MGEQuant

Honest quant research on NQ futures: what we tested, what failed.

Weekly notes:

[MGEQuant @MGEQuant]

Bar-type sequences: nothing passed the bar

We wrote the rules and the bar first, then tested five sequences on NQ. The frequent ones won often but lost net: close targets, wide stops.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

@hhxk233 Thanks! The failures teach us more than the wins. That's why we publish them.

[MGEQuant @MGEQuant]

We tested 13,608 trading rules on NQ. One cleared every discovery gate, then lost to its random control on the confirmation data.
New note on our Substack (link in bio).
#MGEQuant #Backtesting #Futures

[MGEQuant @MGEQuant]

13,608 rules searched, one survived, then failed

On NQ, one rule cleared every discovery gate. On the confirmation data it lost to its random control. The best of thirteen thousand tries is mostly luck.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

Exhaustion Coil: the bug that cancelled every setup

The coil's boundaries weren't locked when it was confirmed, so every setup invalidated itself. Fixed; it stays in research.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

Every candle answers one question: did it break the last one's high, its low, both, or neither?
Four answers turn a busy chart into a sequence you can read.
New note on our Substack (link in bio).
#MGEQuant #PriceAction #NinjaTrader

[MGEQuant @MGEQuant]

FlowEdge: five defects a live chart exposed

INDICATOR ONLY. Heatmap rows built from the wrong bar, layers misplaced, values updating every tick. All five fixed and re-checked.

⚓ The market earns the trade — not me

#MGEQuant #NinjaTrader

[MGEQuant @MGEQuant]

EMA-cross scalper: just under breakeven

Its 30/90-tick bracket needs a 25% win rate. Over ~5 years of NQ, 1,088 trades won 24.3%. With costs: -0.296R a trade.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

Lower-low candle runs: 568 configurations

All four sessions, every entry style and stop/target width. Zero were positive and significant. The best one only broke even.

⚓ The market earns the trade — not me

#MGEQuant #NQ

[MGEQuant @MGEQuant]

Two wins or one loss, then done for the day. A count, not a dollar amount, because a count does not negotiate.
New note on our Substack (link in bio).
#MGEQuant #TradingDiscipline #Futures

[MGEQuant @MGEQuant]

Exhaustion Fade: best of 12 tries

+0.352R on 71 short trades. Corrected for the 12 tries, the 95% interval is -0.162 to +0.866. That includes zero: not enough evidence.

⚓ The market earns the trade — not me

#MGEQuant #QuantTrading

[MGEQuant @MGEQuant]

Two wins or one loss, then done for the day

Our strategies have it built in; hand trades get it from a chart tool. It counts trades, not dollars, and has no undo button.

⚓ The market earns the trade — not me

#MGEQuant #TradingDiscipline

[MGEQuant @MGEQuant]

Two hand-read setups, tested as code

Continuation and Attrition Break. In code, neither beat the same trade taken the opposite way. No edge found; the human read is untested.

⚓ The market earns the trade — not me

#MGEQuant #NQ

[MGEQuant @MGEQuant]

Trader A wins 9 out of 10 and loses money. Trader B wins 4 out of 10 and makes it.
The number on the ad is the one number that cannot tell you whether a trader makes money.
Guru check #1 on our Substack (link in bio).
#MGEQuant #Trading #RiskManagement

[MGEQuant @MGEQuant]

Renko pullback: the edge was a fill artifact

Entries were booked at prices the market had already passed. At realistic fills the average fell from +0.364R to +0.021R.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

Win rate is not the question

A wins 90% at 0.2R and loses 3R: -0.12R a trade. B wins 40% at 2R and loses 1R: +0.2R. Over 100 trades, A ends below zero 88% of the time.

⚓ The market earns the trade — not me

#MGEQuant #RiskManagement

[MGEQuant @MGEQuant]

Factor gate: 24 tests, nothing passed

0 of 24 reached even uncorrected p<0.05. A random-number factor passed one of our own rules, so we dropped the rule.

⚓ The market earns the trade — not me

#MGEQuant #Backtesting

[MGEQuant @MGEQuant]

Our own backtest had two bugs, and the second one made the control look brilliant.
A control that beats the real trade is a bug alert until proven otherwise.
New note on our Substack (link in bio).
#MGEQuant #Backtesting #QuantTrading

[MGEQuant @MGEQuant]

Trend-maturity filter: worse out-of-sample

Significant in-sample (t=2.05), but it failed the multiple-test correction. Frozen onto 2024-2026, the average fell from +0.247R to +0.191R.

⚓ The market earns the trade — not me

#MGEQuant #WalkForward

[MGEQuant @MGEQuant]

NQ returns: about 40 tests, nothing exploitable

Hour of day, session, weekday, autocorrelation: nothing survived. Volatility is the exception: the 5-min range runs from 24 ticks at midnight to 124 at 10:00 ET.

⚓ The market earns the trade — not me

#MGEQuant #NQ