NQ Master · M2 exit & management study · rebuilt 27 Aug 2026

What survives an honest clock

The first version of this study timed entries from a Renko bar's open stamp and aligned prices on the wrong column. Both are fixed at source. The headline barely moved — the two errors were cancelling — but the evidence underneath rearranged, and one earlier “cleanest result” did not survive.

The short version

What to do

Add a breakeven stop at +30 points. It is the one management rule whose advantage survives a session-level significance test: points over taking every signal and waiting for the reversal, with an interval that excludes zero.

Adding a 150-point target and a 250-point disaster stop on top reaches points and cuts maximum drawdown from to . That combination is what I would run — but be clear that its edge over the baseline is not statistically separable, and the target on its own is now worth nothing measurable.

The exit rules are the small half of this study. The rule that ignores new signals while a position is open is worth far more than any of them. Section four.
Against the previous version

What the corrected timing changed

Two faults were found by asking a simple question: is the timestamp in the signal export the bar's open or its close? It was the open. On Flex Renko a brick takes a median of 12 seconds and sometimes half an hour to complete, so the stamp is not when the signal existed.

Holds

Breakeven at +30 is still the one proven addition

Worth points against the baseline with the interval clear of zero. This has now survived a sample increase, a doubling of costs, and a rebuild of the entry timing.

It still wins on magnitude rather than frequency — better on sessions, worse on .

Withdrawn

The 150-point target is not worth anything

The previous version called this the cleanest evidence in the study: significant, positive in all ten quarters, better on 229 sessions against 171. On honest entry timing it is worth points with an interval straddling zero in both directions. It was an artifact of entries that only filled after price had already moved their way.

Holds

Tight stops destroy the edge — now more clearly

25 points costs , 50 points , and the 100-point stop is now significantly negative too. Three stop widths, all significant, all in the same direction. This is the most robust result in the study and it answers the 100-tick disaster-stop question definitively.

Marginal

The full composite is defensible, not proven

Breakeven plus target plus disaster stop returns more than the baseline, but the interval grazes zero. Its real argument is drawdown, not return: the worst peak-to-trough falls by roughly a third.

Holds

The edge is not market drift

NQ rose points over the sample, so the long book has a tailwind. The short book still made into that headwind, and both sides are positive in all three years.

Structural

The raw signal loses money

Held to the opposite signal, the average permit returns points. The tradeable book returns . Everything between those two numbers is the position rule.

Section one

Why the clock mattered

Sierra stamps every bar with its open time. On a time-based chart that is harmless, because a five-minute bar closes five minutes later and everyone knows it. On Flex Renko the bar ends when price has travelled a fixed distance, so the gap between the stamp and the signal is whatever the market decided to do — a couple of seconds in fast trade, half an hour in a quiet Asian session.

The old pipeline looked for an entry from the stamp onward. A quarter of those entries landed before the bar could have finished. Worse, the entry level came from back-adjusted chart prices whose offset had been fitted against the wrong column, leaving it about six points too favourable — so a trade only filled once price had already moved six points the right way. That is a continuation filter: it quietly selected winners and discarded signals that failed to follow through.

Both are now fixed in the study itself. M2 exports a BarCloseTime column, entries are taken at that instant at the price actually trading then, and the chart's price scale is no longer anywhere in the profit-and-loss path.

How long a Renko brick takes to complete — the size of the gap between the old timestamp and the real signal. Median seconds, but the tail is what did the damage: of bricks took longer than two minutes to print.
Section two

Every model on the corrected basis

Portfolio level throughout: one position at a time, new signals ignored while a position is open, flat at the CME daily close (17:00 New York, daylight-saving aware), trades allowed to run a full session. Four ticks round trip charged on every exit.

Sorted by total. Max drawdown is measured trade by trade, not on end-of-session equity. The 90% interval is a bootstrap over sessions, not trades.
Cumulative points per contract by session.
Difference from the baseline with the 90% session-bootstrap interval. Solid bars clear zero; faded bars are not evidence of anything. Note how much of the middle of this chart is faded — that is the honest picture of how much exit tuning is worth here.
Section three

The stop question, settled

A catastrophic stop at 100 ticks — 25 points — was the specific request. It costs about seven thousand points. So does every stop tight enough to fire often.

Expected final result once the trade is x offside What a stop at x pays instead
Signal level, every recorded trade. The curve sits below the stop line at every width, which says a stop helps the average signal — and the portfolio test says the opposite, because stopping out early frees the position slot and puts you straight back into the same chop. The tradeable book is the one that pays you.
Where a stop does earn its place is the tail. The 250-point disaster stop fires on of trades — roughly once a fortnight — and costs nothing measurable in return while cutting the worst drawdown.
Section four

Where the edge actually lives

This is the largest effect in the study and it is not an exit rule.

Held to the opposite signal, the average M2 permit returns points. Taken one at a time under the live position rule, the same stream returns . The entire difference is which signals get skipped.

M2 emits long runs of same-direction permits between flips — median , mean , and the longest run in the sample is permits without a single opposite signal. Every permit in a run ends at the same eventual reversal, so the deeper a permit sits, the more of the final adverse move it absorbs.

Average result by position within a same-direction run. The first eight are profitable. Everything past that is trades at points each, and holding one position at a time skips almost all of them.
Practical consequence. The rule that a second signal in the same direction is ignored is not an execution detail — on this evidence it is carrying the strategy. Any future change that permits concurrent or re-entry positions has to be re-tested against this rather than assumed neutral.
Section five

Consistency and drift

Quarterly totals, points per contract, shaded proportionally. Ten quarters spanning a large rise in NQ and two contract-cycle regimes.
Long and short books by year against the underlying move. Both sides positive every year, in both models, in a market that rose throughout — the control that separates an edge from exposure to the tape.
Section six

Recommendation

Run this

  • Breakeven at +30 points — the only individually proven addition.
  • Target at 150 and a 250-point disaster stop — for drawdown control, on the understanding that neither is independently significant.
  • Opposite signal closes the position, flat at the CME daily close, one position at a time.
  • Optional: stop taking new signals once a session is 400 points down. Slightly lower total, similar drawdown.

What not to do

No stop tighter than roughly 150 points — 25, 50 and 100 are all significantly negative. Do not treat the 150-point target as a source of return; it is drawdown management. Do not let the system hold over the weekend, and do not add re-entry positions without re-testing section four.

Section seven

What this still cannot tell you

The win rate is . The average win is points and the average loss . That distribution is hard to sit through, and no backtest measures whether a person will.

One instrument, one brick size. MNQ on Flex Renko 30, nothing else tested.

Entry logic is untouched. This study takes M2's signal stream as given. Section four says the larger opportunity is upstream: the deep re-entries the position rule currently discards are a signal-quality problem, not an exit problem.

Costs are modelled, not observed. Four ticks round trip is an assumption; fills are taken from 1-second bars, which cannot see queue position or a fast market. The recommendation survives six ticks.


Built by Gnourt · algorithmic trading systems