QT + SSMT, as I read it
Tap Right, Partly or Wrong under each picture. Type only where I'm off. Saves itself.
Cycles inside cycles
Every quarter is the next cycle down. Gold tick = True Open (open of Q2).
One day, zoomed in
The 90-minute cycle is 6 hours long. Its quarters are 90 min.
Premium and discount in time
One open, or two stacked?
Our backtests used one. Daye says two. Which pair do you use?
Which open is live
Q4's open carries into the next Q1. Not in any lecture I have: where's it from?
SSMT: one or two break, the rest hold
Own prior-quarter level, wick through counts. Trade against the break.
The gate
HSSMT: same idea, closes only
When is it dead?
Our rule below. Daye instead: an opposing SSMT of equal or bigger range cancels it.
Higher cycle sets direction, lower cycle times the entry
How I'd take it on SSMT alone
NY AM vs London, beyond TDO
beyond T90O / T90RO
enter MNQ on that 5m close
2 pts past the sweep
London's other side
1Tue 21 Oct 2025 · 15m · daily SSMT
MNQ and MES break their London low. YM holds. Break is below TDO → bullish.
25m · 90m SSMT → entry, stop, target
10:32 MNQ breaks the 09:00–10:30 low, MES and YM hold, below T90RO. 10:35 it closes back above → long.
Honest check
One good day is easy to find. Mechanically, these 5 rules lose. Your corrections are what should fix it.
Same quarter, three cycles in a row
The 10 sequences on your clock
Daily · 90m · Micro quarter numbers line up. 333 = Daily Q3 · 90m Q3 · Micro Q3 = 09:45–10:07:30.
No sequence, no SSMT
The SSMT counts only inside a sequence or with a sequence's high/low. Trade the one that held.
Three True Opens: A+ vs B
Same rule one level up?
My guess at how QS works for bigger SSMTs (you said Stage 1 is Daily or higher). Lethality only showed the intraday one.
"Cheat code": 90m Q4 → next 90m Q1 (e.g. 10:30–12:00 → 12:00–13:30) counts as an SSMT pair.
Does QS pick better SSMTs?
Not as I've coded it: QS ≈ no QS
Every SSMT 2019–2024 (2025–26 held back). Enter the one that held at the quarter close, stop past the level it held, 1R target, costs in.
Win rate at 1:1 ≈ 50% in every bar. Hover a bar for the trade count.
A+ (all 3 True Opens) is worse, not better
Sequence trades only. Price beyond all three opens tends to keep going, not reverse.
His trades, decoded
His indicator rows use the exact digits my engine uses. The turn happens in the lowest cycle of a valid sequence.
Test 2, his way: still a coin flip
Every micro SSMT 2019–2024, entered when the sweeper closes back inside. Chance MNQ goes +30 pts before −30.
stops · 1R/2R/3R · old sequence · BE · EOD
Bought, checked, re-tested
$29.22 of your $33 credit, saved on the SSD
May → 16 Oct 2026 · $2.36
his 4th instrument · $10.33
Daye's FX triad · $6.22
his trade days · $3.23
bond sentiment · $7.09
PortableSSD › Trading-Data › databento (raw + parquet + README). About $3.78 credit left.
My engine sees his trades
His rules lose in every period. One variant wins in all three.
Two periods the rules never saw: 2017–19 and 2025–26. MNQ, costs in, flat 15:55.
The survivor
swept level set in a window, sweep lands in a window
fade MNQ's false move
A+ loses, every period
within 30 min
flat 15:55
≈ $2,980 on 1 MNQ
Treat as a lead, not an edge yet: found by searching 576 combos, small (PF 1.09–1.40), and it breaks two of his rules ("trade the one that held", "A+"). A 40-pt stop is also bigger in % terms back in 2017.
Round 1: stricter test, better trade
Nothing you couldn't do live
What I tried, what stuck
Better per trade, smaller drawdown
was 1.43 · 2019–24: 1.29 (was 1.20)
was +0.27R
was −$2.0k · fits a 50K account
Trade-off: fewer trades (2.2/wk), so total $ on 1 contract is lower (+$13.4k vs +$17.1k since 2017). Per dollar of drawdown it's better (10.4× vs 8.6×). 2017–19 is still flat for both.
The rules now
level set in a QS window, swept in a QS window
ES, YM, RTY all held
Daily · 90m · Micro
enter next bar's open (within 30 min)
breakeven after a bar reaches +60
one trade at a time
Replicable: yes. A money machine on Topstep: no, not yet.
A live bot reproduces the backtest exactly
I wrote a second engine that walks forward one bar at a time and never sees a future bar, like a live bot. Same result, to the cent.
0 mismatches
same entry, exit, reason
1 MNQ, Jul 2017 → Oct 2026
Round 2: no new setup survived
Lone-NQ sweep on the 90m cycle: no setting positive in all 2019–24 blocks. Daily cycle: 1 of 36 settings, 67 trades, neighbours negative = luck. Micro setup stays the only one.
The real 2017–2026 path, one 50K account
Combine → Express Funded → new combine whenever it closes. $49/mo + $149 per pass. Payouts at 90%, after 5 winning days, 50% of balance up to $2k.
| MNQ | combines bought | funded accounts | payouts | paid to you | fees | net / year |
|---|---|---|---|---|---|---|
| 1 | 3 | 3 | 4 | $2.1k | $5.2k | −$334 |
| 3 | 15 | 8 | 13 | $18.1k | $4.8k | +$1,441 |
| 5 | 22 | 9 | 24 | $40.0k | $3.9k | +$3,928 |
Best management found: 5 MNQ throughout, take each payout as soon as allowed. From 14 different start dates: median +$4.6k/yr, worst +$1.2k/yr. Funded accounts usually close within ~2 months at 5 MNQ; the money comes from passing again.
But any single year is close to a coin flip
2,000 simulated years built from reshuffled real weeks. Net to you after fees, one account.
At 5 MNQ: average +$1.4k/yr, median +$0.4k, 44% of years lose money (fees), 36% get no payout at all.
My recommendation
run the live engine on real-time data for 4–8 weeks; signals must match what the chart shows
expect ~3 months to pass; worst realistic year ≈ −$900 in fees
size can't fix this; more independent trades can
Before getting happy
One rule change that held up: reclaim within 10 minutes
MNQ must close back inside its level within 10 min of the sweep (was 30). Picked on 2019–24 only, better in both unseen periods.
| Profit factor | 2017–19 unseen | 2019–24 tuned | 2025–26 unseen | 2022–26 |
|---|---|---|---|---|
| before (C) | 1.01 | 1.28 | 1.73 | 1.51 |
| now (C+) | 1.07 | 1.30 | 1.83 | 1.57 |
Also tested, didn't hold in every block: bigger targets, trailing stops, BE at 1R/2R, NY-only, overnight-only, skip Asia evening, beyond-0-True-Opens, no Mondays, 35/45 stops. "Skip 13:00" looked good but I spotted it after seeing unseen data, so it's not a rule.
The numbers (C+, 1 MNQ)
2022–24: 1.46 · 2025–26: 1.83
avg win +$231 · avg loss −$60
≈ $2.5k/yr · 5 MNQ: +$60.7k
5 MNQ ≈ −$5.6k
best +$1.5k · worst −$745
NY AM trades ~18 min
Month by month (1 MNQ)
| year | trades | win | PF | net 1 MNQ |
|---|---|---|---|---|
| 2022 | 113 | 27% | 1.72 | +$3,308 |
| 2023 | 123 | 34% | 1.61 | +$3,353 |
| 2024 | 108 | 23% | 1.05 | +$236 |
| 2025 | 84 | 31% | 1.73 | +$2,595 |
| 2026 (to Oct) | 71 | 30% | 1.96 | +$2,644 |
When the trades happen
| session (NY time) | share | PF | Dubai time |
|---|---|---|---|
| Asia evening 18:00–24:00 | 32% | 1.34 | 02:00–08:00 |
| NY lunch/PM 12:00–16:00 | 29% | 1.42 | 20:00–00:00 |
| Asia late 00:00–03:00 | 18% | 1.73 | 08:00–11:00 |
| NY pre-open 06:00–09:30 | 12% | 2.35 | 14:00–17:30 |
| NY AM 09:30–12:00 | 9% | 1.61 | 17:30–20:00 |
Busiest single hour: 12:00 NY (the 411 window), 23% of trades. About a third land between 02:00 and 08:00 Dubai, so this needs a bot, not you awake.
Trades per week: 2.0 on average, very uneven
29 weeks with zero trades, 71 with one, 72 with two, 33 with four or more.
How trades end
avg +$319
avg +$89
after +60 · avg −$2.5
avg −$83
Longs PF 1.61, shorts PF 1.52. Every weekday positive (Monday weakest, PF 1.15). Median trade goes 49 pts your way and 40 against.
Topstep 50K, C+, 2022–2026
| 3 MNQ | 5 MNQ | |
|---|---|---|
| Real path from Jan 2022 | +$15.1k net (5 combines passed) | +$35.6k net: one funded account never closed, 20 payouts |
| Every monthly start since 2022: median | −$24 | −$198 |
| …average | +$1,690 | +$4,718 |
| …attempts that lost money | 52% | 62% |
| Median days to pass | 120 | 56 |
| Reshuffled 2022–26 years: median / average | +$681 / +$1,633 | +$2,613 / +$2,879 |
| …years that lose money | 36% | 25% |
Skewed: most attempts lose a few hundred in fees; a few runs pay big. The Jan 2022 start was one of the lucky ones.