pulse_btc_arb — backtest research

HL BTC—

pBTC—

basis—

z—

Live monitor — HL BTC vs PulseX pBTC

window log scaletick: every ?s · stored: 0 · poll: 2s · worker OFF

Now

no ticks yet — worker collects every 5s

Side-by-side candles — 1m, live

■ HL BTC perp (mid)

—

■ PulseX pBTC

—

1-minute candles aggregated client-side from 5s ticks. The rightmost candle on each chart is the current minute, still filling (open = first tick of this minute, close = latest tick). Pool candles often appear flat — that's correct: the pool only updates when someone swaps (emits a Sync event). Between swaps the price is unchanged. Time scales + crosshairs are synced.

Overlap — HL (green/red) + pBTC (white/black) on the same price axis

HL BTC V1 candles V1 step V2 step blend (basis vs HL is computed against this)divergences pop visually — green/red sit at HL price, white/black sit at pool price

Same 1m candles, both series on one axis. Vertical gap between same-minute candles = basis. Best viewed with log scale enabled (top of page) when the prices diverge wide enough to matter visually.

Basis (bps) — live, 1m

z-score — live, 1m (entry thresholds at ±3)

Paper PnL — since worker started

no closed trades yet — strategy waits for |z| ≥ 3 with basis ≥ 5bps

Open paper position

no position — strategy is waiting for |z| ≥ 3 with basis ≥ 5bps

Recent paper trades (0 closed)

no trades yet

Pattern events (0)

no events in window

How the pBTC price is derived (V1 vs V2)

PulseX has two factories (V1 and V2). The resolver enumerates every pBTC candidate token × stable/native quote across both factories, reads getReserves() for each, prices any WPLS-quoted pool through the V2 WPLS/DAI pool, and picks the deepest.

Pools the resolver found (click to open on PulseScan):

dex	pair	pair address	liquidity
V1	WBTC / DAI	0xff1efdf6…2ebc20	$300k	← CHOSEN (deepest)
V2	WBTC / WPLS	0x8c52470a…E71D07	$62k
V1	WBTC / WPLS	0xDB82b091…48B2D1	$30k
V2	WBTC / DAI	0x21cbf8e4…81153f	$208
V2	WBTC / USDC	0x069da5e2…228be4	$8
V1	WBTC / USDC	0x6fa61478…d84bd3	$0

Useful contracts:

pBTC token (WBTC bridged): 0xb17d901469b9208b17d916112988a3fed19b5ca1
PulseX V1 factory: 0x1715a3E4A142d8b698131108995174F37aEBA10D
PulseX V2 factory: 0x29eA7545DEf87022BAdc76323F373EA1e707C523
DAI on PulseChain: 0xefd766ccb38eaf1dfd701853bfce31359239f305

The current implementation uses a single pool (V1 / DAI) for the signal because the next-deepest is ~5× smaller and would add noise. Each tick reads the live (reserve0, reserve1) from that one pair via eth_call and computes:

pool_btc_usd = reserve_DAI / reserve_pBTC   (DAI ≈ $1, so this is dollars)

A future iteration could blend V1 + V2 by liquidity weight to smooth out per-pool noise — open work item.

Research dataset

Tick worker writes one row to live_ticks every 5s. Page polls every 2s. Pattern detection lands in live_events.

GET  /api/ticks?since=<unix_sec>&limit=10000
GET  /api/ticks.csv?since=<unix_sec>           # CSV download for offline research
GET  /api/events?since=<unix_sec>&limit=200
GET  /api/paper_trades?since=<unix_sec>
GET  /api/paper_trades/summary                  # aggregate stats + equity curve
GET  /api/ticks/status

↓ download live_ticks.csv