ICT liquidity sweep: does it work?
A liquidity sweep is price running just past an obvious high or low, where stops are said to sit, then turning back. We tested it the way the shelf tests everything: on 11 futures markets, 18 years of one-minute data from 2008 to October 30, 2026, after costs, against 5,000 coin flips on the years before 2023. Passed on 0, came close on 2, failed on 28, no trades on 3 of 33 tests.
What it means: Not a pass anywhere, as the shelf runs it. The closest: Failed breakdown on CL, 77 (Close: 40 to 69, or a pass that misses a check). That is a lead, not an edge: the next step is your version, with your times and your stop, tested the same way.
Failed breakdown
Yesterday's day-session low or high. A bar trades 0.5 points or more through it, then a close back across it within 60 minutes trades the reclaim, with the stop past the flush. Between 9:30 and 15:00; one trade a day; out at 15:30.
| Market | Score | Trades | 2008 to 2022 | Since 2023 | Coin flips as good (of 5,000) |
|---|---|---|---|---|---|
| ES | 13 Failed | 3,008 | −$76,103 | −$39,418 | 799 |
| NQ | 9 Failed | 3,240 | −$16,982 | −$46,386 | 1,081 |
| YM | 16 Failed | 3,143 | −$22,433 | +$5,623 | 649 |
| RTY | 9 Failed | 3,141 | −$18,369 | −$12,423 | 1,096 |
| GC | 14 Failed | 2,388 | −$15,719 | −$17,079 | 1,580 |
| CL | 77 Close | 619 | +$27,662 | +$247 | 410 |
| ZN | 11 Failed | 23 | −$37 | −$622 | 2,094 |
| ZB | 14 Failed | 146 | −$11,877 | +$45 | 3,578 |
| 6E | No trades | 0 | — | — | — |
| 6J | No trades | 0 | — | — | — |
| NG | No trades | 0 | — | — | — |
Yesterday's high or low
Yesterday's high or low: a close 0.1% beyond it, 10 minutes beyond, then the first close back inside is traded against the sweep. Between 9:30 and 11:30; out an hour later.
| Market | Score | Trades | 2008 to 2022 | Since 2023 | Coin flips as good (of 5,000) |
|---|---|---|---|---|---|
| ES | 28 Failed | 1,253 | +$4,863 | −$7,838 | 47 |
| NQ | 57 Failed, stopped | 1,410 | +$30,335 | −$4,837 | 132 |
| YM | 16 Failed | 1,264 | −$2,827 | −$2,836 | 887 |
| RTY | 31 Failed | 1,404 | +$6,960 | −$6,012 | 169 |
| GC | 35 Failed | 964 | +$16,447 | −$25,891 | 34 |
| CL | 14 Failed | 1,267 | −$15,048 | +$1,346 | 1,555 |
| ZN | 13 Failed | 586 | −$12,142 | −$4,351 | 789 |
| ZB | 13 Failed | 958 | −$41,475 | −$9,236 | 769 |
| 6E | 11 Failed | 773 | −$9,140 | −$626 | 2,145 |
| 6J | 11 Failed | 611 | −$12,318 | −$492 | 4,013 |
| NG | 44 Failed, stopped | 1,225 | +$16,303 | −$10,238 | 42 |
NQ at 57 and NG at 44 are Failed, not Close: each made money in the fifteen years before 2023 and has lost money since, over 329 and 259 trades since, which stops it whatever the score (the NQ page).
Last week's high or low
Price closes 0.1% beyond last week's high or low and stays beyond the level for 10 minutes; the first close back inside the level is the trade, against the sweep. Between 9:30 and 11:30; out an hour later.
| Market | Score | Trades | 2008 to 2022 | Since 2023 | Coin flips as good (of 5,000) |
|---|---|---|---|---|---|
| ES | 11 Failed | 667 | −$20,039 | +$125 | 3,214 |
| NQ | 46 Close | 645 | +$581 | +$25,920 | 1,593 |
| YM | 10 Failed | 558 | −$1,989 | +$770 | 1,519 |
| RTY | 9 Failed | 713 | −$4,301 | +$2,036 | 1,577 |
| GC | 11 Failed | 484 | −$17,048 | −$29,434 | 3,869 |
| CL | 11 Failed | 592 | −$29,911 | +$7,724 | 4,639 |
| ZN | 11 Failed | 266 | −$6,142 | −$2,398 | 1,298 |
| ZB | 13 Failed | 463 | −$16,870 | −$10,762 | 605 |
| 6E | 11 Failed | 357 | −$9,324 | −$4,338 | 4,091 |
| 6J | 28 Failed | 322 | +$1,642 | −$4,563 | 199 |
| NG | 32 Failed | 504 | +$1,573 | −$209 | 614 |
Every number on this page is read from the shelf’s records. All the ICT ideas → · What the winners have in common →