Two Bull Hours at the Open: The Day Closes Up 84.5% of the Time, but Buying Then Pays Off Only 56%

I have a CSV of cleaned 1-minute ES futures bars (Databento's ES.c.0 continuous contract, with columns ts_event, open, high, low, close; timestamps in UTC). Please write Python code that measures what happens on days when the first N RTH bars are all bull bars (or all bear bars). Rules:
1. Convert timestamps to US Eastern time (US/Eastern), handling daylight saving time automatically. Use RTH only, 09:30–16:15, meaning 1-minute bars from 09:30 to 16:14.
2. Build 5, 15, 30 and 60-minute bars from the 1-minute bars, starting at 09:30; a short last bar still counts (for example, the last 60-minute bar is 15:30–16:15). Each bar's open is its first minute's open and its close is its last minute's close.
3. Bull bar: close > open; bear bar: close < open; equal counts as neither.
4. Signal: the first N bars of the day are all bull bars (or all bear bars), with N = 1 to 8 for 5 and 15-minute bars and N = 1 to 6 for 30 and 60-minute bars.
5. For each signal, measure four things (shown for bull bars; reverse everything for bear bars):
   - The day closes up: RTH close > RTH open;
   - Close above the signal close: RTH close > bar N's close, i.e. buying at bar N's close would be profitable at the close;
   - The low of the day is already in: the day's RTH low is inside the first N bars;
   - New high later: after bar N, some bar's high is above the highest high of the first N bars.
6. Print the share of all days closing up and down; then for each bar size, each N, and bull and bear separately, print the number of signal days, their share of all days, and the four shares above.
7. Finally, look separately at days when the first 6 sixty-minute bars are all bull bars: bar 6 closes at 15:30. Print the number of such days, the share closing above the 15:30 close, the number of days falling more than 0.3% from 15:30 to the close, and the median distance from the close to the day's high as a share of the day's range.
Please add comments to every step explaining why it's done that way.

# Consecutive same-color bars at the open: when the first N RTH bars are all bull (bear) bars, how does the day close?
# Input: cleaned 1-minute ES bars (Databento ES.c.0, timestamps in UTC)
import numpy as np
import pandas as pd

df = pd.read_csv("ES-1min_clean.csv", usecols=["ts_event", "open", "high", "low", "close"])
t = pd.to_datetime(df.ts_event, utc=True).dt.tz_convert("US/Eastern").dt.tz_localize(None)
m = t.dt.hour * 60 + t.dt.minute - 570                      # minutes since 09:30
rth = df[(m >= 0) & (m < 405)].assign(date=t.dt.normalize(), m=m)   # RTH 09:30–16:15

day = rth.groupby("date").agg(o=("open", "first"), h=("high", "max"), l=("low", "min"), c=("close", "last"))
print(f"Sample {len(day)} days, closed up {100 * (day.c > day.o).mean():.1f}%, closed down {100 * (day.c < day.o).mean():.1f}%")

for T in [5, 15, 30, 60]:                                   # bar size (a short last bar still counts, e.g. 15:30–16:15 on the 60-minute chart)
    k = rth.assign(k=rth.m // T).groupby(["date", "k"])
    bars = k.agg(o=("open", "first"), h=("high", "max"), l=("low", "min"), c=("close", "last"))
    color = np.sign(bars.c - bars.o).unstack()              # one row per day: is bar 1, 2, 3, … bull (+1) or bear (−1)
    hi = bars.h.unstack().cummax(axis=1)                    # highest high up to bar N
    lo = bars.l.unstack().cummin(axis=1)
    close = bars.c.unstack()
    for N in range(1, (8 if T <= 15 else 6) + 1):            # same as the interactive chart: up to bar 8 for 5 and 15 minutes, bar 6 for 30 and 60
        for s, name in [(1, "bull"), (-1, "bear")]:
            sig = (color.iloc[:, :N] == s).all(axis=1)      # the first N bars are all bull (bear) bars
            d = day[sig]
            sig_close = close.iloc[:, N - 1][sig]           # signal close = the close of bar N
            hiN, loN = hi.iloc[:, N - 1][sig], lo.iloc[:, N - 1][sig]
            if s > 0:
                same = d.c > d.o                             # the day closes up
                beyond = d.c > sig_close                     # closes above the signal close: buying at bar N's close would be profitable at the close
                opp = loN == d.l                             # the low of the day is already inside the first N bars
                more = d.h > hiN                             # price makes a new high later
            else:
                same, beyond, opp, more = d.c < d.o, d.c < sig_close, hiN == d.h, d.l < loN
            print(f"{T:2d} min, first {N} bar{'s' if N > 1 else ''} all {name}: {sig.sum():4d} days ({100 * sig.mean():.1f}%)"
                  f"  day same color {100 * same.mean():.1f}%  close beyond signal close {100 * beyond.mean():.1f}%"
                  f"  opposite extreme already in {100 * opp.mean():.1f}%  new {'high' if s > 0 else 'low'} later {100 * more.mean():.1f}%")

# First 6 sixty-minute bars all bull: after bar 6 closes at 15:30, what happens in the last 45 minutes
k = rth.assign(k=rth.m // 60).groupby(["date", "k"])
h60 = k.agg(o=("open", "first"), c=("close", "last"))
color, close = np.sign(h60.c - h60.o).unstack(), h60.c.unstack()
six = (color.iloc[:, :6] == 1).all(axis=1)
d, c6 = day[six], close.iloc[:, 5][six]
tail = (d.c - c6) / c6 * 100                                # change from 15:30 to the close, in %
print(f"\nFirst 6 sixty-minute bars all bull: {six.sum()} days; closed above 15:30 {100 * (tail > 0).mean():.1f}%, "
      f"fell more than 0.3% into the close on {(tail < -0.3).sum()} days; median distance from the close to the day's high {100 * ((d.h - d.c) / (d.h - d.l)).median():.0f}% of the day's range")