Liquidity Sweep / Turtle Soup (ICT)

The claim

Institutions spike price past an obvious high or low to grab retail stops, then reverse. Wait for the sweep (a false break that closes back inside), then a displacement + market-structure shift, enter on the fair-value-gap retest, stop beyond the raid candle, target the opposite liquidity pool. Pitched at 50–65% win with 1:3 reward-to-risk.

How we tested it

Mechanized the full sweep → displacement → market-structure-shift → entry sequence on 48 Nifty-50 stocks, 5-min, 2 years, using prior-day high/low as the liquidity pool and the real Zerodha MIS cost model + slippage. We ran the ICT confirmation model against a naive no-confirmation fade, scored opposite-pool / 2R / 3R targets, and stripped out all costs to isolate any raw edge — 22,000 confirmed setups.

The data

Our verdict

The most sophisticated reversal model in the set, and the confirmation layer is genuinely doing something — requiring a displacement and a market-structure shift before fading a swept level lifts the win rate from 24% (naive fade) to 34%. The concept is internally coherent and, unlike most, it tells you exactly where it's wrong (a re-break of the raid candle).

But it has no cost-surviving edge. Gross of all costs and slippage the confirmed model sits exactly on the coin-flip line — about ₹0 per trade, 39% win at ~1.5 payoff — because the higher hit-rate is precisely offset by a worse payoff and fewer, later entries. So its net economics are identical to the naive Sneaky-Pivot fade (≈−₹290/trade), and it loses every year and on both the long and short side once real MIS friction is applied. The strategy thrives on hindsight: after the move you can always point to the sweep that reversed, but in real time every level gets poked constantly. The 50–65% / 1:3 figures are vendor numbers, not what the tape pays.