Rules note

Every rule idea we talk about goes here, with the exact wording, where it came from, and its status. Status: IDEA = written down, not tested; TESTING; NO EDGE = tested, did not hold; WORKS = held on 2 years (both halves) and beats cost; LIVE = in a bot. Tests run on the 30 days first; the 2 years only when you say so. Nothing here is live unless it says LIVE.

RN-30DV1 exit: sell if no close above VWAP within N minutesTESTING
RN-29DV1 exit: sell when EMA9 falls clearly under EMA20TESTING
RN-28DV1 exit: EMA9 flattens -> sell unless buyers + heavy volumeNO EDGE
RN-27Don't chase far above the crossNO EDGE
RN-26Volume must support the V before a cross buyTESTING
RN-25Opening buy only if the next bar opens higher than the trigger barNO EDGE
RN-24Avoid cross buys while the EMAs are compressedNO EDGE
RN-23Green-bar check on every buy triggerNO EDGE
RN-22After a cross + sideways run, buy the breakout with state RECLAIMNO EDGE
RN-21Second leg of a deep V counts as the same deep VNO EDGE
RN-20Re-cross after a deep V counts as the same deep VTESTING
RN-19Never buy while the stock is ASLEEPNO EDGE
RN-18Opening style: after the first 5 bars, RUNNING and EMA9 > EMA20 > EMA50 = buyTESTING
RN-17Shallow cross after 11+ asleep bars: wait for RECLAIM, then buyNO EDGE
RN-16At the cross EMA50 must not be fallingNO EDGE
RN-15Asleep at the cross: wait for RECLAIMNO EDGE
RN-14BASING at the buy signal: green candle confirm, else wait for 2 green barsNO EDGE
RN-13Deep-V restart: main entry = EMA9 crosses EMA20 after a deep VLIVE
RN-12After a shallow cross, buy when the state changes to RUNNINGTESTING
RN-11Looser deep V for the morningIDEA
RN-10Price vs EMA50 at the buyIDEA
RN-9Skip cross-during-reclaim buys before 10:30IDEA
RN-8Block when resistance is far above (2%+)IDEA
RN-7Block the buy if the gap to resistance is under 0.8%NO EDGE
RN-6Cross during reclaimNO EDGE (best tilt so far)
RN-5Reclaim after 6+ basing/asleep barsNO EDGE
RN-4EMA cross with BASING as the gateNO EDGE
RN-3Steep V rises, flat V fadesNO EDGE
RN-2APPS marks: first cross 9:53, second ~12:20, never buy with EMAs + VWAP sloping downIDEA
RN-1Deep-V EMA cross (core)NO EDGE (base replaced by RN-13)
RN-30TESTING2026-11-16

DV1 exit: sell if no close above VWAP within N minutes

Rule
After the DV1 buy, if no finished bar has closed above VWAP within N minutes (buy bar counted as minute 1), sell at the next open. Once a close above VWAP happens, the rule is off for that trade. EV3's stop, give-back and C3 unchanged.
Where it came from
11/16, from the RN-29 VWAP test: of the 221 DV1 trades EV3 stopped out (30 days), 125 never closed above VWAP after the buy. User: 'run it: haven't reached VWAP within N minutes'.
Still to decide
N = 10 / 15 / 20 / 30 / 45 / 60 minutes tested.
Result
FLO 11/16: closed above VWAP at 10:07 (3 min after the buy), so the rule never fires; EV3 stop -2.58% unchanged. | 30-day test 11/16 (469 DV1 buys with trade flow, halves split Oct 26, 0.2% cost; EV3 -0.149 = -0.40 / +0.13): N10 -0.139 (-0.25 / -0.01), N15 -0.167, N20 -0.149, N30 -0.130 (-0.30 / +0.06), N45 -0.123 (-0.32 / +0.10), N60 -0.125 (-0.34 / +0.12). N45/N60 = best DV1 exit so far (+0.025/trade overall; the trades it sells gain +0.12 / +0.15 each), fires on 102 / 74 trades; better in half 1, a little worse in half 2 (-0.04 / -0.015) = close to steady but still mixed; nothing above cost. N60: EV3 stop-outs -1.72 -> -0.81 (39 trades), only 10 winners cut.
RN-29TESTING2026-11-16

DV1 exit: sell when EMA9 falls clearly under EMA20

Rule
On every finished bar after the DV1 buy bar: if EMA9 is more than k own candles under EMA20, sell at the next open. Version 'before arm' = only until EV3's give-back arms at +2%; version 'any' = at any time. EV3's stop, give-back and C3 unchanged.
Where it came from
User 11/16, FLO 11/16: asked for the biggest EMA9-EMA20 gap in the first drop (10:10-10:16). Drop 1: EMA9 only 0.22 own candles under EMA20 at 10:15, back above at 10:18 (shakeout). Drop 2: 0.34 at 10:47, 0.58 at 10:49 and widening (real break). User: 'yes run it'.
Still to decide
k = 0.25 / 0.3 / 0.4 own candles tested.
Result
FLO 11/16: EV3 -2.58% (stop 11:08); k 0.25-0.3 sells 10:48 open -1.20%, k 0.4 10:49 -1.39%; all skip the first drop. | 30-day test 11/16 (469 DV1 buys with trade flow, halves split Oct 26, 0.2% cost): EV3 -0.149 (-0.40 / +0.13), win 42%. k 0.25: -0.197 (-0.27 / -0.11); k 0.3: -0.210 (-0.29 / -0.12); k 0.4: -0.156 (-0.26 / -0.04); 'any time' versions about the same (k 0.4 any -0.153). Fires on 73-81% of trades (median hold 27 min vs 79). Helps the bad half, hurts the good half. On EV3's stop-outs it halves the loss (-1.60 -> -0.82, k 0.3), but on EV3's winners it sells on an early dip (+1.83 -> -0.68). Bigger k got better (0.4 nearly = EV3, smoother halves) -> 0.5-0.8 not yet tested. | 30-day test k 0.5-0.8 (11/16, same 469 buys): k 0.5 -0.138 (-0.30 / +0.05), k 0.6 -0.148 (-0.32 / +0.05), k 0.8 -0.132 (-0.33 / +0.09) vs EV3 -0.149 (-0.40 / +0.13); 'any time' versions ~same. k 0.5 and 0.8 are slightly better overall (+0.01 / +0.02 per trade) and steadier, but still better in half 1 and worse in half 2 = mixed; nothing above cost. k 0.8 fires on 188 trades, median hold 61 min; on EV3 stop-outs -1.73 -> -1.14, on EV3 winners only 14 cut. FLO: k 0.5 sells 10:50 -1.48%, k 0.6-0.8 11:04 -2.21%. | 11/16 user: 'RN-29 just after it crosses VWAP' = rule armed only after a close above VWAP since the buy. 30-day: k 0.25 -0.222, 0.3 -0.224, 0.4 -0.186, 0.5 -0.148 (-0.34 / +0.07), 0.8 -0.164 vs plain RN-29 and EV3 -0.149 -> WORSE than plain at every k. Reason: most stop-outs never reach VWAP (only 96 of 221 did), so the VWAP version loses the main saving, but still sells winners that crossed VWAP.
RN-28NO EDGE2026-11-16

DV1 exit: EMA9 flattens -> sell unless buyers + heavy volume

Rule
On every finished bar after the DV1 buy bar, before EV3's give-back arms (+2%): if EMA9's change over the last bar <= f own candles (flat or turning down), HOLD only if the 5-min buy share >= b AND the last 5 bars' volume >= v x normal for that time of day; otherwise sell at the next open. EV3's stop, give-back and C3 stay as they are.
Where it came from
User 11/16 after FLO 11/16 (DV1 buy 10:04 at 5.48, high 5.53 = +0.9%, give-back never armed, stopped 11:08 at 5.35 = -2.37%): 'create the exit rules for DV1: if EMA9 flats out, check if we still have buy pressure with heavy volume; if not, sell'.
Still to decide
My first settings, all tested: f = 0.0 / 0.1 / 0.2 own candles, b = 50% / 55%, v = 1.0x / 1.5x. Should it also run after the give-back arms?
Result
FLO 11/16: EV3 -2.58% (stop 11:08); every variant sold in the FIRST drop: f 0.1-0.2 at 10:09 -0.47%, f 0.0 at 10:12-10:15 -1.1 to -1.5%. | 30-day test 11/16 (Oct 5 - Nov 13, 469 DV1 buys with trade flow, halves split Oct 26, 0.2% cost): EV3 exit -0.149%/trade (-0.40 / +0.13), win 42%, median hold 79 min. All 12 variants -0.16 to -0.21 (best F0.0B0.5V1.0 -0.158: -0.15 / -0.17), win 22-25%. The rule fires on ~95% of trades with a median hold of 3 min = 'sell at the first pause'. On trades EV3 stopped out (218) it saves a lot (-1.60 -> -0.33), but on EV3's winners (give-back/X1, 120) it turns +1.82 into -0.17. Hold condition (buyers AND heavy volume) is rarely met. Needs a later / softer trigger. | 11/16 winners-vs-losers at the moment the rule fires (446 trades, 30 days; EV3 winners 179 / losers 267): winners had already moved more (peak 0.90 vs 0.50 own candles), were closer to VWAP (-1.0 vs -1.9 candles) and EMA50 (-0.1 vs -0.4), gapped up more (+1.0% vs +0.1%), bought earlier (10:44 vs 11:00 median). Buy share and volume do NOT separate them (5-min 49% vs 49%, 15-min 54% vs 55%, 5-bar volume 0.91x vs 0.85x). No single 'hold if' condition beats the rule alone in both halves by more than +0.04%/trade (best: run under EMA20 <= 32 bars, dip <= 8.9 candles - entry traits, not sell-time traits).
RN-27NO EDGE2026-11-15

Don't chase far above the cross

Rule
A: skip any cross buy (deep, re-cross, 2nd leg, shallow-asleep) if the buy price is more than N own candles above the cross bar's close. B: for the wait-for-RECLAIM buys only (RN-15 asleep, RN-17 shallow + asleep), skip if VWAP was more than N own candles above the close at the cross. N = 2, 3, 4 tested.
Where it came from
User 11/15, CLSK 2025-03-06 buy #6: 'entry is way too high above the cross'. Shallow + asleep cross 14:40 at 8.125, VWAP 8.269 (1.8% above); RN-17 waited for RECLAIM, one big bar to 8.31 at 14:48, buy 14:49 at 8.31 = +2.3% (about 6 own candles) above the cross, -2% first.
Still to decide
Which version and which N.
Result
30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on, halves split at Oct 26; FULL net -0.30 / +0.16, random -0.25 / -0.23). Every variant blocks buys that did WORSE than FULL in both halves, but they are rare (1-3% of buys) so the system total barely moves (+0.01). Blocked buys net (1st / 2nd half): A2 40 buys -0.43 / -0.07; A3 21 buys -0.40 / -0.35 (clearest); A4 16 buys -0.37 / -0.28; B2 22 buys -0.37 / -0.02; B3 13 buys -0.65 / -0.16; B4 12 buys -0.65 / -0.14. Samples tiny (6-20 per half). CLSK 2025-03-06 14:49 blocked by all. 2-year not run (user to decide). | 2-year test 11/15 (A only; Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; FULL net -0.17 / -0.26, random -0.31 / -0.34). Blocked buys net yr1 / yr2: A2 966 buys -0.45 / -0.06; A3 572 buys -0.51 / -0.05; A4 344 buys -0.57 / +0.15 = good to block in year 1, but in year 2 the blocked buys did BETTER than FULL -> not steady, NO EDGE as a whole. By type, steady in both years: chased shallow-asleep buys (A3 -0.51 / -0.60 vs all shallow-asleep -0.42 / -0.31) and chased re-crosses (A3 -0.55 / -0.30 vs all re-cross -0.17 / -0.20). Chased deep and 2nd-leg buys mixed (year 2 big winners). Lead only (subset picked after seeing the data): A3 for re-cross + shallow-asleep only, 368 buys over 2 years.
RN-26TESTING2026-11-15

Volume must support the V before a cross buy

Rule
For deep-V, re-cross and 2nd-leg cross buys (not opening, breakout or shallow-asleep). A: average volume per minute on the rise (V low to cross) must be at least the average on the drop (60-bar high to V low). B (user's favourite): at the cross bar, volume over the last 15 bars must be at least the day's pace so far (rv_day >= 1.0). If it fails, skip that cross.
Where it came from
User 11/15, RCAT 2025-02-11 buy #3 (11:42 deep-V cross, -2% first): 'has no volume to support the V'. Checked: drop into the V 27.4k sh/min (1.23x normal), rise out of it 13.0k (0.53x day pace), cross bars 1.3k-3.9k.
Still to decide
Thresholds 1.0x are first settings. Checked at the cross bar, not at the buy bar.
Result
30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on, halves split at Oct 26; random net -0.25 / -0.23; FULL V-cross buys net -0.27 / +0.05). A (rise vol >= drop vol): blocks 469 of 798 V-cross buys; blocked net -0.26 / +0.11 vs kept -0.28 / -0.06 = blocked did better in both halves (hurts) -> NO EDGE. B (15-bar vol >= day pace at cross): blocks 600 of 798 (75%); kept -0.16 / -0.09 vs blocked -0.30 / +0.09 = helps half 1, hurts half 2 (mixed). Kept buys are steadier between halves. By type: deep kept -0.22 / -0.16 vs blocked -0.44 / +0.21 (mixed); re-cross kept +0.10 / +0.08 vs blocked -0.04 / +0.02 (better in both halves, only 17 + 25 kept). FULL with B overall -0.32 / +0.21 vs FULL -0.30 / +0.16 = about the same. RCAT 2025-02-11 11:42 blocked by both. 2-year not run.
RN-25NO EDGE2026-11-15

Opening buy only if the next bar opens higher than the trigger bar

Rule
Add-on to RN-18 (opening 9:34-9:59, RUNNING + EMA9 above EMA20 and EMA50 + green trigger bar): buy at the next bar's open only if that open is higher than the trigger bar. If it is not, no buy from this trigger.
Where it came from
User 11/15, after the 2-year charts: 'add to the open rules: after the buy signal is triggered, the next open bar has to be higher than the trigger bar to allow the buy'.
Still to decide
'Higher than the trigger bar' = above its CLOSE or above its HIGH? And after a failed check: skip the opening buy that day, or keep looking for the next trigger before 10:00? Testing all four.
Result
30-day test 11/15 (Oct 5 - Nov 13, guards on, halves split at Oct 26; random 9:35-10:00 net -0.18 / -0.00). RN-18 as now: 463 buys, net -0.34 / +0.41. Open > trigger CLOSE: keeps only 163 of 463; kept buys net -0.47 / +0.06, blocked buys -0.27 / +0.60 = the blocked buys did BETTER in both halves (hurts). Open > trigger HIGH: keeps 56; kept -0.58 / +0.54 vs blocked -0.31 / +0.39 = mixed, tiny sample. 'Wait for next trigger' versions: close -0.39 / +0.42, high -0.50 / +0.35 = no better than RN-18 as now. Reading: a higher next open means paying more for the same move. 2-year not run (user to decide).
RN-24NO EDGE2026-11-15

Avoid cross buys while the EMAs are compressed

Rule
Applied 11/15 (user chose A): if EMA9, EMA20 and EMA50 stayed within 1 own candle of each other (highest minus lowest) for all 30 bars before a cross, skip that cross buy; only a breakout (RN-22) can buy out of the base. Opening (RN-18) and RN-22 not affected. Exception (user 11/15): a shallow + asleep cross that waits for RECLAIM (RN-17) still buys.
Where it came from
User 11/15, SOUN 2025-04-09 (base 11:30-13:15, then spring + breakout 13:19): 'as long as EMAs are compressed together, avoid buying with the cross; consider the distance at 11:00 as the acceptable gap'.
Still to decide
30 bars and 1 own candle are my first settings (user gave none).
Result
Charts 11/15 (RN-16 off), with the RN-17 exception: 3 buys removed - SOUN 12:36 +2% first, ACHR 14:01 -2% first, UUUU 10:56 neither. APPS 12:17 (+2%, user's mark) and SOUN 13:20 kept. With RN-16 on: only ACHR 14:01 (-2%) removed. Literal 'gap at SOUN 11:00' version (2.8c at the cross) would have blocked all 32 cross buys. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Removes 48 buys: half 1 diff +0.0 net -0.47 (good to block), half 2 diff +39.3 net +0.76 (blocked winners). Overall the blocked buys beat FULL (net +0.25 vs -0.08) = hurts. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). The 582 buys it blocks did better (net -0.04 / -0.13) than the rest of FULL (-0.17 / -0.26) = hurts.
RN-23NO EDGE2026-11-15

Green-bar check on every buy trigger

Rule
Applied 11/15 (user chose A): on every entry the bar that triggers the buy must close green (and be awake), buy the next open. If it is red, buy after the first green bar while the setup is still valid: EMA9 above EMA20; reclaim waits and RN-22 also need price above VWAP; opening must still be lined up before 10:00. RN-14 keeps its stricter 2-green-bar wait.
Where it came from
User 11/15: 'for every buy trigger, it will check if next bar will be green, before buying it'. Today only RN-14 (basing at the cross) checks candle colour.
Still to decide
None for now; judge it on the 30 days.
Result
Charts 11/15 (RN-16 off), 6 buys changed, all others already had a green trigger bar. Opening: AMCI 9:36 (-2%) and REPL 9:35 (-2%) gone, VIVO 9:35 -2% -> 9:37 +2%, JOBY 9:50 -> 9:51 still -2%. Deep V: APPS 11:45 -> 11:46 still +2%, CRML 13:31 -> 13:34 still neither. Opening now 4 / 8 / 2 (was 3 / 11 / 2). Same 4 opening changes with RN-16 on. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 147, removes 182. Added: diff -4.3 / +21.8, net -0.03 / -0.04; removed: -5.6 / +10.8, net -0.12 / -0.21. Better in both halves - the most steady rule. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). Buys it keeps net -0.06 / -0.40 vs the ones it drops -0.08 / -0.35. No help (30-day gain did not hold).
RN-22NO EDGE2026-11-14

After a cross + sideways run, buy the breakout with state RECLAIM

Rule
Applied 11/14 (user: breakout = B, EMA9 turns up sharply; sideways = more than 10 bars): an EMA9-over-EMA20 cross in the last 60 bars, then the 11 bars before are sideways (closes inside 4 own candles), then EMA9 rises 0.8+ own candle in one bar with state RECLAIM, EMA9 above EMA20, not asleep -> buy next open. One per sideways run. Separate entry, works without a deep V.
Where it came from
User 11/14, SOUN 2025-04-09: 'after a cross and side run, enter after ema9 breakout like what happened on this stock around 13:20 with state = reclaim'. SOUN 13:19: a big green bar (7.005 to 7.26) breaks the 12:39-13:18 range (high 7.239), state RECLAIM, EMA9 back over EMA20, buy 13:20 at 7.27, +2% first. (Already a buy under RN-20 + RN-21.)
Still to decide
My first settings: 4-candle sideways range, 0.8-candle EMA9 jump, cross within 60 bars. SOUN 13:02 is kept out only by the jump size (0.66c vs 0.8 needed); 13:19 jump was 1.03c.
Result
Charts 11/14: fires only twice on the 20 days, both on a bar another rule already buys: SOUN 13:20 +2% first (also RN-20) and FJET 12:59 -2% first (also RN-17). Adds no new buys here. Not tested on 30 days. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds only 12 buys: diff -8.3, net -0.59. Too few and negative. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). 148 buys, diff +9.7 / +0.0, net -0.25 / -0.40. Not steady.
RN-21NO EDGE2026-11-14

Second leg of a deep V counts as the same deep V

Rule
Applied (user chose A, 11/14): a deep-V cross that never bought, price stayed under VWAP since, and the new low holds within 1 own candle of the deep-V low -> the next cross counts as the same deep V (same waits RN-14/15/19); dip measured from the deep-V high. A buy ends it. It also lets RN-20 re-cross chain after it.
Where it came from
User 11/14, CDE 2025-03-04: 'why there is no buy signal at around 11:30'. The 10:35 deep-V cross (65 bars under, 19c dip) never bought: basing, flat cross candle, and no 2 green bars in a row before EMA9 went back under at 10:46. The 11:22 cross had 36 bars under but its dip was measured only from the 10:11 high 5.02 to the 11:08 low 4.89 = 2.6% = 6.5 own candles (needs 8) -> plain shallow. 11:31 cross = 6 bars under after 11:22, but 11:22 was not deep, so RN-20 re-cross did not apply.
Still to decide
1 own candle tolerance is my first setting.
Result
Charts 11/14 (RN-16 off): 7 new buys, 5 +2% first, 0 -2% first, 2 neither. RN-21 itself: CDE 11:23 +2%, SOUN 12:36 +2%. Re-crosses that now chain after it: CDE 11:32 +2%, SOUN 13:20 +2%, JOBY 13:12 +2%, JOBY 14:38 neither, UUUU 14:04 neither. EMA50 on: adds SOUN 13:20 +2%, JOBY 14:38 neither. Not tested on 30 days. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 45: diff +6.7 / -6.7, net -0.08 / -0.61. Not steady. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). 603 buys, diff +2.7 / -2.9, net -0.28 / -0.29 = about random.
RN-20TESTING2026-11-14

Re-cross after a deep V counts as the same deep V

Rule
If EMA9 dips back under EMA20 for 10 bars or fewer right after a deep-V cross, the next cross counts as part of the same deep V (same waits RN-14 / RN-15 / RN-19).
Where it came from
User 11/14, ASST 2025-05-21: 'around 11:03-05 we should have a buy signal at that cross'. The 10:46 deep-V cross was asleep, its reclaim wait was cancelled when EMA9 dipped under at 11:00; the 11:05 cross was only 5 bars under = shallow, so nothing fired.
Still to decide
10 bars is my first setting.
Result
Charts 11/14 (RN-16 off): 5 re-cross buys - ASST 11:06 +2% (the user's mark), REPL 13:05 +2%, ACHR 14:01 -2%, SOUN 11:21 -2%, APLD 2/7 12:46 neither. Not tested on the 30 days yet. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 211 re-cross buys: diff +0.0 / +12.7, net +0.00 / +0.03 vs random -0.25 / -0.23. Better than random on net in both halves; still about break-even after cost. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). 3,512 re-cross buys: diff +1.8 / +2.5 vs random -0.9 / -1.0; net -0.19 / -0.19 vs -0.25. Small steady tilt both years, below cost.
RN-19NO EDGE2026-11-14

Never buy while the stock is ASLEEP

Rule
The bar that triggers a buy must not have the ASLEEP flag. Waits go on while EMA9 stays above EMA20.
Where it came from
User 11/14, APLD 2025-03-11 buy #3 (12:37, shallow + asleep): 'state was asleep, so we should add the rule - dont buy as state is asleep'.
Still to decide
Clashes with RN-17 (which waits for a reclaim after a long sleep): with RN-19 on, the RECLAIM bar must also be awake.
Result
Charts 11/14: removed APLD 3/11 12:37 (-2%), UUUU 11:57 (neither) and CRML 11:54 (+2%, asleep the whole reclaim). APPS ~12:20 mark kept: the 12:16 cross bar was already RECLAIM and awake, buy 12:17 (+2%). | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 30, removes 84 (asleep buys). Removed: diff -5.5 / +13.8, net -0.56 / -0.38; added net -0.41 / +0.16. Better than without it in both halves on net. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). The asleep buys it blocks did BETTER (net -0.25 / -0.18) than what replaces them (-0.35 / -0.35). Hurts (30-day gain did not hold).
RN-18TESTING2026-11-14

Opening style: after the first 5 bars, RUNNING and EMA9 > EMA20 > EMA50 = buy

Rule
The first bar from 9:34 to 9:59 with state RUNNING and EMA9 above both EMA20 and EMA50: buy the next bar open. One per day.
Where it came from
User 11/14, UPXI 2025-07-21: no deep V at the open, but right after the first 5 bars state is running and EMA9 > EMA20 > EMA50. A separate entry style.
Still to decide
11/14 change (user, APLD 2025-03-11: 'first buy should happen at 9:38, next open after the cross'): was EMA9 > EMA20 > EMA50; EMA20 was still a hair under EMA50 at 9:37, so now only EMA9 has to be above both. UPXI still buys 9:38. Cut-off 10:00 is my choice.
Result
Charts 11/14 on the 20 deep-V days: 16 opening buys - 3 hit +2% first (UPXI 9:38, QUBT 9:35, FJET 9:45), 11 hit -2% first, 2 neither. Not tested on the 30 days yet. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 463 opening buys: diff -13.0 / +23.2 vs random 9:35-10:00 -9.2 / +3.7; net -0.34 / +0.41. Big swing with the market; worse than random in half 1. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). 8,676 opening buys: diff +3.8 / -1.7 vs random at the same time -2.3 / -6.8; net -0.09 / -0.33 vs -0.53 / -0.71. Beats same-time random by about +6 diff and +0.4% net in BOTH years = the best tilt in the set, but still below cost.
RN-17NO EDGE2026-11-14

Shallow cross after 11+ asleep bars: wait for RECLAIM, then buy

Rule
At an EMA9/EMA20 cross that is not a deep V, if the stock was asleep 11+ bars in a row during the dip, wait for state RECLAIM (no time limit while EMA9 stays above EMA20), buy the next bar open.
Where it came from
User 11/14.
Still to decide
'Asleep for >10 bars' read as 11+ asleep bars in a row inside the dip (from EMA9 going under EMA20 to the cross). 11/14 fix: the cross bar itself counts for the reclaim wait.
Result
Charts 11/14 (with RN-19): 2 buys - APPS 12:17 +2% (user's ~12:20 mark), FJET 12:59 -2%. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 61 buys: diff +6.5 / +10.0 (above random both halves) but low rise (+0.85%) and net -0.30 / -0.23 = same as random. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). 1,017 buys, diff -4.9 / +2.0, net -0.41 / -0.34 = worse than random.
RN-16NO EDGE2026-11-14

At the cross EMA50 must not be falling

Rule
At the cross, block the buy when EMA50 is lower than 5 bars ago.
Where it came from
User 11/14.
Still to decide
'Not falling' read as EMA50 now >= EMA50 5 bars ago.
Result
Charts 11/14 on the 20 deep-V stock-days: deepv_new_rules_entries.pdf (all rules) and deepv_new_rules_entries_no_ema50.pdf (RN-16 off). Not tested on the 30 days yet. Blocks almost every cross: 59 of 72 deep-V / shallow-asleep crosses blocked, winners and losers alike (a deep V almost always has EMA50 still falling at the cross). With it on, almost only the opening entries are left. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Kept 589: diff -10.4 / +19.9, net -0.30 / +0.29. Blocked 750: -2.5 / +12.0, net -0.30 / +0.05. Worse in half 1, better in half 2 = not steady. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). Kept net -0.12 / -0.30 vs blocked -0.21 / -0.22. Better year 1, worse year 2.
RN-15NO EDGE2026-11-14

Asleep at the cross: wait for RECLAIM

Rule
If the stock is ASLEEP at the deep-V cross, wait until the state changes to RECLAIM, buy the next bar open. Cancel if EMA9 drops back under EMA20.
Where it came from
User 11/14.
Still to decide
If the cross is both BASING and ASLEEP, this rule wins over RN-14.
Result
Charts 11/14 on the 20 deep-V stock-days: deepv_new_rules_entries.pdf (all rules) and deepv_new_rules_entries_no_ema50.pdf (RN-16 off). Not tested on the 30 days yet. Removed the APPS 14:11 fire (EMA9 went back under before any reclaim) - one of the user's 'should never happen' marks. Also removed SOUN 4/9/25 12:07 (a winner in the old charts). | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Adds 19, removes 67. Added net -0.15 / -0.35, removed +0.44 / -0.23. Mixed, no clear help. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). Adds net -0.24 / -0.21, removes -0.28 / -0.17. Mixed.
RN-14NO EDGE2026-11-14

BASING at the buy signal: green candle confirm, else wait for 2 green bars

Rule
If the state is BASING at the deep-V cross: if the cross candle is green, buy the next bar open; if not, wait for 2 green bars in a row, then buy the next bar open. Cancel if EMA9 drops back under EMA20.
Where it came from
User 11/14.
Still to decide
'2 green bars' read as 2 in a row.
Result
Charts 11/14 on the 20 deep-V stock-days: deepv_new_rules_entries.pdf (all rules) and deepv_new_rules_entries_no_ema50.pdf (RN-16 off). Not tested on the 30 days yet. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). Moves basing buys to a green bar. Buys it adds: diff -9.5 / +10.9, net -0.73 / +0.24; buys it removes: -9.2 / +1.7, net -1.03 / +0.01. Better than without it in both halves (small). | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). Its buys net -0.36 / -0.14 vs the ones it replaces -0.32 / -0.15. No help (30-day gain did not hold).
RN-13LIVE2026-11-14

Deep-V restart: main entry = EMA9 crosses EMA20 after a deep V

Rule
Buy at the next bar open after EMA9 crosses above EMA20 after a deep V. Deep V for now = v1 values: EMA9 under EMA20 for 20+ bars and a dip (60-bar high to the low) of 8+ own candles. Buy window 9:35-15:00. Guards on. No other filters.
Where it came from
User 11/14: 'lets start the deep-V rules over, the main entry is: enter when the ema9 cross ema20 after a deep v'. Replaces RN-1 as the base; further deep-V rules get added on top one at a time.
Still to decide
Deep-V definition kept at v1 (20 bars, 8 candles) until you change it.
Result
Charts rebuilt 11/14 (deepv_restart_all_signals.pdf, 20 stock-days, every signal marked). Moving the window to 9:35 added no fires on these 20 days: a 20-bar deep V cannot finish much before 10:00. 2-year result for the same rule is RN-1: about random. | 30-day test 11/15 (Oct 5 - Nov 13, 881 stock-days, guards on; halves split at Oct 26; random: diff -1.4 / +3.7, net -0.25 / -0.23). RN-13 alone: 715 buys, diff -4.6 / +7.7 (random 10:00-15:00 +0.6 / +3.7), net -0.44 / +0.04. Worse than random in half 1, better in half 2. | 2-year test 11/15 (Oct 2024 - Oct 2026, 503 days, 14,754 stock-days; year 1 / year 2; random 10:00-15:00 diff -0.9 / -1.0, net -0.25 / -0.25; random 9:35-10:00 -2.3 / -6.8, net -0.53 / -0.71). RN-13 alone: 11,541 buys, diff +2.2 / +0.5, net -0.17 / -0.22. Small tilt over random, below cost. | 11/16 user: LIVE on the trade bot (paper) from 11/16 as EV3 setup DV1 = plain RN-13, FIRST deep V of the day only, EV3 exit (stop under V low, GB2 + C3), no-trade gate exempt, guards in the setup. 2y first-deep-V entry only: diff +2.3 / +0.2, net -0.15 / -0.24 (below cost); EV3 exit on it untested. Code dv1_setup.py; checked bar by bar = 84/84 test buys on 6 days. Off: config ev3.setups.DV1.enabled false.
RN-12TESTING2026-11-13

After a shallow cross, buy when the state changes to RUNNING

Rule
When EMA9 crosses above EMA20 but the cross is shallow (not a deep V: under 20 bars below, or a dip under 8 own candles), do not buy on the cross itself unless the state is RUNNING. Buy at the next bar open on the first bar where the state is RUNNING (close above VWAP) - the cross bar itself counts. No time limit after the cross, as long as EMA9 stays above EMA20; if EMA9 drops back under EMA20, wait for a new cross. Buy window 9:35-15:00 (at least 5 bars to see how the stock behaves).
Where it came from
QUBT 2025-01-24: crosses at 9:59 and 10:12 were too shallow for the deep-V rule (12 and 1 bars under, 4.4 and 4.3 own candles) and nothing bought; price then ran 11.60 to 12.18 by 10:26.
Still to decide
Read 'drop below EMA20' as EMA9 going back under EMA20 (the thing that needs a new cross).
Result
30 days (10/5-11/13), first shallow-cross buy per stock-day, guards on: 686 buys, +2% first 32.5% / -2% first 28.4% (diff +4.1), hold60 net +0.00 vs random -0.30. 9:35-10:00: 512 buys diff +1.2 vs random same slot -5.4. 10:00-10:30: 51 buys diff +29.4, net +0.51 (small sample). 10:30-15:00: 123 buys +5.7, net -0.20. Waited 6+ bars after the cross: +13.3; RUNNING on the cross bar: -1.6. QUBT 1/24/25: would buy 9:35 (+2%), 10:06 (-2%), 10:13 (+2%), 11:29 (-2%). 2-year run waits for the user.
RN-11IDEA2026-11-13

Looser deep V for the morning

Rule
Before about 10:30, accept a cross after 10+ bars under and a 4+ own-candle dip (instead of 20 bars and 8 candles).
Where it came from
QUBT 2025-01-24 10:12 and APPS 9:53: fast morning Vs the deep-V rule is too strict to catch.
Still to decide
Exact morning cut-off and numbers.
Result
Not tested.
RN-10IDEA2026-11-13

Price vs EMA50 at the buy

Rule
Block cross-during-reclaim buys when the close is more than X% above EMA50.
Where it came from
Losers PDF with EMA50: all 40 winners and losers bought with EMA50 falling and EMA9 under EMA50, so those cannot block. Cross-during-reclaim losers were a little further above EMA50 (+0.87% vs +0.52%).
Still to decide
X to sweep.
Result
20 charts only - hint, not tested.
RN-9IDEA2026-11-13

Skip cross-during-reclaim buys before 10:30

Rule
Cross-during-reclaim entry, buy window 10:30-15:00 instead of 10:00-15:00.
Where it came from
Cross-during-reclaim charts: 7 of 10 losers bought 10:00-10:22, only 3 of 10 winners before 10:30.
Result
20 charts only - not tested.
RN-8IDEA2026-11-13

Block when resistance is far above (2%+)

Rule
Reclaim rule, block the buy when the nearest resistance above is 2% or more away.
Where it came from
2-year resistance test: buys with the nearest resistance 2%+ above were the worst group in both years (hold60 net -0.54 / -0.33).
Result
Seen inside RN-7 test, not tested on its own.
RN-7NO EDGE2026-11-13

Block the buy if the gap to resistance is under 0.8%

Rule
Reclaim rule, block when the nearest resistance above (swing levels, open, first-15-min high/low) is under 0.8% away.
Where it came from
User rule on top of the reclaim rule.
Result
30 days looked good (kept + cross last +13.5 diff). 2 years: kept buys diff +0.8 / +1.5, hold60 net -0.32 / -0.25 = worse than no block (-0.22 / -0.21); the blocked ones held better. Near resistance = smaller moves both ways.
RN-6NO EDGE (best tilt so far)2026-11-13

Cross during reclaim

Rule
RN-5, and the EMA cross bar is at or after the first bar of the run above VWAP.
Where it came from
Inside the reclaim rule (RN-5): the EMA9/EMA20 cross happens during the reclaim run, not before it.
Still to decide
Exit not built (stop under base low, +1/+2% target, time stop).
Result
2 years: diff +6.0 / +3.2 (+2% first minus -2% first), hold60 net -0.20 / -0.18. Best entry tilt we have, still under the 0.2% cost on a 60-min hold. Charts: cross_during_reclaim_winners_losers.pdf.
RN-5NO EDGE2026-11-13

Reclaim after 6+ basing/asleep bars

Rule
State RECLAIM (close above VWAP, first 5 bars of the run, 15 of the 20 bars before under VWAP), at least 6 BASING-or-asleep bars right before, EMA9 > EMA20. Buy next bar open 10:00-15:00. Guards on.
Where it came from
User: 'the state should be reclaim, after it been in an asleep or basing for more the 5 bars', v1 values (SOUN pages).
Result
2 years: diff +1.9 / +1.5, hold60 net -0.22 / -0.21 (random -0.30 / -0.26).
RN-4NO EDGE2026-11-13

EMA cross with BASING as the gate

Rule
EMA9 crosses above EMA20 while the state is BASING; settings swept (EMA gap, range, asleep).
Where it came from
User: use the EMA cross to fire and the basing state to allow the buy.
Result
2 years: quiet bases best (EMAs together <= 0.3 candles: diff +3.5 / +2.7, net -0.10 / -0.17), all below cost. 30 days alone showed nothing.
RN-3NO EDGE2026-11-13

Steep V rises, flat V fades

Rule
Judge deep-V fires by the V angle (own candles per minute).
Where it came from
User: 'the vs with more vertical angle usually rise, the vs with less angle if rises it will take longer otherwise it fades'.
Result
2 years: half true - steep Vs move faster both ways (+2 first and -2 first both ~27-30%), flat Vs slower; no edge.
RN-2IDEA2026-11-12

APPS marks: first cross 9:53, second ~12:20, never buy with EMAs + VWAP sloping down

Rule
Block a cross buy when EMA9, EMA20 and VWAP are all sloping down; allow early (before 10:00) Vs.
Where it came from
User marks on APPS 2025-06-25: 9:53 and the 12:16 cross are the right buys, 11:45 is not, the 14:11 fire should never happen.
Still to decide
3 tests offered, not run.
Result
Not tested.
RN-1NO EDGE (base replaced by RN-13)2026-11-11

Deep-V EMA cross (core)

Rule
Buy next bar open after EMA9 crosses above EMA20, EMA9 under EMA20 for 20+ bars, V dip 8+ own candles, 10:00-15:00, guards on.
Where it came from
Per-stock logic study.
Result
2 years: about random; only cross-above-VWAP tilt, below cost.