Stack Builder · v1
Build your own stack. Then try to break it.
Pick engines, set contracts, and get a combined backtest in real time — with the out-of-sample and Monte Carlo checks that separate an edge from a curve fit. Save it to the board, add it to your cart, or hand it to Nexus Cloud to run.
Presets load an official stack's engines at your account's micros. The builder composes each engine's full published curve (every tune, per micro); the official stacks run hand-picked tunes per engine, so preset figures are an approximation of the stack pages, not a copy.
ORB BUILDER
$100/mo
Asia 20:00 ET + NY 09:30 ET · L/S
+$6.0k · DD $749 / micro
ORB 5SEC
$40/mo
Asia 20:00 ET + NY 09:30 ET · L/S
+$2.2k · DD $224 / micro
CONTINUATION LEGACY BUNDLE
$70/mo
NY 08:15 range + 09:30 open · LONG / L-S
+$1.6k · DD $116 / micro
PIVOT FADE NQ
$35/mo
NY 10:00–15:00 ET · LONG
+$6.8k · DD $572 / micro
RSI-2 PULLBACK NQ
$35/mo
NY 09:30–15:00 ET · L/S
+$12.7k · DD $2,124 / micro
IGNITION NQ
$35/mo
NY 09:30–15:00 ET · L/S
+$23.2k · DD $2,015 / micro
MOMENTUM BUNDLE
$50/mo
NY day session · L/S
+$2.8k · DD $431 / micro
KELTNER 5S NQ
$40/mo
NY 09:30–16:00 ET · L/S
+$3.8k · DD $513 / micro
EMA PULLBACK NQ
$40/mo
NY 09:30–16:00 ET · L/S
+$7.4k · DD $1,298 / micro
Combined backtest · 50K account · 21 micros max
2024-12-29 → 2026-07-03 · 551 days · end-of-day drawdown
Robustness
70/100
Net profit
$49,669
whole span
Max DD
$2,539
vs $2,000 trailing
MAR
12.97
Per month
$2,744
18 / 20 green months
Out-of-sample test
The first 70% of history is the in-sample you tuned on. The last 30% (after 2026-01-21) was never used to pick this mix — if the edge is real, it should hold there too.
In-sample
$25,172
MAR 12.99
Out-of-sample
$24,497
MAR 21.62
Out-of-sample MAR holds at 166% of in-sample — the edge survives unseen data.
Monte Carlo stress · 500 shuffles
Same daily results, random order. Real trading won't replay history in sequence, so the question is: what drawdown could the same days have produced?
White line = $2,000 trailing drawdown for a 50K account.
The 95th-percentile shuffle blows the $2,000 trailing DD. Fewer micros or a bigger account.
Every engine pair is verified inside an official stack
100% verified
All nine engines trade MNQ long and short in overlapping New York sessions. Pairs that ship together in an official stack have had their entries reviewed so they never hedge each other into a wash. Unverified pairs are not wrong — they just haven't had that review, so treat their combined curve as optimistic until they have.
Save to the leaderboard
The server recomputes every number from the source data before ranking, so the board can't be gamed. Ranked by out-of-sample MAR, damped by robustness.
Run this stack
3 engines · $210/mo list · 20% build-a-bundle discount → about $168/mo. Set the micros above in the Connector, or hand it to Nexus Cloud to run for you.
Builder leaderboard
Backtest board, ranked by out-of-sample MAR × robustness. Live board follows once Nexus Cloud paper-runs the top entries.
Why these backtests matter
Most backtests are optimistic. Here is what we do about it.
A strategy that only works on the data it was tuned on is worthless. Every result in the builder passes through the same six checks.
01
Out-of-sample split
History is cut 70 / 30 by date. The mix is judged on the last 30% it never saw. A stack whose out-of-sample MAR collapses was fitted to the past, not to the market.
02
Monte Carlo reshuffle
Five hundred random re-orderings of the same daily results. The 95th-percentile drawdown is what you should size to — the historical sequence is just one draw.
03
Account headroom
Every drawdown figure is compared to the trailing DD of the account size you picked. Below ~35% headroom the stack fails: live intraday drawdown runs deeper than end-of-day.
04
Pair verification
Engine pairs that ship together in an official stack have been reviewed so they never hedge each other into a wash. Unverified pairs are shown, not hidden.
05
No self-reported numbers
The leaderboard API recomputes every statistic from the source equity data. What you submit is the recipe — the server does the cooking.
06
Same data as the storefront
The builder runs on the exact per-micro daily curves published for each engine. If a number here disagrees with a product page, that is a bug, not marketing.
v1 composes audited engines. Parameter-level building (your own entries, stops, sessions) with walk-forward optimisation is on the roadmap.
Ready to deploy?
Start with the highest-MAR stack, or build your own bundle and save up to 33%.