 2x+16.8%Train / test half9.1% / 9.5%Enter / exit20%/yr gap · gap flipsCoins watched41 · 3 held max -
Listing snipe
Buys a coin seconds after an Upbit or Binance listing. Only fires when it gets there first: if the price has already moved 5%, it skips. Out within five minutes.
Listing snipeHyperliquid · 5x84%of 19 trades wonMedian trade+10.4%Notices tested108 since 2024Exit−10% stop · 5 minSize10% of pool · 5x
Flush engine on a 100 SOL pool. One year, every signal.
Every signal from Oct 2025 to Oct 2026, first five a day, costs included. Ideas that looked good and failed out of sample are listed beside it. None of them run.
Engines tested and cut.
Speed decides listings. The move lands in two seconds.
Median price path of 108 Upbit KRW listings and Binance futures launches, Apr 2024 to Oct 2026, on one-second Binance data. The move lands in the first two seconds, then holds.
Listing delay and trade return.
Same 108 notices, bought at different delays and sold after five minutes or at a 10% stop. Fees and 0.3% slippage per side included.
Listing entry filter.
If the price has already moved 5% by the time the order is ready, the move is gone. WIRE skips those and only takes the ones it reached first.
Listing replay. 108 real notices.
Each row is a real listing notice and what WIRE's rules would have done with it one second later, priced on Binance's one-second data. Click any row.
A listing trade. Notice to exit in five minutes.
Target timings from a server in Tokyo. Paper mode records every one of these on live order books before real money is used.
- 0 ms
Notice goes up
Upbit or Binance publishes a listing.
- 1–3 s
WIRE sees it
Upbit is checked every second, Binance every 3 seconds.
- + 5 ms
Rules check
Seen within 4 s, perp exists, price up less than 5% in 10 s, limits clear.
- + 150 ms
Order fills
Immediate-or-cancel buy on Hyperliquid, 3% max slippage.
- + 200 ms
Stop is set
Exchange-side stop 10% under entry.
- 5 min
Out
Closed at market. Every trade posted with the notice that caused it.
Rules and pool. Every number is in the config.
The dev seeds the pool with his own SOL at launch. Creator fees from $WIRE trading grow it. A share of every winning trade buys $WIRE on the open market.