Applying the Signal: A Before-and-After Test

When you evaluate a new trading signal, the practical question is: does this signal improve a strategy that I am already running?

In this post, we run a simple “before and after” comparison using the same trend-following strategy in two configurations:

• Trend-only: a simple trend-following intraday strategy.
• Trend + Signal: the trend strategy only takes long/short positions when both the trend filter and our signal agree (a strict confirmation filter).

This is an intentionally naive way to apply our signal, but it creates a clean, easy-to-interpret picture of what changes when the signal is added.

When we refer to “our signal,” we mean the Level II / order-book model we described in Introducing our model.

‍

The two strategies

1) Trend-only: comparison baseline

The baseline is a simple trend follower. Trend following has largely fallen out of favor in recent years, and in many markets it has generally underperformed. This is especially true when regimes become choppy or mean-reverting. In this sample, it draws down over time.

Implementation note: the baseline here uses the same execution and signal-to-position framework we describe in How we implement the benchmarking strategy behind the signal. We’re holding the mechanics consistent so the comparison stays focused on the impact of adding the signal.

2) Trend + Signal: confirmation filter

The combined strategy uses the same trend logic, but adds our signal as a trade-quality filter:

• Go long only when trend = up and signal = up
• Go short only when trend = down and signal = down
• Otherwise: make no trade

This “agreement only” approach is intentionally strict. It will miss some winning trades but the goal is to avoid low-quality trades where the trend indicator is flashing but the microstructure signal disagrees.

What the window shows (NQ and ES)

We assume that we start with $100,000 in 24Q3, and we track performance through 25Q4.

Why a microstructure signal can help a trend strategy

Trend systems often struggle when price action becomes choppy, mean-reverting, or dominated by short-lived liquidity conditions. A Level II / order-book-driven signal can add value by detecting when:

• liquidity and participation conditions are deteriorating,
• short-term buying/selling pressure doesn’t match the trend signal,
• there are more price reversals and “false breaks”

In practice, that tends to show up as fewer trades, but better trades which is exactly what a strict confirmation filter is designed to do.

Other ways in which to blend your strategy and our signal

Using our signal as a strict gate is not the only way to combine signals and strategies.

A more sophisticated approach is to treat “trend” and “signal” as experts and blend them, for example:

• Weighted experts / ensemble allocation: size positions based on each expert’s confidence or recent performance.
• Soft agreement instead of hard agreement: reduce exposure when they disagree rather than going fully flat.
• Regime-aware weights: let the strategy learn when trend should dominate vs. when microstructure should dominate.

That’s important because strong systems usually don’t ask, “Which one is right?” They ask, “How much should I trust each input right now?”

This post is meant to show the directional impact clearly: even a strict, simple integration improved results substantially versus trend alone.

Bottom line

A trend-only strategy can be a useful benchmark—but as this 5-quarter snapshot shows, it can also bleed in the wrong conditions.

Adding our signal, even as a strict confirmation filter, creates a clear before/after improvement:

• NQ: trend-only drawdown → Trend+Signal profitability
• ES: materially smaller drawdown and better capital preservation

Risk note: These results are for research/backtest-style illustration and are not investment advice. Trading involves risk, including the risk of loss.