APXTT

Coming soon · MetaTrader 5 indicator

ATR said ~12 pips.
NFP moved 43.

The News Volatility Warning forecasts how big the next hour's range will be, in pips, and warns you before scheduled releases: NFP, CPI, FOMC and central-bank decisions. So your stop and lot size fit the hour you are actually trading.

Join the waitlist (free) How we test it

Size, not direction · next hour only · prototype, no live results yet

Replay of real EURUSD data · 2 Oct 2026 · not live EURUSD · H1
MT5 chart of EURUSD on 2 October 2026 with the APXTT panel: NFP in 5 minutes, expect about 56 pips, usual hour about 18, ATR(14) about 12
Forecast ~56 · Actual ~43 · ATR ~12 Closer than ATR, not exact

APXTT  News Volatility Warning   EURUSD

This hour 10:00: expect ~18.6 pips (usual ~12.4, ATR(14) ~11.3)

This hour: 2 of 19 pips used (12%)

Next hour 11:00: ~15.1 pips (usual ~12.3)

NFP in 5h 30m - expect ~53 pips (usual hour: ~18 pips, ATR(14) says ~11)

hour after: ~16 pips (event at 2026.10.02 15:30 chart time)

Model FULL (factor pairs 27/27) | events: event file (ends 2026-12)

Clock: server = NY+7 (forced)

Size, not direction. Prototype. Hours after the current one use the latest data (approx.).

Real EURUSD minute data, replayed fast around the US jobs report (NFP, 15:30 chart time = 12:30 UTC). The panel shows the indicator's numbers for each hour, recomputed from the same data. Each hour the grey box is the range traded so far. The green band is the expected range (the forecast), centred on the middle of what has traded, so it moves with the mid; it turns amber once the hour exceeds it. The faint dashed lines show where price ends if the whole expected range plays out in one direction. None of these are targets: the indicator forecasts the size of the range, not its direction.

The same hour in MetaTrader 5: a Strategy Tester replay of the indicator on real data (screen recording, no sound).

What you see on your chart

Three moments of the same US jobs report (NFP), in MetaTrader 5's Strategy Tester: a replay of 2 Oct 2026 on real data, not a live chart and not a performance record. The panel shows how much of this hour's expected range has been used and a lot size scaled to it. On the chart, the blue band is the expected range, centred on the middle of what has traded this hour and moving with it; the faint lines are the hour's low + the expected range and the hour's high − the expected range; the grey bar on the right is the next hour's expected size.

MT5 M5 chart of EURUSD at 14:42, 48 minutes before NFP: panel says this hour 15 of 17 pips used; a band of about 17 pips centred on the hour's range; a tall grey bar marks the next hour at about 58 pips
48 minutes before NFP. This hour: 15 of 17 pips used. The band is the 17-pip expected range, centred on what has traded. The tall bar on the right: the next hour, the NFP hour, is expected at ~58 pips. The vol-sized lot (1% risk, stop = expected range) is 0.59 here and 0.17 in the NFP hour.
MT5 M5 chart of EURUSD at 15:56, 26 minutes after the NFP release: this hour 40 of 56 pips used; a tall band of about 56 pips around the hour's range, its bottom edge just below the visible price scale
26 minutes after the release. This hour: 40 of 56 pips used (71%). The band shows the full expected range around what has traded so far; which way the rest goes, the indicator does not say. Where an edge falls off MT5's price scale, a note shows its price.
MT5 M5 chart of EURUSD at 16:58, the hour after NFP: this hour 32 of 27 pips used, range exceeded, band in amber
The hour after: range exceeded. 32 of 27 pips traded, so the band turns amber and the faint lines disappear: the expected size is already reached. It says nothing about whether the move continues.

Replay of 2 Oct 2026, tester (MT5 Strategy Tester, real EURUSD minute data). Size, not direction. Back-tested on one broker's data; your broker's ranges may differ.

How we test it

Most indicators are sold on a backtest fitted to the same data it is shown on. We do the opposite: the model was fitted on 2018–2023 and scored on 2024–2026 data it never saw, and it is always compared with the free tools you already have.

EURUSD, release hourAPXTT errorSame-hour averageATR(14)
NFP (30 releases)27 pips41 pips51 pips
FOMC (21 decisions)24 pipsabout 45 pips

Across 28 pairs, in the hour of a release for the pair's own currency, its range forecast missed by about a third as much as the best free tool and a fifth as much as ATR. ATR systematically under-calls release hours by a factor of 2–3. On ordinary hours the gain is small (4.35 vs 5.08 pips error on a 14-pip average hour).

Technical details (for the quants)
  • Model: HAR-type regression on log hourly realized volatility (daily / weekly / monthly lags), with a continuous-vs-jump split (HAR-CJ), an intraday seasonality profile on the broker clock, a cross-pair market volatility factor (27 G8 pairs), and dummies for 12 scheduled macro events.
  • Out-of-sample R² = 0.665 on log hourly realized volatility (median of 28 G8 pairs), fitted 2018-06..2023-12, scored 2024-01..2026-08. Most of it comes from HAR persistence; the events matter most around releases.
  • Next-hour high-low range (median of 28 pairs, 16,151 hours each): MSE on log range 0.157 vs 0.204 for the best free baseline (same-hour 20-day average) and 0.242 for ATR(14); QLIKE 0.471 vs 0.666 / 0.842.
  • Diebold-Mariano (HAC): significantly better than the best free baseline in 28/28 pairs on MSE (pooled t = 25.0) and 26/28 on QLIKE, never significantly worse. A later 9-week window confirmed the ranking (25/28 pairs on MSE).
  • Release hours: MSE ratio 0.35 vs the best free baseline and 0.20 vs ATR(14).
  • Known limits: no edge on the next-day range (ties ATR(14) D1); one broker's feed; later hours in the histogram are approximations, only the current hour is measured.

What it does not do

  • It forecasts the size of the move, never its direction. It is not a signal.
  • Next hour only. It is no better than ATR for tomorrow's range, so we do not claim that.
  • Back-tested on one broker's data; your broker's ranges may differ. Past accuracy does not guarantee future accuracy.
  • Prototype. The public live track record starts on 19 October 2026 (below); until it has run for a while, every number above is from back-data. Not financial advice.

Live track record

Back-tests can be tuned after the fact. A live log cannot. From Monday 19 October 2026, every hour, a public script commits the forecast for the next hour on 7 major pairs to GitHub before the hour starts, next to the two free yardsticks (ATR(14) and the 20-day same-hour average). After the hour, the actual range is added and scored. The rules were fixed and published in advance, and the results are published whatever they show.

Next big tests (times UTC)

DateEventHour scored
Wed 28 OctBank of Canada13:00-14:00
Wed 28 OctFed (FOMC)18:00-19:00
Thu 29 OctECB13:00-14:00
Fri 30 OctBank of Japan (time varies)night, ~02:00-04:00
Fri 6 NovUS jobs report (NFP)13:00-14:00
Tue 10 NovUS CPI13:00-14:00

Each forecast is committed during the hour before (at :15, :35 or :50). Check the log's commit time yourself, then come back after the hour for the result. We post every one of these, hits and misses.

Loading the live scoreboard…

Rows marked "not scored" had no on-time receipt and are excluded by the preregistered rules; hours with no forecast at all are listed in the log's AMENDMENTS. Data: Dukascopy BID prices. Error = average absolute miss of the hour's high-low range, in pips; lower is better. Release hours = the hour of a scheduled release for one of the pair's two currencies. Code and full log · The rules (preregistration) · Raw scoreboard

Get it first

Free to join. One email at launch, with early-access pricing. No spam.

Join the waitlist

Meanwhile: our automated Excel trading journal.