The Fair Value Gap strategy lost on 273 live SPX trades, and on SPY, QQQ and gold
Live paper record 8 Sep 2026 to 13 Oct 2026. Updated 13 Oct 2026.
Short answer: no, not the way I traded it. Four versions of the fair value gap strategy took 273 paper trades on same-day SPX options between 8 Sep and 13 Oct, filled at real prices. Together they lost $5,240, and every version finished below zero.
The strange part is that on the chart it made a little money: the trades gained about 0.3 of an index point each on average. That wasn't nearly enough to pay for getting in and out of an option. I then tried it on SPY and QQQ shares, where costs are tiny, and on gold. It didn't work there either.
-$5,240net, 1 contract a trade
273live paper trades
48%made money
13 of 22sessions lost money
Running P&L by session. Red: all versions together. Grey: each version on its own. Paper trades bought at the real ask and sold at the real bid.
What is a fair value gap?
Take three candles in a row. If the middle one moves so fast that the first and third don't overlap, the empty space between them is a fair value gap, or FVG. It comes from ICT (Inner Circle Trader) teaching. The idea is that the market moved too quickly to trade fairly through that range, so price tends to come back and "rebalance" it, and you can trade that return.
It's one of the most taught setups in retail day trading, it's easy to spot on any chart, and it backtests beautifully. That's exactly why I wanted to see what happens with real prices and real costs.
The exact rules I tested
These rules were fixed before the first live trade and didn't change during the test.
Chart: the S&P 500 index (SPX) on 1-minute candles, regular session only.
The gap: a bullish gap is when candle 1's high is below candle 3's low; a bearish gap is the reverse.
Minimum size: the gap must be bigger than a fraction of the previous week's average 1-minute candle range: 0.4 for the "smaller gaps" versions, 0.6 for the "larger gaps" ones, and never smaller than 1 SPX point. Scaling to last week's volatility keeps the filter fair as the market quietens or speeds up.
Trend filter: only gaps that form against the trend are traded: a bullish gap while SPX is still below a falling 50-period EMA, or a bearish gap above a rising one. The gap is then faded in the trend's direction (short a bullish gap, long a bearish one).
Entry: touches of the gap in the first 5 minutes after it forms are ignored. The first touch of the gap's near edge between 5 and 15 minutes after it formed triggers the trade: buy an at-the-money same-day SPX put (to fade a bullish gap) or call (to fade a bearish one) at the asking price.
Exits: the target and stop are multiples of the gap's own size measured on the index: 2× target and 2× stop for two versions, 4× target and 3× stop for the "wide" ones. Once price has closed the whole gap and moved a little past it (2 points, or 1 for the wide versions), the stop moves up to the gap's far edge. The option is sold at the bid. No new trades after 15:30, and anything open is closed then.
One position at a time per version.
Results, version by version
Version
Trades
Won
Index points
Net
Smaller gaps, 2x target / 2x stop
77
40
+17.5
-$1,700
Larger gaps, 2x target / 2x stop
60
32
+21.9
-$1,155
Smaller gaps, 4x target / 3x stop
77
31
+9.5
-$2,115
Larger gaps, 4x target / 3x stop
59
27
+43.4
-$270
All versions
273
130
+92.2
-$5,240
"Index points" is what the trades made on the SPX chart itself, before the option. Over all 273 trades that was +92 points, which on a chart looks like a working strategy.
Why it failed
The option cost more than the move was worth. Fitting a line through every trade gives a simple rule of thumb: each trade made about $45 per index point, minus $35. Every trade started $35 behind from the bid/ask spread, and an at-the-money option only moves about half as much as the index. To break even, the strategy needed about +0.8 points a trade. It averaged +0.3.
The gaps were tiny. The median gap was 1.5 SPX points. When the whole expected move is that small, a slightly late fill or a quote that's a second old is a big share of it, and a 1-minute backtest can't see either.
The trailing stop "locked in" profits that weren't there. In the wide versions, 70% of trades went the right way on the chart, but only 43% made money. Many closed the gap, came back and stopped out at its far edge: a small win on the chart, a loss after the spread. Of the 66 trades that ended on the trailed stop, 63 had gained on the chart, and 40 of those 63 made nothing or lost money.
The direction didn't matter. Replaying every trade with the opposite option gave about the same result. The touch of the gap tells you that something is about to happen, not which way.
The backtest was fitted to its own data. Before going live, my backtest on 1-minute candles promised almost two index points a trade and a 70% win rate. The exits had been chosen from more than 120 combinations on the same three months. Live, the win rate in dollars was 48%.
I also tried it on SPY, QQQ and gold
Options are expensive to trade in and out of, so I tested the same rules on things that are cheap to trade.
SPY and QQQ shares: the same rules on 100 shares, scaled to each ETF's price. On 1-minute candles, every version on QQQ lost and two of four on SPY lost; the two SPY winners were small and not convincing. Replayed on real bid/ask quotes, all eight versions lost.
A trend-plus-FVG version on SPY and QQQ: fading stretched gaps in the direction of a short-term trend looked positive at first, but a version with no gap required at all earned exactly the same. The gap added nothing.
Gold: ten years of 1-minute gold prices, plus the gold and silver ETFs, with the FVG combined with an EMA trend and pullback. None of the 60 exit settings made money, and bigger gaps didn't help.
Did I test it wrong?
Quite possibly. I tested the version of Fair Value Gap (FVG) described above, with settings I thought were fair, and it didn't work. That doesn't prove every version fails. I may have picked the wrong settings, the wrong market or timeframe, or missed a filter that experienced traders take for granted. Many strategies also depend on judgement that a set of fixed rules can't capture.
If you trade Fair Value Gap (FVG) and it works for you, I'd genuinely like to hear how. Tell me the chart and timeframe, what counts as a setup, any filters, where you get in, where your stop and target go, and when you get out. I'll test your version on the same data, the same way, and publish the result here, whichever way it comes out.
A version that makes at least one index point per trade after realistic fills, tested on tick data rather than 1-minute candles. Trading it through index futures, where the spread is a fraction of an option's, would be the version worth testing next. If you have that version, the button above is for you.
Questions
What is a fair value gap?
A three-candle pattern where the middle candle moves so fast that the first and third candles don't overlap. The untraded range between them is the gap. ICT-style traders expect price to come back and fill it.
Does the fair value gap strategy work?
Not in my test. Four versions traded 273 times on SPX same-day options over 22 sessions and lost $5,240 combined. Backtests on SPY, QQQ and gold didn't work either.
Why do fair value gap backtests look so good?
Mine predicted almost two index points a trade and a 70% win rate. It was scored on one-minute candles, which can't show where price really traded inside the minute, and the exits were picked from over 120 combinations on the same months of data.
Does a bigger gap or a trend filter fix it?
Neither did in my tests. Every live trade already used a trend filter, and gaps were scaled to recent volatility. Bigger gaps were rare and the results for them weren't consistent.
How it was tested
Live paper trading: every trade above was taken by an engine running in real time, filled at real market bid/ask quotes on same-day SPX options, one contract each, without commissions. Nothing was added or removed afterwards.
Versions share signals: several versions trade the same setups with different exits, so the trade count overstates how many independent bets there were.
Backtests: the SPY, QQQ and gold tests used historical 1-minute data and, where available, recorded bid/ask quotes. They are backtests, not live trades.
This page records my own testing. It isn't financial advice. Same-day options can go to zero within minutes.
