The Cable Short That Entered on Its Lowest Confidence Reading

The Cable short on September 2 was refused four times at high confidence and taken once at low confidence. From 14:13 UTC the gate scored the setup at 79, then 81, then 82, then 84 percent, and returned WAIT on every one. At 14:22 it scored 62 and entered short at 1.34989 with a stop at 1.35140, a risk band of about 15.5 pips. The first target at 1.34835 filled for +1.02R. Read as a ranking, that sequence is backwards. The system passed on its four best-looking reads and acted on its worst. Read correctly, it is the only sequence that makes sense, because the four high scores and the one low score were not measuring the same thing.

A bearish session that had already paid out

London had set a clearly bearish tone in Cable. Price sold from the London high at 1.35148 down to 1.34739, breaking beneath the prior day's low at 1.35050 and holding under the 60-minute fast and slow EMA structure for most of the session.

By the New York open that move had already done its work. The London selloff had reached an intraday extreme and produced exactly the mean reversion you would expect from one, with the NY overlap lifting price back into the 1.3510-1.3515 resistance band. The bearish case was intact. The cheap entry into it was gone.

The macro frame supported the direction without qualification. The Macro Agent was bearish GBPUSD at 82% confidence on dollar strength, Fed and BoE divergence, and a weak UK backdrop. The Trend Agent was bearish at 62% but explicitly transitioning, naming 1.35148 as the invalidation level.

One input did not cooperate. DXY sat at 99.46, slightly below its 5-day EMA of 99.529, so the dollar was not confirming the dollar-strength half of the thesis. VIX at 15.8 against a 5-day EMA of 15.51 kept the environment mildly risk-off, which leans against sustained Cable upside, but the currency leg was a genuine gap in the case rather than a rounding error.

London Mean-Reversion Failure Short. This was never a market sell. It was a sell-failure setup, which means the entry did not exist until the market first tried to rally and could not hold it, and that distinction is what separates the four refusals from the one entry.

Two different numbers wearing the same label

The first four evaluations scored the bearish thesis, and the thesis deserved every point it got. London direction bearish, Macro Agent bearish at 82%, 60-minute structure below the 5-day EMA and the slow EMA with MACD still under zero. Scores of 79 to 84 are an accurate description of that picture.

What those scores were not describing is an available trade. Price was mid mean-reversion, climbing into the 1.3510-1.3515 band and pressing the Trend Agent's invalidation at 1.35148. Selling into that is selling into strength with the stop already close to being hit. There is no entry there at any level of conviction about the direction.

What the trigger actually required

The setup was conditional on three things happening in sequence: a rejection at 1.35110-1.35148, then a 5-minute candle closing back below 1.35055, then a failed retest of 1.3500 and VWAP from underneath. Nothing before that third event is the trade.

When all three printed at 14:22, the entry became live and was scored on its own merits rather than on the thesis behind it. Six of seven confluences cleared. DXY did not, sitting below its 5-day EMA. The 5-minute and 15-minute momentum were still bullish, which is precisely why this is a sell-failure and not a market sell. Score: 62, with a Trend Agent instruction to reduce size.

Why the low number is the honest one

Professional traders: a system that scored this entry in the eighties, on the grounds that the direction had been obvious since London, would be telling you something false about the risk you were taking. The DXY miss was real and the counter-trend momentum was real.

SkyAnalyst does not favor any single strategy and it does not let a strong directional view inflate the score of a flawed entry. It reads the tape first, scores the entry that is actually in front of it, and sizes down when the confluence is incomplete. The trade was right and the reservation was also right; both went in the record.

What this trade teaches

The lesson is that a confidence score is attached to an entry, not to an opinion, and the two move independently.

For nine minutes this desk was more certain about Cable than it was at the moment it actually sold Cable. Nothing about the direction weakened in that window; the bearish case at 14:22 was the same one that scored 84 at 14:20. What changed is that a specific, pre-declared sequence of price events completed, which converted a view into a position with a defined invalidation 15.5 pips away.

The second lesson is about scoring flaws honestly. The DXY miss did not disqualify the trade, and it did not get quietly absorbed into a rounder number either. It cost the entry roughly twenty points of confidence and a reduction in size, and the position was still profitable. A system that inflates scores to match its convictions loses the ability to size correctly, which is the only thing the score is really for.

From the desk

This is the fourth of five consecutive winners since August 31, and part of twelve wins in the last thirteen trades, a run that began on August 24.

It also lands on the right side of a book we have been watching closely. GBPUSD was the instrument that broke August: the long book took ten trades, won two and gave back 6.18R, more than double the entire month's net loss of -2.67R, while the GBPUSD short book lost a negligible 0.11R across two trades. The pair was not the problem. One direction of it was.

On August 27 every instrument trader was split into independent long and short books precisely so that distinction could be acted on. The GBPUSD long book came off. The short book, which had done nothing wrong, kept trading. This entry is that short book working, on a day when the pair rallied first and failed.

There is a detail in the record worth explaining. The trade shows a stop-hit exit reason alongside a first-target fill, which looks contradictory. The broker closes the full position at TP1, so the position was finished the moment 1.34835 traded. The tracker keeps following the instrument afterwards and eventually logs where the original stop would have been reached, long after there was anything left in the market. The realized result is the first target, +1.02R, and that is what enters the year-to-date record.