Harness benchmark

Cross-Venue Funding Spread Arb (maker/taker)
Idea: Long the perp where funding is low, short where it's high; make one leg, take the other to avoid eating the spread.

- Rules: Every 1 min compute funding diff between 2 perp venues for the same asset. ENTER when annualised diff > 25% AND mark-price spread between venues < expected 3-day funding gain: post maker limit on venue A, on fill instantly market the hedge on venue B. EXIT when diff < 5% APR or after 72h, whichever first; unwind maker-first.
- Data: Real-time funding + order books on 2+ venues
- Venue / timeframe: Hyperliquid vs Lighter/Extended/Drift; 1m loop
- Krexa risk settings: maxTrade 20% per leg; daily loss 1%; close both legs if combined PnL < -1%; leverage <=2x per leg
- Main failure mode: Price spread between venues offsets the funding (author's core warning); legging risk when one leg fills and the other slips.
- Paper test plan: 3 weeks, >=20 round trips. Measure realised funding minus cross-venue spread & fees; leg-fill latency. PASS: >60% of cycles net positive, avg net > 0.5bp/h, worst leg-slip < 0.3%.

OI + Spot-CVD + Funding Trend Rider
Idea: Only ride breakouts where spot buyers lead (spot CVD up, OI up, funding positive but not extreme); exit when spot CVD stalls.

- Rules: 4h BTC/ETH. LONG when price makes 20-bar high AND spot CVD makes 20-bar high AND OI +5% over 24h AND 0 < funding < 0.03%/8h. EXIT/flip-short trigger: price new high but spot CVD lower high for 2 bars (divergence). SL below last 4h swing low; trail with 2x ATR.
- Data: Spot CVD (aggregated), OI, funding, 4h OHLC
- Venue / timeframe: Hyperliquid perps; 4h
- Krexa risk settings: maxTrade 10%; daily loss 2%; SL ~3%; leverage 2x
- Main failure mode: CVD data quality differs by provider; perp-led fakeouts that look spot-led on one venue.
- Paper test plan: 6 weeks, >=12 trades. Compare vs same breakout WITHOUT the CVD/OI filter. PASS: filter improves profit factor by >=0.3 and cuts losing trades >=25%.

Post-Flush OI-Reset Long
Idea: After an aggressive OI build-up gets flushed, go long only once OI has reset, support holds and spot CVD diverges positively.

- Rules: 1h/4h BTC. ARM when OI rose >15% in 7d. TRIGGER after OI drops >=8% within 24h (flush) AND price holds the prior consolidation low on a 4h close AND spot CVD makes a higher low while price makes equal/lower low. LONG on next 4h close; SL below flush wick; TP1 at 50% retrace of flush, TP2 at pre-flush high.
- Data: OI, liquidations feed, spot CVD, 4h OHLC
- Venue / timeframe: Hyperliquid perps; 4h
- Krexa risk settings: maxTrade 10%; daily loss 2%; SL 3-4% (below wick); leverage 2x
- Main failure mode: Second leg of the cascade - 'reset' that keeps resetting lower.
- Paper test plan: Event-based: collect >=8 flush events (backtest 2024-26 + live paper 6 weeks). PASS: avg R > 1.0, no single loss > 1.5R.

Daily Donchian-55 Breakout (Turtle) + 200EMA + ATR Trail
Idea: Classic trend following: buy 55-day highs only above a rising 200EMA, trail with ATR, no fixed TP. 4h Turtle variant for more trades.

- Rules: Daily: LONG when close > 55-day high AND close > 200EMA AND 200EMA rising (optional ADX(14) > 20). Initial SL 2x ATR(20). Trail stop at max(close - 3x ATR, 20-day low). Risk 1% equity per trade. Variant B: same on 4h with 20/55 bars (Turtle).
- Data: Daily/4h OHLC, EMA200, ATR, ADX
- Venue / timeframe: BTC/ETH/SOL perps or spot; 1D (and 4h variant)
- Krexa risk settings: maxTrade sized so 1R = 1% equity (~15-20% notional); daily loss 2%; SL 2x ATR (~6-8%); leverage 1x
- Main failure mode: Long flat periods: author shows it spent ~96% of the test below prior equity high; few trades => overfit risk.
- Paper test plan: Too slow for paper alone: run walk-forward backtest 2019-2026 + 8 weeks paper of the 4h variant (>=10 trades). PASS: OOS profit factor > 1.5, max DD < 25% at 1% risk.

RSI(14)<30 Dip-Buy, Time-Capped
Idea: Buy BTC when daily RSI drops below 30, exit on RSI recovery >50 or after 14 days. Do NOT add a 200DMA filter (author showed it killed returns).

- Rules: Daily BTC/ETH. LONG at close when RSI(14) < 30. EXIT when RSI(14) > 50 or 14 days elapsed. Catastrophe SL -12%. One position at a time.
- Data: Daily OHLC, RSI
- Venue / timeframe: Spot BTC/ETH on Jupiter/Base DEX; 1D
- Krexa risk settings: maxTrade 20%; daily loss 3%; SL 12%; leverage 1x
- Main failure mode: Catching knives in bear markets (author's separate dip-buy run nearly wiped out with aggressive sizing).
- Paper test plan: Few signals: backtest 2018-2026 + run 1h-RSI variant in paper 4 weeks (>=15 trades). PASS: PF > 1.4, avg win/avg loss > 1.5. Author-claimed daily backtest: +12.44%, PF 1.90, 51.85% WR.

15m RSI Mean-Reversion Scalper
Idea: On 15m BTC, momentum loses; RSI extremes revert: long RSI<30, short RSI>70, hold ~1 bar.

- Rules: 15m BTC perp. LONG when RSI(14) < 30 at candle close; SHORT when > 70. EXIT at next 15m close (or 2 bars max). Only trade if spread+fees < 0.04% round trip (use maker entries). Skip during top-of-hour news windows.
- Data: 15m OHLC, RSI, live spread
- Venue / timeframe: Hyperliquid BTC perp (maker); 15m
- Krexa risk settings: maxTrade 5%; daily loss 1.5%; SL 0.6%; leverage 2x
- Main failure mode: Edge is thin (author: 54.6% WR vs 51.8% break-even on Polymarket); fees/slippage eat it on perps if taking.
- Paper test plan: 2 weeks, >=150 trades. Track WR, avg net bp/trade after fees. PASS: net > +1bp/trade and WR > 53%.

VWMA 10/80 Crossover + ATR Trail (4h)
Idea: Volume-weighted MA crossover on BTC 4h with ATR trailing stop.

- Rules: 4h BTC. LONG when VWMA(10) crosses above VWMA(80); flat (or short, variant) on cross below. Trail stop = highest close - 1.5x ATR(14). Re-enter only on a fresh cross.
- Data: 4h OHLCV, VWMA, ATR
- Venue / timeframe: BTC perp/spot; 4h
- Krexa risk settings: maxTrade 15%; daily loss 2%; SL 1.5x ATR (~2-3%); leverage 1-2x
- Main failure mode: Parameters were optimised in-sample (overfit risk); whipsaw in ranges.
- Paper test plan: Walk-forward on 2022-2026 then 6 weeks paper (>=8 trades). Also test 8/100 & 12/60 neighbours. PASS: neighbours also profitable (robustness), PF > 1.2. Author-claimed Sharpe 2.47 / DD -2.5% - unverified.

SuperTrend / EMA-Chandelier Trend (4h)
Idea: Ride BTC 4h trends with SuperTrend flips; chandelier exit from EMA.

- Rules: 4h BTC. LONG on SuperTrend(10,3) flip to up AND close > EMA(50). SHORT on flip down AND close < EMA(50). Exit on opposite flip or chandelier stop (22-bar high - 3x ATR). Tune ONE parameter max.
- Data: 4h OHLC, ATR, EMA
- Venue / timeframe: BTC/ETH perps; 4h
- Krexa risk settings: maxTrade 10%; daily loss 2%; SL = SuperTrend line (~3-5%); leverage 2x
- Main failure mode: Chop; the author notes a single parameter change doubled Sharpe - classic overfit flag.
- Paper test plan: 6-8 weeks paper, >=10 trades + backtest with default params only. PASS: PF > 1.3 with default params.

Bull Market Support Band Regime Switch on BTC, weekly.
Entry: buy BTC (hold long) when the weekly close is above both the 20-week SMA and the 21-week EMA.
Exit: sell to USDC when the weekly close is below both the 20-week SMA and the 21-week EMA.
Stop: the regime flip is the stop. Leverage 1x, long only.

Fear & Greed + Drawdown Accumulator
Idea: Buy BTC in Extreme Fear when price is >=30% off its high; distribute in Extreme Greed at new highs.

- Rules: Daily. BUY 1 tranche (5% of sleeve) per day when F&G <= 20 AND BTC drawdown from ATH >= 30%. SELL 1 tranche per day when F&G >= 80 AND BTC within 3% of ATH. Neutral = hold. Max 20 tranches.
- Data: CMC/alt.me Fear & Greed API, daily price, ATH
- Venue / timeframe: Spot BTC/ETH; daily
- Krexa risk settings: maxTrade 5%/day; daily loss 3%; no SL (accumulation); leverage 1x
- Main failure mode: Fear can last months (early buys deep underwater); greed can persist through a long bull.
- Paper test plan: Backtest 2018-2026 vs plain weekly DCA (cost basis, final value). Paper can only verify plumbing (signals rare). PASS: beats DCA on cost basis by >10% in backtest.

Relative-strength pair trade on Hyperliquid perps, daily: long HYPE / short SOL, equal notional.
Entry: when the ratio R = HYPE/SOL closes above its 50-day moving average AND the 20-day change of R is positive, go long HYPE and short SOL with equal dollar size.
Exit: close both legs when R closes below its 50-day moving average.
Stop: close both legs if the pair PnL reaches -6%. Leverage 2x gross.

Dollar-Neutral Cross-Sectional Momentum
Idea: Weekly: rank a liquid universe by trailing return, long the top, short the bottom, zero net exposure.

- Rules: Universe: top 20 Hyperliquid perps by OI. Every Monday 00:00 UTC rank by 28-day return (skip last 1 day). LONG top 4, SHORT bottom 4, equal-weight, dollar-neutral; optionally beta-neutral vs BTC. Hold 7 days, rebalance.
- Data: Daily prices, OI, funding for 20 perps
- Venue / timeframe: Hyperliquid perps; weekly
- Krexa risk settings: maxTrade 6% per name; daily loss 2%; SL 15% per name; leverage 1.5x gross
- Main failure mode: Momentum crashes (sharp reversals hit shorts); funding drag on crowded longs.
- Paper test plan: Backtest 2023-2026 with costs; paper 8 weeks (8 rebalances). PASS: OOS Sharpe > 1, net beta to BTC |b| < 0.2. betadotfund's Sharpe ~1.4 is their claim (on equities).

Indicator-gated grid on SOL, 1h bars, 1x.
Entry: inside the 20-period Bollinger Bands (2 sd), buy a 3% grid level each time price closes 1% lower than the last buy, but only while MACD(12,26,9) is above its signal line and above 0 and RSI(14) is below 70.
Exit: sell each grid level when price closes 1% above its buy price, or sell everything when MACD crosses below its signal line below 0.
Stop: close all if price closes more than 1 ATR(14) outside the Bollinger Bands.

Day-of-Week Tilt (Sunday-night buy, Monday exit; avoid Thursday)
Idea: Exploit the reported Monday strength / Thursday weakness with a time-based long.

- Rules: BTC. BUY Sunday 22:00 UTC, SELL Monday 23:59 UTC. Optional short leg Thursday 00:00-23:59 UTC only if BTC < 20-day MA. No indicators. Also usable as a DCA timing rule (DCA on Sunday night).
- Data: Hourly price, calendar
- Venue / timeframe: BTC perp/spot; weekly cadence
- Krexa risk settings: maxTrade 15%; daily loss 2%; SL 4%; leverage 1x
- Main failure mode: Calendar effects are small and decay once published; 6-month samples are noise.
- Paper test plan: Backtest 2018-2026 by year to see stability; paper 8 weeks. PASS: positive in >=6 of 8 calendar years in backtest. SuperTrade's 6-month numbers are their claim.

CME Gap Magnet
Idea: Weekend BTC moves away from the Friday CME close; trade the tendency to revisit the gap early in the week.

- Rules: Record CME BTC Friday close (21:00 UTC). At Sunday CME open (22:00 UTC) if |spot - CME close| > 1.5%: trade toward the gap with entry on first 1h close moving toward it. TP = gap fill level; SL = 1.2x the distance beyond the weekend extreme; time stop Wednesday 00:00 UTC.
- Data: CME BTC futures close (public), spot 1h
- Venue / timeframe: BTC perp; 1h
- Krexa risk settings: maxTrade 10%; daily loss 2%; SL ~2-3%; leverage 2x
- Main failure mode: Gaps can stay unfilled for months in trending markets.
- Paper test plan: Backtest every weekend 2020-2026 (fill rate within 3 days); paper 8 weekends. PASS: fill-within-time-stop >60% and positive expectancy.

High-FDV / Low-Float Listing Short
Idea: Short newly listed high-FDV tokens on perps after the launch pump; historically 'down only'.

- Rules: Universe: tokens listed on Hyperliquid/perp DEX in the last 30 days with FDV/MCAP > 5 and upcoming unlocks. SHORT when price closes below the 3-day VWAP after day 2 AND funding > 0. SL above listing-day high; TP -35% or 21-day time stop. Size small, never more than 3 at once.
- Data: Listing date, FDV, circulating supply, unlock calendar, perp funding
- Venue / timeframe: Hyperliquid perps; 4h/daily
- Krexa risk settings: maxTrade 4% per token; daily loss 2%; SL 15-20% (above launch high); leverage 1x
- Main failure mode: Low-float squeezes (insiders control supply - RAVE-style pumps); author's own plays were hedging an airdrop, not naked shorts.
- Paper test plan: Backtest on 2025-26 listings (>=30 tokens); paper 6 weeks. PASS: >55% of shorts profitable, no loss > 1R.

STRAT Trap-Bar Fade (with regime filter)
Idea: A candle that breaks the prior high, sweeps stops and closes red is a trap: fade it, but only with the higher-timeframe trend.

- Rules: 1h BTC/ETH. SHORT trigger: candle takes out prior bar high (2U) and closes red below prior close, AND 4h trend down (price < 4h EMA50). LONG mirror: takes out prior low (2D), closes green, 4h trend up. SL beyond sweep extreme. TP1 1R (move SL to break-even), TP2 2R, TP3 3R.
- Data: 1h & 4h OHLC, EMA
- Venue / timeframe: BTC/ETH perps; 1h
- Krexa risk settings: maxTrade 8%; daily loss 2%; SL beyond sweep (~1-2%); leverage 2x
- Main failure mode: Without the regime filter it 'fades every wiggle' (author's own warning).
- Paper test plan: 4 weeks, >=30 trades. PASS: avg R > 0.3, WR > 45%. Note author says indicator session stats are NOT a backtest.