2:11 a.m. I was staring at a number I hadn't expected: 1,802.40. That was gold's low, fifteen minutes after the NFP print. My long was from 1,810.20. Stop at 1,806. I was out before I finished reading the headline.
Twenty minutes. $3,600 gone. 60% of my account. I sat in the dark in my Tiong Bahru flat and did not move until the sun came up over the HDB block across the road.
For nine years after that, every time NFP Friday came around, I did something I never told anyone. I rebuilt the trade. On paper. Same entry. Same stop. Same size. Then I changed one variable. Next month, another. I have 47 versions of that trade in an old notebook. The 48th was the first one that actually worked.
What actually happened that Friday
The number came in at 271k against a 185k forecast. Gold dropped 80 pips in under 15 minutes. Spreads on my broker widened from 18 points to 210 in the first 90 seconds.
My stop was 42 pips wide. My position was 0.4 lots on a $6,000 account. That is 1:100 leverage doing exactly what 1:100 leverage does. The stop was inside the noise. The size was outside the account. Both of those things were my fault.
But here is what I got wrong for nine more years: I thought the problem was the stop.
The explanation everyone gives is wrong
The internet told me to widen the stop. So I did. Version 12 had a 120-pip stop. Version 19 had 180. Version 31 had no stop at all, just a mental one. That one cost me $2,100 on a Tuesday in 2019 because I was in the bathroom.
Widening the stop doesn't fix a news trade. It just changes how slowly you lose.
The real problem with NFP is not volatility. It is that the first 90 seconds of gold after the print is not a market. It is a queue. Market makers pull quotes, spreads blow out, and the price you see is not a price you can trade. You are not trading gold. You are trading whatever the last guy was willing to pay to get out.
Every version of my trade that entered before the print, or in the first three minutes after, was doomed. Not because of the direction. Because of the venue.
What version 48 did
I stopped trying to trade the event. I started trading the reaction to the reaction.
Version 48: no position into the print. Wait 12 minutes. Let the spread come back under 40 points. Then look at the H1 close. If gold has closed the first hour back inside the pre-news range, I take the fade. If it closed outside, I wait for the retest and take the continuation.
Entry is never earlier than 12 minutes after the print. Size is half my normal. Stop is 1.5x the ATR of the previous four H1 candles, not a fixed pip number.
I ran that version on paper for 14 NFP Fridays. Nine winners. Five losers. Average win 38 pips. Average loss 22. That is not a spectacular edge. It is a survivable one.
The first live version was March 2024. Gold spiked to 2,195 on the print, came back to 2,178 by minute 14, closed the H1 at 2,181. I went long at 2,183. Stop at 2,171. Out at 2,214 two hours later. $840 on 0.15 lots.
It was the smallest NFP win of my life. It was also the first one I didn't have to lie down after.
What the notebook taught me
Forty-seven versions. The first 30 all tried to predict the number. Versions 31 through 40 tried to predict the reaction. Versions 41 through 47 tried to predict the reaction to the reaction. Version 48 just waited for the market to finish having its seizure and then asked one question: did the move hold?
The lesson was not about stops, or size, or leverage, or any of the things I spent nine years blaming. The lesson was that I had been trying to be early. On NFP, being early is the same as being wrong. The only people who make money in the first three minutes are the ones selling the liquidity to everyone else.
I deleted the indicator set that Friday in 2015. I should have deleted the impulse to be first.
If you are still trying to catch the NFP candle, you are not trading gold. You are paying a tax to the people who are. Wait twelve minutes. Let the spread normalise. Trade the close, not the print. The move that lasts is always the second one.