CME Gap Tracker for Gold and Bitcoin: Open Gaps, Fill Rate and a Fade-Trade Test - indicator for MetaTrader 5

CME Gap Tracker for Gold and Bitcoin: Open Gaps, Fill Rate and a Fade-Trade Test

Finds the weekend CME gap on gold and on 24/7 Bitcoin charts, keeps every unfilled gap on the chart until price returns to it, and shows how often gaps filled, how fast, how far price ran first, and what fading them would have made. Non-repainting, with buffers for an Expert Advisor.

The idea

"CME gaps always fill" is something you hear in every Bitcoin chat and in a lot of gold channels too. CME futures stop trading on Friday at 16:00 Chicago time and start again on Sunday at 17:00. Whatever the market did in between shows up as a gap on the futures chart, and the belief is that price comes back to close it.

I wanted two answers before I believed it. How often does the gap actually fill? And if it does fill, can you make money trading toward it, or does price run the other way first and stop you out? This indicator answers both on your own chart.

What it does

• Converts the CME Friday close and Sunday open to your broker's server time, with the US daylight saving rule for Chicago and the rule of your server clock.
• Takes the price of your chart at those two moments. The difference is the gap. Gaps smaller than InpMinGapPct (0.30% by default) are ignored.
• Draws each gap as a box and labels it with its size. An unfilled gap stays on the chart, extended to the right, until a closed bar trades back to the Friday close. Then the box turns gray and the label shows how many hours the fill took.
• Puts an arrow on the bar where CME opened: up for a gap down, down for a gap up, in the direction of the fill.
• Shows a scorecard panel for the last InpWeeksBack weeks (104 by default).

On a Bitcoin chart that trades all weekend there is no visible gap between Friday's last bar and Monday's first bar, so an ordinary weekend-gap tool finds nothing. The CME gap only appears when you look at the price at the two CME moments, which is what this indicator does. On gold the spot market closes over the weekend too, so the CME gap and the broker's weekend gap are almost the same thing.

One thing to be clear about: the indicator uses the prices of the chart it runs on, not the futures prices. On a spot or CFD chart you get the spot move over the CME weekend. The gap on the real futures chart can differ a little because futures trade at a premium or discount to spot.

What the panel shows

• The number of gaps, split into up and down, and how many weekends were below the minimum size.
• How many gaps filled, and what share filled within 24 hours and within 7 days.
• The median time to fill, and the time the slowest 10% of gaps needed.
• Heat before fill: how far price ran away from the fill level before it came back, in multiples of the gap size. A heat of 2.0x on a 1% gap means price first moved another 2% the wrong way.
• A fade-trade test: enter at the CME open toward the Friday close, stop at InpStopGapMult times the gap size on the other side, exit after InpMaxHoldDays days if neither level is hit. The spread of the entry bar is charged. When the stop and the target are both inside one bar, the stop is counted.
• The number of open gaps in the scored window, and the nearest open fill level above and below the current price (these two look at every gap in the loaded history).

What it found on gold

XAUUSD H1, MetaQuotes-Demo data, default inputs except where noted. The two windows overlap by a few days at the end of December 2025.

So the saying is mostly true on gold: 83% to 97% of the gaps filled. The trade is where it falls apart. Among the gaps that filled, the median one first ran about twice its own size the wrong way, so most fade trades with a stop of one gap size lost.

I tried wider stops with the 0.30% minimum. A wider stop wins more often, but each loss is bigger and the total stayed negative in both years:

2025 had only 6 gaps of 0.30% or more, so treat that column as a small sample. What I take from it is simple. On gold the gap is a good place to expect price to visit, but fading it at the open with a fixed stop didn't pay in either year.

Some gaps don't fill for a long time. As of 18 September 2026 one gold gap from 2025 was still open with a fill level of 3201.00, far below the price.

Fig. 1. XAUUSD H1, late January 2026: a +0.43% gap filled in 106 hours and a -2.47% gap filled in 32 hours

Fig. 2. The scorecard panel for the four runs in the table

About Bitcoin

I couldn't test it on Bitcoin. My test broker, MetaQuotes-Demo, has no 24/7 BTCUSD, only a Bitcoin ETF that trades during Nasdaq hours. The code path for a 24/7 chart is the same one that runs on gold: the price at the Friday close time and at the Sunday open time. Before you rely on it for BTC, check the first few gaps against a CME futures chart, and set the server offset inputs for your broker.

Inputs

The defaults (GMT+2 in winter, GMT+3 in summer, European rule) match MetaQuotes-Demo and many other MetaTrader brokers. If your broker's server runs on UTC all year, set both offsets to 0 and the rule to fixed. Use H1 or a lower timeframe. On H4 and above the CME times fall inside a bar and the prices are less exact.

Using it from an Expert Advisor

The gap is written on the bar where CME opened, as soon as that bar opens, so an EA reads it at shift 0. It only uses the Friday close and the open price of that bar, so it never changes afterwards. I checked this with a test EA that recorded every gap as it appeared and compared the list with the indicator's buffers at the end of the run: 24 of 24 gaps matched on gold in 2026.

int      handle;
datetime lastBar = 0;

int OnInit()
  {
   handle = iCustom(_Symbol, PERIOD_H1, "CMEGapTracker");
   return(handle == INVALID_HANDLE ? INIT_FAILED : INIT_SUCCEEDED);
  }

void OnTick()
  {
   //--- a gap is written on the CME open bar as soon as that bar opens
   datetime bar = iTime(_Symbol, PERIOD_H1, 0);
   if(bar == lastBar)
      return;
   lastBar = bar;

   double up[1], down[1], target[1], stop[1];
   if(CopyBuffer(handle, 1, 0, 1, up) < 1 || CopyBuffer(handle, 0, 0, 1, down) < 1 ||
      CopyBuffer(handle, 2, 0, 1, target) < 1 || CopyBuffer(handle, 3, 0, 1, stop) < 1)
      return;

   if(up[0] != EMPTY_VALUE)
      Print("Gap up, fill level ", target[0], ", fade stop ", stop[0]);
   if(down[0] != EMPTY_VALUE)
      Print("Gap down, fill level ", target[0], ", fade stop ", stop[0]);
  }

Buffers 4 and 5 are handy for a different kind of EA, one that uses open gaps as targets for trades it takes for other reasons. The indicator has no input groups on purpose, because MQL5 counts each "input group" as a parameter of iCustom() and every value after it would shift by one.

Limits

It works on bar data, so it can't tell the order of events inside one bar. It uses your chart's prices at the CME times, not the futures prices. US holidays that shorten the CME week are not handled separately. The fade test charges the spread of the entry bar but no commission and no slippage, and weekend opens are exactly when slippage is worst. All the numbers above describe past gaps on one broker's data.

Ali Rajput