91% of Wickless ES Trades Hit Their Target. Only 16.5% Make Money.

Two years ago the internet decided that a candle with no lower wick marks a level worth buying. We mechanised that claim on the Nasdaq in July and killed it. We have now run the identical machine on twelve and a half years of our own S&P archive. The headline number replicates to the decimal. With a 0.25-ATR target against a 3-ATR stop, 91.1% of ES trades reach their target.

They also lose $34.63 each.

That gap is the whole article. A target-hit rate is not a win rate, and on a five-minute S&P bar the difference between them is $29.50.

The rule, unchanged from the NQ test

• 5-minute ES bars, close above the EMA-50 for longs, mirrored for shorts.
• A bullish candle whose low equals its open — no lower wick at all.
• The candle low becomes the level.
• Limit entry on the retest of that level. One position at a time.
• Three exits, each in ATR(14) multiples: balanced 1R:1R, a 2R target against a 1R stop, and the 0.25-ATR target against a 3-ATR stop that manufactures the viral number.

Nothing was retuned for the S&P. The archive gave us 868,567 five-minute bars between 2 January 2014 and 10 September 2026, containing 68,726 wickless levels. The control — the same trend filter, the same retest entry, the same three exits — ran on 102,673 ordinary lows that had a normal wick.

Target-hit rate is not win rate

Here is the same set of trades counted two ways. The left number is the win rate before costs. The right number is how often the trade finished above zero after the round trip came out.

Wickless levels, 63,904 trades	Win rate before costs	Net wins	Per trade	Total	PF
0.25 ATR TP / 3 ATR SL	91.1%	16.5%	−$34.63	−$2,212,763	0.089
Balanced 1R:1R	54.3%	43.9%	−$25.91	−$1,655,844	0.513
2R target	37.9%	37.5%	−$22.52	−$1,438,850	0.679

Look at the top row. The trade reached its target 58,191 times out of 63,904 and was stopped out only 5,637 times. Nine stops in a hundred, and the strategy still ends at a profit factor of 0.089 — it returns under nine cents for every dollar it risks. The t-statistic is −81.18, which is not a marginal result, it is the most reliably negative number this series has produced.

The mechanism is arithmetic, not markets. A quarter of one ATR on a five-minute ES bar is frequently a smaller amount of money than $29.50. When the target is smaller than the round trip, hitting it is a loss. Of that row, 74.6% of trades hit their target and lost anyway.

The other two rows show the same thing without the drama. As the target widens, the gap between the two win rates narrows — 10.4 points at 1R:1R, 0.4 points at 2R — because a wider target clears the cost. What does not change is the sign of the result.

Gross of costs, there is nothing here

The obvious objection is that we killed a real edge with execution assumptions. So here is every configuration before a cent of cost comes out.

Average per trade, gross	Wickless	Wicked control
Balanced 1R:1R	+$3.59	+$1.24
2R target	+$6.98	+$9.08
0.25 ATR TP / 3 ATR SL	−$5.13	−$17.74

Five of the six cells sit inside a ten-dollar band around zero, and the two tiny-target cells are negative even gross — the geometry loses money before anyone is paid. Nothing in this table is close to the $29.50 it would need to survive. There is no cost assumption that rescues this; halve our slippage and every row is still red.

The difference from the Nasdaq: the control dies too

The NQ version of this test had a second finding that we thought was the more interesting one. The placebo — ordinary lows with a wick — made about +$575k where the wickless levels lost. We concluded that the viral setup was sitting on top of generic trend-pullback continuation and then filtering it down to the worst candle type, and we put the control on the live board as a paper algo.

On the S&P the control does not make money either.

Balanced 1R:1R	Trades	Net wins	Per trade	Total	PF
Wickless levels	63,904	43.9%	−$25.91	−$1,655,844	0.513
Ordinary wicked lows	90,407	50.8%	−$28.26	−$2,555,182	0.733

Both lines go down, and the wickless setup ends above the control: −$1,655,844 against −$2,555,182. The control keeps the better profit factor, 0.733 against 0.513, and takes the worse dollar total, because it runs 26,503 more trades at a similar loss per trade. That reverses the NQ ranking, where the wicked control was the profitable one and the wickless filter was what ruined it. Here the filter is the less expensive of two losing ways to trade a retest.

The reason is in the gross table above. It is also the single most useful sentence we can write for anyone moving an intraday setup between indices. At five-minute scale on the S&P, the trend-pullback effect is not there before costs, so costs decide the outcome. The control averages +$1.24 a trade gross. It is not being killed by our fill model. There is nothing for the fill model to kill.

That is a different failure from the NQ one. On the Nasdaq the wickless filter destroyed a real effect. On the S&P there is no effect at that timeframe for the filter to destroy, and both level types converge on the same answer, which is the cost of trading them.

What we changed in our own book

Nothing, and that is the point of running it.

1. Wickline stays NQ-only. The sleeve was built and validated on the Nasdaq, and this test says the S&P port would trade 90,407 times to lose $2.6m. We had considered extending it. We are not extending it.
2. We report target-hit rate and net win rate separately from now on. Our own live algo pages quote net. Any strategy claiming a win rate above 85% is quoting the other number until it says otherwise.
3. Cost-to-target ratio is now a pre-registered check. If the target is smaller than the round trip, the backtest does not need running. That check would have disqualified the 0.25-ATR configuration in one line, on either index.

The twelve-and-a-half-year window matters less here than it did in our RSI-2 re-run. This result is not close enough to zero for five extra years to change the verdict, not with 63,904 trades and t = −81.18. What the long window buys is the certainty that we are not looking at one regime.

The ES tick archive this ran on, aggressor side on every print back to 2014, is in our historical data packages.

Methodology: ES continuous front-month, 5-minute bars, 868,567 of them from 2 January 2014 to 10 September 2026, built from our own trade prints. Signal: an up candle that opens on its own low, taken only while the close sits above the EMA-50, with the short side mirrored. The candle low becomes the level and the order is a limit sitting on it, filled when price comes back. Never more than one position open at once. Exits in ATR(14) multiples as stated; when a bar touches both, the stop is taken first. One contract, $4.50 commission plus two ticks of slippage per round trip, $29.50 on ES. Control: the identical machine on bullish candles whose low is below their open. 68,726 wickless levels and 102,673 wicked levels produced 63,904 and 90,407 trades respectively. The NQ comparison figure comes from our July 2026 NQ study, which covers 2019–2026.