Get AMT Gold ↗
← Research library

Trend Bender Dashboard

2019-01-02 to 2026-05-27 | 1,899 sessions | ES 5-min bars | WMA 5 / EMA 6 / Shift 1

View Study Details→

The question & finding

Do color transitions carry through the session?

For ES, buy transitions had 51.8% session-close continuation and sell transitions had 45%. Direction matters; a transition is not, by itself, proof of a profitable entry.

20,127 signals across 1,899 sessions. These are directional outcomes at the session close, not trade win rates. Historical results do not guarantee future outcomes.

Buy vs Sell Accuracy

Session-Close Continuation Rate

BUY — bearish-to-bullish transition

BUY
51.8%
48.2%

SELL — bullish-to-bearish transition

SELL
45%
55.0%

Buy signals outperform sell by 6.8pp. This reflects the structural long bias in ES — bullish transitions hold better than bearish ones.

Signal Distribution

BUY
50%
SELL
50%

Count Breakdown

Buy total: 10,065
Sell total: 10,062
Buy continued: 5,214
Sell continued: 4,528

Buy and sell signals are roughly evenly split. The edge lies in the directional accuracy, not signal count.

Sample counts & setup
20,127
Total Signals
10,065
Buy Signals
10,062
Sell Signals
51.8%
Buy Accuracy
45%
Sell Accuracy
10.6
Avg Signals/Session
Parameter Set & Time-of-Day

Accuracy by Parameter Set

35%40%45%50%55%60%51.8%45%DefaultWMA 5 / EMA 6 / Shift 1~10.6/session52.2%44.9%MediumWMA 8 / EMA 10 / Shift 2~5.7/session51.1%44.6%SlowWMA 14 / EMA 14 / Shift 3~3.7/session
Buy AccuracySell Accuracy

All three parameter sets produce virtually identical accuracy (~48.4%). Slower settings reduce signal count from ~10.6 to ~3.7/session but don't improve edge.

Time-of-Day Accuracy

35%40%45%50%55%60%55.1%41.8%First Hourn=1,24853.8%44.4%Mid-Morningn=5,02053.7%45.1%Middayn=6,61948%45.7%Laten=7,240
BuySell
Yearly Stability

Accuracy by Year (Buy / Sell Split)

35%40%45%50%55%60%2019n=28912020n=28502021n=27572022n=27322023n=26232024n=26702025n=26832026n=921
BuySellCombined

The buy/sell asymmetry is consistent across all years. Combined accuracy (cyan dots) stays flat around 49-50% while buy (green) consistently exceeds sell (red).

Forward Accuracy & Trend Duration

Forward Bar Accuracy

After a signal, is price moving in the predicted direction after N bars?

35%40%45%50%55%60%51.9%46.4%5 barsn=18k52.7%46.2%10 barsn=17k53.2%44.7%20 barsn=14k54.5%43.5%30 barsn=11k
BuySell

Trend Duration (bars)

How long does each trend streak last before flipping? On 5-min bars with short MAs, trends are short-lived.

Bullish Avg
6.6
11,045 streaks
Bearish Avg
5.6
10,982 streaks
Bullish Median
4
bars per streak
Bearish Median
4
bars per streak

Median trend duration is 4 bars (20 min). Most crossovers reverse quickly — the indicator is highly sensitive with default settings.

MFE / MAE Analysis

Buy Signal Excursion (points)

Avg MFE
14.05
Favorable excursion
Avg MAE
15.56
Adverse excursion
Median MFE
9
Favorable excursion
Median MAE
9.25
Adverse excursion

MFE/MAE ratio: 0.90:1 avg, 0.97:1 median.

Sell Signal Excursion (points)

Avg MFE
14.77
Favorable excursion
Avg MAE
14.85
Adverse excursion
Median MFE
8.25
Favorable excursion
Median MAE
9.75
Adverse excursion

MFE/MAE ratio: 0.99:1 avg, 0.85:1 median.

Definitions & interpretation

What is Trend Bender?

Trend Bender detects trend direction by comparing a Weighted Moving Average (WMA) to a shifted Exponential Moving Average (EMA) on 5-minute bars. When the WMA is above the shifted EMA, the bar is colored bullish; when below, bearish. A color transition signals a potential trend change.

BUY Signal — Bearish-to-bullish transition. WMA crosses above the shifted EMA, suggesting the short-term trend is turning up.
SELL Signal — Bullish-to-bearish transition. WMA crosses below the shifted EMA, suggesting the short-term trend is turning down.

Key findings

  • Combined accuracy is sub-50% — At ~48.4% combined, individual crossovers do not carry a standalone directional edge. The signal is noisy even on 5-min bars with ~10.6 transitions per session.
  • Persistent buy/sell asymmetry — Buy signals (bearish→bullish) consistently outperform sell signals by ~6-9pp across all parameter sets and years. This reflects the structural long bias in equities.
  • Parameter sensitivity is low — Default, Medium, and Slow settings all produce nearly identical accuracy (~48%). Longer settings reduce signal count from ~10.6 to ~3.7/session but don't improve edge.