OPEN-SOURCE SCRIPT
已更新 Day-Extreme Reversal Signals (SPY 5m)
What you'll see on the chart
5-minute scouts (small triangles + small labels). When a bar makes a new session low with one of the study shapes, it's marked with its honest hold-rate: 5m RevClose: holds 28% (base 14%) (green bar closing in its top third at the low — the most common day-low shape), 5m Spring: holds 17% (big lower wick, close held up), 5m V-confirm: holds 27% (green bar defending the prior bar's flush). Deliberately subtle styling — at 17–28% these are scouts, not entries. High-side scouts get the same treatment with their (weak) 11–12% numbers.
Hourly signals (full background band + large label). When a completed hourly bar prints one of the shapes at a session extreme, the next 12 five-minute bars get a colored band and a loud label: HOURLY RevClose at session low: DAY LOW 64% (base 35%). This is the real signal tier — teal bands for lows, orange for highs (with the high label honestly adding "expect a retest attempt," since even the best high pattern only holds 47%). Alerts fire on both hourly signal types.
The decision panel (top-right)
- Live session state: current session low/high, when each was set, and — the piece I think you'll use most — "Odds the day extreme has ALREADY printed: low ~X%, high ~Y%", interpolated live from the measured time-of-day curves as the session progresses. At 10:30 ET it'll say the low is ~47% likely already in but the high only ~32%.
- Hourly forming row: what shape the current, incomplete hourly bar is building (labeled provisional — it only becomes a signal at the hour close).
- The two structural rules, always visible: lows are V-shaped — 55% never retest within 30 minutes, don't wait for confirmation; highs retest 82% of the time — first touch is rarely final, exits can be patient.
- Timing reference (lows in by 10:00/10:30/11:30 on 37/47/60% of days; highs skew late) and the honesty footer distinguishing scout-grade from signal-grade percentages.
Design honesty, as always
The percentages are P(that extreme holds as the day's extreme) from real-time simulation on 60 days of 5m and 730 days of hourly data — the number you actually face in the moment, which is why they're much lower than the hindsight lifts from the study (a 3.9× lift collapsed to 28% precision once you can't peek at the future). None of this has been tested as an entry system with P&L — it's decision support for the scalps you're already taking within the daily-bias framework. The natural marriage: an hourly RevClose band firing while the bias panel shows an active long tier is the highest-alignment moment this whole project can currently produce. If you want, the next study can quantify exactly that intersection — hold-rate and forward return when hourly low-signals fire during active T1–T3 windows.
5-minute scouts (small triangles + small labels). When a bar makes a new session low with one of the study shapes, it's marked with its honest hold-rate: 5m RevClose: holds 28% (base 14%) (green bar closing in its top third at the low — the most common day-low shape), 5m Spring: holds 17% (big lower wick, close held up), 5m V-confirm: holds 27% (green bar defending the prior bar's flush). Deliberately subtle styling — at 17–28% these are scouts, not entries. High-side scouts get the same treatment with their (weak) 11–12% numbers.
Hourly signals (full background band + large label). When a completed hourly bar prints one of the shapes at a session extreme, the next 12 five-minute bars get a colored band and a loud label: HOURLY RevClose at session low: DAY LOW 64% (base 35%). This is the real signal tier — teal bands for lows, orange for highs (with the high label honestly adding "expect a retest attempt," since even the best high pattern only holds 47%). Alerts fire on both hourly signal types.
The decision panel (top-right)
- Live session state: current session low/high, when each was set, and — the piece I think you'll use most — "Odds the day extreme has ALREADY printed: low ~X%, high ~Y%", interpolated live from the measured time-of-day curves as the session progresses. At 10:30 ET it'll say the low is ~47% likely already in but the high only ~32%.
- Hourly forming row: what shape the current, incomplete hourly bar is building (labeled provisional — it only becomes a signal at the hour close).
- The two structural rules, always visible: lows are V-shaped — 55% never retest within 30 minutes, don't wait for confirmation; highs retest 82% of the time — first touch is rarely final, exits can be patient.
- Timing reference (lows in by 10:00/10:30/11:30 on 37/47/60% of days; highs skew late) and the honesty footer distinguishing scout-grade from signal-grade percentages.
Design honesty, as always
The percentages are P(that extreme holds as the day's extreme) from real-time simulation on 60 days of 5m and 730 days of hourly data — the number you actually face in the moment, which is why they're much lower than the hindsight lifts from the study (a 3.9× lift collapsed to 28% precision once you can't peek at the future). None of this has been tested as an entry system with P&L — it's decision support for the scalps you're already taking within the daily-bias framework. The natural marriage: an hourly RevClose band firing while the bias panel shows an active long tier is the highest-alignment moment this whole project can currently produce. If you want, the next study can quantify exactly that intersection — hold-rate and forward return when hourly low-signals fire during active T1–T3 windows.
發行說明
small change so that I can show the actual chart開源腳本
秉持TradingView一貫精神，這個腳本的創作者將其設為開源，以便交易者檢視並驗證其功能。向作者致敬！您可以免費使用此腳本，但請注意，重新發佈代碼需遵守我們的社群規範。
Get more in-depth analysis on my website: joerodtrades.substack.com
免責聲明
這些資訊和出版物並非旨在提供，也不構成TradingView提供或認可的任何形式的財務、投資、交易或其他類型的建議或推薦。請閱讀使用條款以了解更多資訊。
開源腳本
秉持TradingView一貫精神，這個腳本的創作者將其設為開源，以便交易者檢視並驗證其功能。向作者致敬！您可以免費使用此腳本，但請注意，重新發佈代碼需遵守我們的社群規範。
Get more in-depth analysis on my website: joerodtrades.substack.com
免責聲明
這些資訊和出版物並非旨在提供，也不構成TradingView提供或認可的任何形式的財務、投資、交易或其他類型的建議或推薦。請閱讀使用條款以了解更多資訊。