The headline: playing the daily stocks-in-play instead of a fixed list is decisive — and for the first time the edge survives out-of-sample. The disciplined stack (in-play + morning + daily-trend + matched exits) held: TRAIN +0.36R → TEST +0.12R (test N=1,483, 40% win). Three plays hold strongly. LAB — not live-confirmed; trade small first.
Plain-English key (read this first)
R How much you planned to lose if the trade went wrong. Every result is counted in R. "+0.15R" = on average you made 15¢ for every $1 you risked.
Avg R The average result of all the trades, in R. Above 0 = winning overall; below 0 = losing.
Win % Out of every 100 trades, how many made money.
SIP (stocks in play) The stocks that are actually busy and moving that day — not the sleepy ones. You trade only these.
Selective Only the best trades: an in-play stock, in the morning, going the SAME way as the daily trend.
Train The older data we studied to build the rules — like a practice test.
Test (out-of-sample) Newer data we hid from ourselves, then graded cold — the real exam. THIS is the number that counts.
Holds Positive on BOTH the train and the test data — a sign it's real, not luck.
① The progression — how each change moved the number
Start with everything, then add one discipline rule at a time. Watch the edge climb as the trade count drops — fewer, better trades. "$ / 100 trades" = dollars made per 100 trades, each risking $100.
What we changed
Trades
Avg R
Win
$ / 100 trades
Lift
Baseline — fixed mega-cap list, all plays, all day
12,988
+0.023
34%
$+231
—
+ trade the daily stocks-in-play (not a fixed list)
32,227
+0.100
36%
$+998
+0.077
+ only the morning (9:30–11:00)
7,245
+0.241
42%
$+2,407
+0.141
+ only with the daily trend (price vs its 20-day)
3,520
+0.258
42%
$+2,581
+0.017
+ only the 4 durable plays (drop the losers)
2,183
+0.368
42%
$+3,685
+0.110
From basically break-even (+0.02R, a coin-flip 34% win → ~$230 per 100 trades) to +0.37R (~$3,700) — roughly a 16× swing in profit, by trading less but better. The two biggest single levers: trading the morning (+0.14R) and keeping only the durable plays (+0.11R). Crucially it holds out-of-sample (§②) — so the realistic live expectation is the TEST number, about +0.17R (~$1,700/100), not the full +0.37.
the old fixed-list vs in-play A/B (what step 1 above replaced)
Method
Signals
Avg R
Win %
OLD — fixed 8 mega-caps, all day
12,988
+0.023
34%
NEW — daily SIP in-play, all day
32,227
+0.100
36%
OLD — fixed 8, selective (morning+daily-aligned)
1,332
+0.061
39%
NEW — SIP, selective (morning+daily-aligned)
3,520
+0.258
42%
② The out-of-sample verdict (the one that counts)
Trained on 2026-02-27→2026-05-20 (first ~2/3), tested on 2026-05-20→2026-07-02 (held-out last ~1/3). A play is real only if it's positive in both halves. STRONG = test ≥ +0.10R on a real sample.
Play (SIP, selective)
Train N
Train R
Test N
Test R
Verdict
backside
2
+0.666
0
+0.000
no
backside_roll
27
-0.259
30
+0.029
no
band_break
138
+0.534
83
+0.039
weak
open_drive
83
+0.332
31
+0.114
STRONG
revbreak_long
781
+0.597
532
+0.147
STRONG
trend_ride
291
+0.456
167
+0.207
STRONG
v_bottom
234
-0.019
162
+0.016
no
vwap_pull_in
368
+0.050
293
+0.054
weak
vwap_reject_short
113
+0.231
185
+0.192
STRONG
Overall selective SIP stack: TRAIN 2,037/+0.361 → TEST 1,483/+0.118. The long trend/reversal plays (revbreak, trend_ride) and the short (vwap_reject_short) are the durable core.
③ Every play, in plain English
All 9 plays the backtest runs. Click any play name → its labeled chart + exact specs + the real code open right here (no page change). Or the full Plays page.
Play
Side
What it is
Out-of-sample
vwap_reject_short ↗
SHORT
Price is below VWAP, pops up to it, but fails and closes back under. You short the failure. (Your MU trade.)
HOLDS
revbreak_long ↗
LONG
Price is falling, breaks its downtrend, pulls back to test the breakout, then goes. You buy the turn. (Your afternoon reversal.)
HOLDS
trend_ride ↗
LONG
In a clear uptrend, price pulls back to its fast average, then takes off again. You buy the continuation.
HOLDS
band_break ↗
LONG
Price sits calmly above VWAP, coils tight, then bursts up through the top band. You ride the pop.
weak
open_drive ↗
LONG
In the first hour, price squeezes into a tight bar then breaks out upward. You buy the breakout.
HOLDS
vwap_pull_in ↗
LONG
Price dips just under VWAP, then closes back above it. You buy the recovery into value.
weak
v_bottom ↗
LONG
Price drops hard to the bottom band, then snaps back up in a “V.” You buy the rebound.
no
backside_roll ↗
SHORT
Price is stretched way above VWAP and slowly rolls over, losing its rising average. You short the fade back to VWAP.
no
backside ↗
SHORT
Price stretches far above VWAP and slaps down with a rejection candle. You short the snap-back to VWAP.
no
④ Honest caveats
• Test < train (+0.36 → +0.12) — real edge decays out-of-sample; expect live to look like the test number, not the train.
• One split, ~90 days. A single held-out period — more periods + more names would harden it.
• Free-tier skips: some stock-days were dropped on rate limits (partial coverage).
• "Liquid-mover" reconstruction (top $-volume names that moved), a proxy for the real SIP scanner's catalyst weighting.
• LAB only — nothing promoted to the Pine; trade small/live by hand before trusting it.
⑤ What's next · and what we tried and rejected
Next: automate the cockpit + a real push-alert when a play fires, and build the MTF higher-timeframe "which-setups-actually-run" selection gate. What we rejected (kept off this page on purpose — this page is the durable results): the inverse mirror-shorts (regime artifact), the v_bottom hybrid exit (refuted), the backside long (falling knife). The full journey incl. every dead end is in the Experiment Log, and which stocks make these plays pay is in Nuance.
Generated from sip90.py → sip90.csv, rendered by j3p_sip90_page.py. Reconstructed daily SIP via Massive grouped-daily. 2026-07-05.
vwap_reject_shortSHORTSTRONG OOS
1below VWAP, downtrend→2pops up, tags VWAP→3rejects, closes back below → SHORT→4falls to −2σ / −4σ
▶ DO: SHORT the VWAP-loss close · stop above the pop · cover the fade into −2σ / −4σ (mean-revert)
🔍 Full detail — specs · management · code
Exact specs
Context: price below session VWAP, down-context (VWAP slope ≤ 0 OR price below the 2-day AVWAP).
Trigger: this bar CLOSES back below VWAP after the prior bar was at/above it (a cross-down), sitting in the VWAP→−1σ zone with −2σ below the close.
Entry = the VWAP-loss close · Stop = above the shakeout high / 2-day AVWAP +0.25·ATR · T1 = −2σ (60%) · T2 = −4σ (20%) · runner covers into VWAP.
Management plans (pick by context)
Plan A — cover the fade (default): Cover into the −2σ target — a mean-reversion short, so bank it. Stop above the recent high / a reclaim of VWAP.
Invalidation: Reclaim of VWAP (the short is wrong). · +0.19 practice / +0.11 fresh — the one short that held OOS
Plan B — topping-short trail (multi-day): If it's a multi-day top losing its daily MAs, take partial at −2σ and TRAIL the falling 20-day down (the MU / AMZN swing case). Add only on failed retests of the falling MA.
Invalidation: Reclaim of the falling 20-day. · MU 7/2 short rode toward the 50-DMA — see the Swing-DMA page
How it's implemented (from the scan code)
cross_dn = (r.close < r.vwap) and (p.close >= p.vwap) # loses VWAP from above
below_multi = (avwap2 is not None) and (r.close < avwap2)
down_ctx = (r.vwap_slope <= 0) or below_multi
value_zone = (r.close >= r.lb2) and (r.close <= r.vwap) # VWAP→−1σ
sigma_below = r.lb3 < r.close # −2σ below close
if cross_dn and down_ctx and value_zone and sigma_below:
entry = r.close; stop = shake_high + BK_STOP_BUF*atr
t1 = r.lb3; t2 = r.lb5
revbreak_longLONGSTRONG OOS
1downtrend→2breaks the trend up→3pulls back, retests the break→4holds → BUY, runs
▶ DO: BUY the retest-and-go · stop under the retest low · ride the 20-EMA (the band is target #1)
🔍 Full detail — specs · management · code
Exact specs
Context: an intraday downtrend (a real down-leg below the recent resistance).
Trigger: price breaks ABOVE the prior resistance in the last few bars, then this bar pulls back to RETEST that level and closes back above it (an up bar) — the retest holds.
Entry = the retest-and-go close · Stop = below the retest low −0.1·ATR · T1 = entry +1.5·ATR · T2 = entry +3·ATR · trend-trail the runner.
Management plans (pick by context)
Plan A — EMA ride (default): Ride the whole position. After the +2σ band tags, move the stop to BREAKEVEN; exit 100% on the first 2-min close back through the 20-EMA.
Invalidation: Close under the 20-EMA once trending (or the initial stop before that). · +$1,471 per 100 — ride-all beat cap-at-band ~7× (fresh data)
Plan B — 1/3 partial + ride: Take 1/3 off at the +2σ band, move the stop to breakeven, ride the other 2/3 to the 20-EMA break. Smoother, higher hit-rate, a bit less total.
Invalidation: Same 20-EMA break; the breakeven stop protects the runner. · +$879 per 100 (46% win vs 42%)
⛰ If it's running into a known CEILING (200-day, prior high, anchored VWAP, supply wall) → switch to momo-burst: sell 50–70% into the burst + MANDATORY trim at the ceiling (+5.22R on IBIT). Never hold for a target sitting in overhead supply.
How it's implemented (from the scan code)
resist = high[i-10:i-2].max() # prior down-leg resistance
dipped = low[i-8:i].min() < resist - 1.0*a # a real down-leg happened
broke = max(close[i-3],close[i-2],close[i-1]) > resist # broke above it
retest = (low[i] <= resist+0.15*a) and (close[i]>resist) and (close[i]>open[i])
if dipped and broke and retest:
entry=close[i]; stop=min(low[i-1],low[i])-0.1*a
t1=entry+1.5*a; t2=entry+3.0*a
trend_rideLONGSTRONG OOS
1uptrend (5>20>50)→2pulls back to 5-MA→3resumes → BUY→4continues up
▶ DO: BUY the pullback that resumes · stop under the pullback low · ride the 20-EMA
Trigger: the prior bar pulls back to the 5-MA, then this bar breaks the pullback-bar high, closes up and above the 5-MA. RVOL ≥ 1.2.
Entry = the go bar's close · Stop = below the pullback low −0.1·ATR · T1 = +2σ · T2 = +4σ · trend-trail the runner.
Management plans (pick by context)
Plan A — EMA ride (default): Ride the whole position. After the +2σ band tags, move the stop to BREAKEVEN; exit 100% on the first 2-min close back through the 20-EMA.
Invalidation: Close under the 20-EMA once trending (or the initial stop before that). · +$1,471 per 100 — ride-all beat cap-at-band ~7× (fresh data)
Plan B — 1/3 partial + ride: Take 1/3 off at the +2σ band, move the stop to breakeven, ride the other 2/3 to the 20-EMA break. Smoother, higher hit-rate, a bit less total.
Invalidation: Same 20-EMA break; the breakeven stop protects the runner. · +$879 per 100 (46% win vs 42%)
⛰ If it's running into a known CEILING (200-day, prior high, anchored VWAP, supply wall) → switch to momo-burst: sell 50–70% into the burst + MANDATORY trim at the ceiling (+5.22R on IBIT). Never hold for a target sitting in overhead supply.
How it's implemented (from the scan code)
up_regime = (r.close>r.vwap) and (r.vwap_slope>0) and bool(r.ma_align)
established = r.close > r.ub1 # already above +0.5σ
pulled_back = p.low <= p.ma5 # prior bar tagged 5-MA
go = (r.close>p.high) and (r.close>r.open) and (r.close>r.ma5)
if up_regime and established and pulled_back and go and rvol>=1.2:
entry=r.close; stop=min(p.low,r.low)-ATR_BUF_MULT*atr; t1=r.ub3; t2=r.ub5
band_breakLONGweak OOS
1coils above VWAP→2breaks the top band→3BUY→4pops up
▶ DO: BUY the band break · stop under the coil · ride it, trim on MA breaks
🔍 Full detail — specs · management · code
Exact specs
Context: a 5-bar coil HOLDING above VWAP (≥60% of closes above), range ≤ 2·ATR, prior bar still inside the coil (fresh).
Trigger: this bar closes decisively above the coil high (> coil-high +0.1·ATR), is an up bar, and closes at/above +0.5σ. VWAP rising, RVOL ≥ 1.2.
Entry = the breakout close · Stop = below the coil low −0.1·ATR · T1 = +2σ · T2 = +4σ · trend-trail the runner.
Management plans (pick by context)
Plan A — EMA ride (default): Ride the whole position. After the +2σ band tags, move the stop to BREAKEVEN; exit 100% on the first 2-min close back through the 20-EMA.
Invalidation: Close under the 20-EMA once trending (or the initial stop before that). · +$1,471 per 100 — ride-all beat cap-at-band ~7× (fresh data)
Plan B — 1/3 partial + ride: Take 1/3 off at the +2σ band, move the stop to breakeven, ride the other 2/3 to the 20-EMA break. Smoother, higher hit-rate, a bit less total.
Invalidation: Same 20-EMA break; the breakeven stop protects the runner. · +$879 per 100 (46% win vs 42%)
⛰ If it's running into a known CEILING (200-day, prior high, anchored VWAP, supply wall) → switch to momo-burst: sell 50–70% into the burst + MANDATORY trim at the ceiling (+5.22R on IBIT). Never hold for a target sitting in overhead supply.
How it's implemented (from the scan code)
win=dt.iloc[i-K:i]; coil_hi=win.high.max(); coil_lo=win.low.min()
held=(win.close>win.vwap).mean()>=0.6; contained=(coil_hi-coil_lo)<=2*atr
fresh=p.close<=p_coil_hi
broke=(r.close>coil_hi+0.1*atr) and (r.close>r.open) and (r.close>=r.ub1)
if held and contained and fresh and broke and r.close>r.vwap and r.vwap_slope>=0 and rvol>=1.2:
entry=r.close; stop=coil_lo-ATR_BUF_MULT*atr; t1=r.ub3; t2=r.ub5
open_driveLONGweak OOS
1inside / coil bar→2breaks up→3BUY (open window)→4drives up
▶ DO: BUY the inside-bar break (open only) · stop under the range · ride the 20-EMA
🔍 Full detail — specs · management · code
Exact specs
Window: OPEN only (before 10:30 ET).
Trigger: the prior bar is an inside bar (its high < the bar-before's high AND its low > that low); this bar breaks and closes above the inside-bar high, close > VWAP, up bar. RVOL ≥ 1.5.
Entry = the breakout close · Stop = under the inside-bar low −0.1·ATR · T1 = entry +2·ATR · T2 = entry +4·ATR.
Management plans (pick by context)
Plan A — EMA ride (default): Ride the whole position. After the +2σ band tags, move the stop to BREAKEVEN; exit 100% on the first 2-min close back through the 20-EMA.
Invalidation: Close under the 20-EMA once trending (or the initial stop before that). · +$1,471 per 100 — ride-all beat cap-at-band ~7× (fresh data)
Plan B — 1/3 partial + ride: Take 1/3 off at the +2σ band, move the stop to breakeven, ride the other 2/3 to the 20-EMA break. Smoother, higher hit-rate, a bit less total.
Invalidation: Same 20-EMA break; the breakeven stop protects the runner. · +$879 per 100 (46% win vs 42%)
⛰ If it's running into a known CEILING (200-day, prior high, anchored VWAP, supply wall) → switch to momo-burst: sell 50–70% into the burst + MANDATORY trim at the ceiling (+5.22R on IBIT). Never hold for a target sitting in overhead supply.
How it's implemented (from the scan code)
is_inside = (p.high<p2.high) and (p.low>p2.low) # inside/coil bar
broke = r.high>p.high and r.close>p.high
if mins<OPEN_END and is_inside and broke and r.close>r.vwap and r.close>r.open and rvol>=RVOL_THRESH:
entry=r.close; stop=p.low-ATR_BUF_MULT*atr
t1=entry+2*atr; t2=entry+4*atr
▶ DO: BUY the VWAP recapture · stop under the shakeout · ride to +2σ / +4σ
🔍 Full detail — specs · management · code
Exact specs
Context: VWAP slope ≥ 0.
Trigger: price CLOSES back above VWAP after the prior bar was at/below it (a cross-up), sitting in the VWAP→+1σ value zone with +2σ above the close.
Entry = the recapture close · Stop = below the shakeout low −0.1·ATR · T1 = +2σ (60%) · T2 = +4σ (20%) · runner trailed.
Management plans (pick by context)
Plan A — take the target (default): Mean-reversion: there's no trend to ride, so take the +2σ target. Stop below the shakeout low; move to breakeven after the first partial.
Invalidation: Loss of the entry level / the shakeout low. · marginal edge — bank it, don't hold
How it's implemented (from the scan code)
cross_up = (r.close > r.vwap) and (p.close <= p.vwap) # recaptures VWAP
recapture = cross_up and (r.vwap_slope>=0) and (r.ub3>r.close) and (r.close<=r.ub2)
if recapture:
entry=r.close; stop=r.shake_low-ATR_BUF_MULT*atr
t1=r.ub3; t2=r.ub5
v_bottomLONGno OOS
1flush to −2σ→2V-turn (5-MA up)→3reclaims → BUY→4to VWAP
▶ DO: BUY the V-reclaim, small · stop under the V low · T1 = VWAP (mean-revert, don't chase)
🔍 Full detail — specs · management · code
Exact specs
Context: a real flush — the last ~5 bars' low reaches ≤ −2σ.
Trigger: the 5-MA slope turns UP (a V), and this bar closes back above the 5-MA on an up bar, still ≤ +0.5σ (not extended). No RVOL gate — the reclaim is a quiet bar.
Entry = the reclaim close · Stop = below the V low (3-bar) −0.1·ATR · T1 = VWAP · T2 = +1σ · runner trailed.
Management plans (pick by context)
Plan A — take the target (default): Mean-reversion: there's no trend to ride, so take the VWAP, then +1σ target. Stop below the shakeout low; move to breakeven after the first partial.
Invalidation: Loss of the entry level / the shakeout low. · marginal edge — bank it, don't hold
How it's implemented (from the scan code)
flushed = dt.low[i-5:i+1].min() <= r.lb3 # ≤ −2σ capitulation
ma5_vturn = (r.ma5>p.ma5) and (p.ma5<=p2.ma5) # 5-MA slope turns up
reclaim = (r.close>r.ma5) and (r.close>r.open)
if flushed and ma5_vturn and reclaim and r.close<=r.ub1:
entry=r.close; stop=v_low-ATR_BUF_MULT*atr; t1=r.vwap; t2=r.ub2
backside_rollSHORTno OOS
1extended above VWAP→2rolls over→3loses 5-MA → SHORT→4fades to VWAP
▶ DO: SHORT the lost 5-MA when extended · stop above the high · COVER at VWAP (fade)
🔍 Full detail — specs · management · code
Exact specs
Context: extended ≥ 2·ATR above VWAP, and the 5-MA had been rising (vs 3 bars ago).
Trigger: price just LOST the 5-MA (crossed below it) on a down bar. RVOL ≥ 1.0.
Entry = the close · Stop = above the recent high +0.25·ATR · Exit = cover 100% at VWAP (a mean-reversion fade — do NOT trend-trail).
Management plans (pick by context)
Plan A — cover into VWAP (default): Backside fade: cover into VWAP (the mean). Stop above the recent high + buffer. A scalp — take it.
Invalidation: Push back above the recent high. · not durable OOS — smallest size, quick
How it's implemented (from the scan code)
extended = (r.close-r.vwap) >= BK_EXT_ATR*atr
ma5_rising = r.ma5 > dt.iloc[i-3].ma5
lost_5 = broke_5 or (p.close>=p.ma5 and r.close<r.ma5)
if extended and ma5_rising and lost_5 and (r.close<r.open) and rvol>=1.0:
entry=r.close; stop=r.recent_high+BK_STOP_BUF*atr; tgt=r.vwap
backsideSHORTno OOS
1≥2σ extended→2rejection candle→3SHORT→4snaps to VWAP
▶ DO: SHORT the rejection candle when extended · stop above the high · COVER at VWAP (fade)
🔍 Full detail — specs · management · code
Exact specs
Context: extended ≥ 2·ATR above VWAP, 5-MA rolling over (or just lost).
Trigger: a rejection candle — took the prior 5-bar high, closed down, and closed in the lower 40% of its range. RVOL ≥ 1.5.
Entry = the close · Stop = above the recent high +0.25·ATR · Exit = cover 100% at VWAP (mean-reversion fade).
Management plans (pick by context)
Plan A — cover into VWAP (default): Backside fade: cover into VWAP (the mean). Stop above the recent high + buffer. A scalp — take it.
Invalidation: Push back above the recent high. · not durable OOS — smallest size, quick
How it's implemented (from the scan code)
extended = (r.close-r.vwap) >= BK_EXT_ATR*atr
rej = (r.high>=r.recent_high_prev and r.close<r.open
and r.close < (r.high-(r.high-r.low)*0.4)) # rejection candle
if extended and (ma_roll or broke_5) and rej and rvol>=RVOL_THRESH:
entry=r.close; stop=r.recent_high+BK_STOP_BUF*atr; tgt=r.vwap