Technical trend indicators

Download the data and code

• technical_indicators_monthly.csv12,820 rows · 732 KBEvery rule for every asset and month, with the trend uncertainty index, buy share and disagreement.
  Download signals
• technical_summary_monthly.csv910 months · 163 KBOne row per month: trend uncertainty, buy share and disagreement for each asset, and the commodity averages.
  Download summary
• update_technical.py23 KBPython script that downloads the prices and rebuilds both files. Needs pandas.
  Download Python
• update_technical.m21 KBMATLAB script that does the same.
  Download MATLAB
• replicate_neely.py4 KBChecks the rules against the paper’s replication files, if you have them.
  Download

The paper’s S&P 500 signals for 1950 to 2011 are written into both scripts; everything later is computed from prices downloaded when the script runs. Python and MATLAB give identical output.

Read the code

Loading the script…

The fourteen rules

Each rule gives a buy signal (1) or a sell signal (0) at the end of every month. The S&P 500 uses all fourteen. The commodities use the eight price rules, because there is no public volume series for them.

Three summaries are given for every asset and month:

• Trend uncertainty index, trend_uncertainty: the share of the asset’s rules on a sell signal, from 0 (every rule on buy) to 1 (every rule on sell).
• Share on buy, buy_share: one minus the index.
• Trend disagreement, disagreement, a secondary series: 1 − |2 × buy_share − 1|, which is 0 when every rule points the same way and 1 when they are split evenly.

Replication

Before extending anything, the indicators were rebuilt from the data file in Neely, Rapach, Tu and Zhou’s replication package with the rules in their code. All 12,468 indicator-months in their results file are reproduced exactly, and so is every bivariate in-sample regression in their Table 2. The table shows the slopes; the t-statistics and R-squared values match to the same precision. replicate_neely.py repeats the check given their two files.

January 1951 to December 2011, White t-statistics, as in the paper. The other eight indicators match equally closely.

The extension then has to use public data. Rebuilt from the S&P 500 index on Yahoo Finance over the paper’s own sample, the eight price rules match the paper in every month and the six volume rules in 99.6% of months. The share of rules on a buy signal has a correlation of 0.9993 with the paper’s, and the majority signal is the same in every month. The S&P 500 series therefore keeps the paper’s values to December 2011 and uses the public data from January 2012.

Assets and sources

Each commodity starts in the January after the last year before 1990 in which at least half of its monthly prices were unchanged from the month before, and signals begin once there are twelve months of prices. That leaves out eras of fixed or posted prices, such as gold at $35 an ounce and posted oil prices, where a trend rule has nothing to measure. Iron ore, coal and European gas are left out altogether because their prices were set in annual contracts for decades.

The trend uncertainty index

The trend uncertainty index is the share of an asset’s technical rules on a sell signal. For the stock market it behaves like a measure of uncertainty: falling markets are volatile markets, and the index moves with the VIX. Trend disagreement, the secondary series, was tested as an alternative and does not: it peaks at turning points, when the short rules have changed direction and the long ones have not.

S&P 500 realised volatility from daily index returns, December 1950 to September 2026; VIX from 1990. Computed in October 2026.

For commodities the index measures how widely prices are trending down, but it does not track their volatility. Across the eighteen, the average index has a correlation of −0.17 with the average absolute monthly price change, because commodity prices tend to be most volatile in booms rather than slumps. Treat the commodity index as a measure of how many trend rules point down, not of price risk. Commodity disagreement is highest at turning points such as June 2009, March 2002 and August 2011.

Before you use it

• Month-end against monthly average. S&P 500 signals use the month-end close, as in the paper. Commodity signals use the World Bank’s monthly average prices, which are smoother, so their signals change less often than month-end signals would.
• Timing. A signal for month t uses prices up to the end of t, so it can be used to forecast month t+1. The World Bank publishes each month’s averages early in the following month, so commodity signals for t are public a few days into t+1.
• Revisions. The World Bank occasionally revises recent prices, and Yahoo Finance has revised historical S&P 500 volume, which is why the public volume rules match the paper in 99.6% of months rather than all of them.
• Ties. When two moving averages are equal, the rule gives a sell signal, as the original code’s strict comparison does. Repeated prices in the commodity data make this more common than for the S&P 500.
• Data terms. Only the signals are published here, not the prices. The scripts download the S&P 500 index from Yahoo Finance and the commodity prices from the World Bank each time they run.

Why it matters

Neely, Rapach, Tu and Zhou (2014) showed that these simple trend rules forecast the US equity premium over 1951 to 2011 about as well as the standard macroeconomic predictors, and that the two kinds of predictor carry different information: technical indicators pick up the fall in the premium near business-cycle peaks, macroeconomic variables the rise near troughs. Updating the signals, and computing the same rules for oil, metals and crops, lets that evidence be tested on data the paper never saw and on markets it did not cover.

How to cite

If you use the signals, please cite the paper that defined them, and the World Bank for the commodity prices. A reference for this page is included in case it helps.

@article{NeelyRapachTuZhou2014,
  author  = {Neely, Christopher J. and Rapach, David E. and Tu, Jun and Zhou, Guofu},
  title   = {Forecasting the equity risk premium: The role of technical indicators},
  journal = {Management Science},
  volume  = {60},
  number  = {7},
  pages   = {1772--1791},
  year    = {2014},
  doi     = {10.1287/mnsc.2013.1838}
}

@misc{WorldBankPinkSheet,
  author       = {{World Bank}},
  title        = {Commodity Price Data (The Pink Sheet), monthly prices},
  howpublished = {\url{https://www.worldbank.org/en/research/commodity-markets}},
  year         = {2026}
}

@misc{PybisTechnical,
  author       = {Pybis, Sam},
  title        = {Technical trend indicators for the {S\&P} 500 and 18 commodities, monthly, 1950 to September 2026},
  year         = {2026},
  howpublished = {\url{https://sampybis.co.uk/updated-data/technical.html}},
  note         = {Updated 20 October 2026}
}