ETH vs RTH: which session should you actually chart?
RTH — regular trading hours — is the 09:30 to 16:00 ET session. ETH, extended trading hours, is everything around it: the pre-market from 04:00 and the after-hours session to 20:00, or in futures the near-24-hour tape. The choice matters more than most charts admit. Measured on 60 days of five-minute bars, only 40.4% of QQQ’s bars fall inside RTH, a regular-session bar travels about twice as far as an extended-hours one, and on 93.2% of days the high or the low of the full session printed outside regular hours.
Published 2026-08-18 · five-minute bars, 2026-05-22 to 2026-08-17 · every number reproducible from research/measure_eth_rth.py
What do RTH and ETH mean?
RTH (regular trading hours) is the official session an exchange quotes for: 09:30–16:00 ET for US stocks and ETFs. ETH (extended trading hours) covers the pre-market and after-hours windows on either side — for US equities typically 04:00–09:30 and 16:00–20:00 ET. Index futures like ES blur the line further: they trade nearly around the clock, so their “regular session” is a convention laid on top of a continuous tape rather than a period when the market is otherwise shut.
How much of the chart is extended hours?
Most of it. Fewer than half of a US equity’s five-minute bars fall inside regular hours, and for ES futures it is closer to a quarter. If a chart is set to show everything, the majority of what is on screen is not the regular session:
| Symbol | Days | Bars inside RTH | RTH bar range | ETH bar range | Ratio |
|---|---|---|---|---|---|
| QQQ | 59 | 40.4% | 13.7 bp | 6.8 bp | 2.02× |
| SPY | 59 | 41.0% | 8.1 bp | 4.1 bp | 1.97× |
| NVDA | 59 | 40.4% | 26.1 bp | 14.1 bp | 1.85× |
| ES=F | 49 | 27.9% | 8.7 bp | 4.3 bp | 2.02× |
Bar range is the median high-to-low travel of one five-minute bar, in basis points of price.
Is extended-hours price action quieter?
Yes, and by a remarkably stable factor: a regular-session five-minute bar moves roughly twice as far as an extended-hours one — 1.85× to 2.02× across every symbol measured. That ratio is why mixing the two windows quietly distorts anything computed from bar size. Any indicator with a volatility term — average true range, most volatility bands, any zigzag or swing detector with a size threshold — reads a different market depending on which bars are in the series, even though the price data is identical.
Where does the day’s high or low actually print?
Usually outside regular hours. On 93.2% of QQQ days, the extreme of the full extended session — its high, its low, or both — printed outside 09:30–16:00. Because thin pre-market prints can spike a wick without much trading behind them, the same measurement was repeated using only bar closes, which is immune to that; the effect shrinks but does not go away:
| Day’s extreme printed outside RTH | Using bar highs / lows | Using bar closes only |
|---|---|---|
| QQQ | 93.2% | 84.7% |
| SPY | 93.2% | 69.5% |
| NVDA | 100.0% | 76.3% |
| ES=F | 71.4% | 75.5% |
The gap between the two columns is itself information. SPY falls from 93.2% to 69.5% — a good share of its extended-hours extremes were thin prints rather than real trade. QQQ only falls to 84.7%, so most of its overnight range was genuine movement. ES=F barely moves at all, and is the one symbol where the close-based figure is the higher of the two, which fits a market that trades continuously rather than reopening cold.
So why does MADDOG read the regular session?
Because extremes and structure are different questions. The numbers above say the widest price of the day is often set outside regular hours; they do not say the day’s shape is decided there. MADDOG reads the regular session because that is where the patterns it was trained on resolve — its day-type model learned on RTH sessions, and its published accuracy is measured over a complete regular session.
One clarification worth making, because the two figures look like they disagree: MADDOG’s opening-range article reports that 81.2% of sessions have one side of their extreme in by the 90-minute mark. That is measured within the regular session on 16 years of ES data. This article measures the full extended window over 60 days. Different scopes, different samples, no contradiction — and neither number should be quoted as if it were the other.
Which session should you use?
Pick the window your method was calibrated on, and never change it halfway. If a strategy’s statistics were built on regular-hours bars, charting extended hours will hand it a different volatility baseline and a different set of swing points. If it was built on the continuous tape, cutting to regular hours does the same in reverse. The practical rule: use extended hours to see where price has already been overnight, and regular hours to judge what the session is doing — but measure in one window only.
How this was measured
Five-minute bars including pre- and post-market, bucketed in US Eastern time, with RTH defined as 09:30–16:00 — the same window MADDOG’s own pipeline uses. A day counts only if it has at least 60 regular-session bars and some extended-hours trade, which drops half-days and gaps in the data. Sample: 59 days for the equities and 49 for ES, from 2026-05-22 to 2026-08-17.
That 60-day ceiling is the limit of freely available five-minute history, and it is a far smaller sample than the 16-year ES work behind MADDOG’s day-type statistics. Treat these as well-measured facts about a recent quarter, not as multi-year base rates. The script that produces every number above ships in the repository as research/measure_eth_rth.py.
See a live read
MADDOG reads the regular session 90 minutes after each market’s open and returns five probabilities for how the day could finish. Read real sample emails, or leave an email and one ticker on the homepage for one free read — no account, no card.