Fades survive · continuation dies Scalping / orderflow Bybit perps (183 symbols) 600 days · 5m bars

Scalping Strategies —
Four Backtested Setups

Four specific entries/exits derived from the scalping research program, each backtested across the 183-symbol Bybit perp universe (ATR≥0.5% filter, 600 days of 5m bars) — net of Bybit taker fees + slippage (~0.14% round trip).

E8 + E4 (overextension fades) survive costs · E3 + E5 do NOT
+93,840R
Best strategy (E4)
IVA break-in fade · 96.8% win rate · PF 3.51
+91,893R
E8 VWAP-deviation fade
535k trades · 53.7% WR · PF 1.36
−31,816R
E3 IVA breakout
Continuation does NOT survive taker costs
4
Strategies tested
E8 · E4 · E5 · E3

Verdict

The overextension-fade family is the edge; continuation is not. E8 (VWAP-deviation fade) nets +91,893R (PF 1.355) and E4 (IVA break-in fade) nets +93,840R (PF 3.508, 96.8% win rate) after taker fees + slippage, with P(loss)=0.0 from resampling.

E5 (CVD-divergence fade) loses −14,873R (PF 0.775) and E3 (IVA confirmed breakout) loses −31,816R (PF 0.749) net of costs — the continuation family fails the taker-cost test, consistent with the research finding that fades are the edge on this universe.

All four share the same scalping skeleton: 5m bars, session-anchored reference levels, 1×ATR stop, TP at 2R or a value level, 60-minute time stop.

The four strategies

IDStrategyTypeEntryExit
E8VWAP-Deviation Fadefade (reversion)5m close > 2× ATR from session VWAPstop 1× ATR, TP 2R (or VWAP), 60-min time stop
E4IVA Break-In Fadefade (failed expansion)close beyond 30m IVA then back inside ≤12 barsstop 1× ATR, TP IVA POC, 60-min time stop
E5CVD-Divergence Fadefade (divergence)new session extreme without CVD confirmationstop 1× ATR, TP 2R, 60-min time stop
E3IVA Confirmed Breakoutcontinuationclose beyond IVA + volZ≥2 + acceptancestop behind IVA, TP next node/2× ATR, 2h runner

Headline results (net of taker fees + slippage ~0.14% RT)

StrategyTradesWin%Net Total RNet PFP(loss)
E8 VWAP-Fade535,00253.7%+91,893R1.3550.0
E4 IVA-BreakIn556,40196.8%+93,840R3.5080.0
E5 CVD-Divergence109,23643.3%−14,873R0.7751.0
E3 IVA-Breakout265,93144.1%−31,816R0.7490.991
Read: E4’s 96.8% win rate with PF 3.51 is a high-hit-rate fade — most trades revert quickly; the few losses are capped at 1R. E8 is the high-frequency workhorse. The two winners are both overextension fades; the loser of the fades (E5) is the one that relies on CVD divergence as a standalone trigger rather than confluence.

How it was tested

Universe

183 Bybit perps with ATR≥0.5%, 600 days of 5m bars, UTC days. Screened-first design from the scalping-research pipeline.

Costs

Bybit taker fee + slippage ≈ 0.14% round trip applied to every trade. Gross-vs-net pages show the full cost impact.

Per-strategy pages

Overview (KPIs, exit mix, net equity curve), regime breakdown by ATR, entry-filter testing (Spearman ρ, p, train/test, held-up verdict), costs, and a Monte Carlo net-R resampling with P(loss).

Validation gates

t>3, walk-forward OOS, net-of-cost — the research program's hard gates before a strategy earns “survives” status.

Run it yourself

cd /home/jose/workspace/scalping-research
python3 backtest/strategy_engine.py --store scalping-data/store_wide --out-dir scalping-data \
  --screened-list scalping-data/screened_list.txt
python3 backtest/build_strategy_analysis.py --store scalping-data/store_wide --out-dir scalping-data
python3 backtest/compute_costs.py && python3 backtest/compute_filters.py
python3 backtest/compute_regime.py && python3 backtest/compute_montecarlo.py
python3 backtest/build_strategy_site.py