How High and Low Compare to Yesterday’s High and Low, from Alex Castaldo
I believe there are 4 cases, conventionally called:
h(t) > h(t-1) and l(t) > l(t-1) an uptrend day
h(t) < h(t-1) and l(t) > l(t-1) an inside day
h(t) > h(t-1) and l(t) < l(t-1) an outside day
h(t) < h(t-1) and l(t) < l(t-1) a downtrend day
Victor Niederhoffer writes:
Excuse me?
Alex Castaldo replies:
OK, OK, let's call then upshift day and downshift day then.
;-)
Kim Zussman adds:
What about outie and innie?
Using SPY (2000-present), checked for outside days (today's high > yesterday's high, today's low < yesterday's low). Then checked the return for the next day.
Days after outies were also checked if the market was up-trending (20DMA > 100DMA), and, in addition to up-trending, if the intra-day return was up (C>O). Here are the c-c returns after (vs zero), for all days (c-c ret), all days after outside days outie+1), days after outside days and up-trending (and 20>100), days after outside days and up-trending, and the outside day was up intra-day (and C>O):
One-Sample T: c-c ret, outie+1, and 20>100, and C>O
Test of mu = 0 vs not = 0
Variable N Mean StDev SE Mean 95% CI T
c-c ret 3550 0.00022 0.0131 0.0002 (-0.0002, 0.0006) 1.01
outie+1 353 0.00049 0.0118 0.0006 (-0.0007, 0.0017) 0.77
and 20>100 234 0.00061 0.0090 0.0005 (-0.0005, 0.0017) 1.03
and C>O 89 -0.00047 0.0090 0.0009 (-0.0023, 0.0014) -0.49
Zeroish, with none significant, and days after outside days and up-trending winning the race at 0.06% per day avg.