KodoQuant · the record

The layer can only take exposure off.

KodoQuant runs a legacy rule set that decides direction, and one layer on top of it that decides size. The layer has a hard ceiling equal to the baseline’s fixed position, so it can subtract exposure and never add it. Twenty-four years later it is ahead on return and shallower on drawdown. The table below is that whole claim, in figures.

The sheets themselves are behind the sign-in ↓

Two pens, one set of rules

Both runs take the same signals on the same bars. The baseline holds one full contract whenever it is in the market. KodoQuant holds between none and one, in tenths. Everything in the table below is a consequence of that single difference.

2002-10-09 → 2026-07-09 · simulated · fixed size, no leverage
	KodoQuant	Baseline
Final equity	$927,362	$865,835
Sharpe	1.204	1.008
Max drawdown	−8.26%	−12.85%
CAGR	9.84%	—
Closed trades	388	140
Orders	783	281
Largest position	1 NQ	1 NQ

Beating a baseline on return alone is a leverage decision. Beating it on return and drawdown, under the same ceiling, is not — that only comes from being smaller at the right times.

Multiplying position size by any fixed number scores exactly 100 on the project’s own measure. Only the timing of the reductions can move it, which is the claim being made and the one worth testing.

Baseline CAGR is not stated in the run that produced this record, so the cell is empty rather than estimated.

The pen cannot go above the line

Exposure is carried in MNQ contracts at $2 a point, so a position can be split into tenths. Ten MNQ is one NQ — exactly the baseline’s fixed size, and a hard ceiling asserted in the strategy code and its tests. Across all 6,007 bars the largest position ever held is 10.0.

The record holds a position on 5,333 days. On 4,194 of them it is at full size, matching the baseline exactly. The remaining 1,139 days are the entire strategy: the layer is subtracting, and there is no arrangement of its inputs that lets it do anything else.

A reader who wants to disprove this looks for one bar above the cap. The exposure sheet is behind the sign-in, and there is not one in it.

What it costs, in the same ink

Holding less through a flagged stretch means lagging when the flag is wrong. That is a recorded property of this strategy, not a caveat about it.

It lags 7.0% of the time. On 423 of 6,007 days KodoQuant sits behind the baseline it beats overall. An allocator reading a monthly statement will see those months.

It scores 143.8 against a target of 150. 108.0% of baseline return plus a 35.7% drawdown cut. Up from 114.0 before the day-quality scores existed, and short of the target the project set itself.

Its ceiling is 142.8%. Avoiding every single losing trade in the record would still cap return at 142.8% of the baseline. This layer cannot be scaled into a different order of result.