Trend
10 points each for price above the 50-day average, the 50-day above the 200-day, and price above the 200-day average.
RESEARCH LIBRARY
RESEARCH INVESTIGATION 002
The 100-Point Stock Score
CLOSED · DID NOT HOLD UP
THE QUESTION
Can a blended score of five popular technical signals pick stocks that beat the market?
This was the first stock-ranking model Market Data Hunter tested. It combined trend, momentum, relative strength, nearness to a 52-week high and volume into a single score out of 100.
Tested on the stocks that were actually in the S&P 500 at each date from 2015 to 2026, the highest-scoring stocks trailed SPY on average. The model was closed and is kept here as part of the research record.
Rules Frozen
COMPLETEPortfolio Test
COMPLETEDiagnostics
COMPLETEClosed
ARCHIVEDWHY PUBLISH A FAILURE
Most scoring systems
are never tested.
Stock screeners and newsletters often rank stocks with a points-based blend of indicators. The weights usually sound sensible, and the examples shown are usually winners.
This study fixed the score before looking at any results and then measured what happened to every eligible stock it ranked. The answer was not what the score's design suggested.
HOW THE SCORE WORKED
Five signals.
One hundred points.
Each stock received points for five characteristics that are widely used in technical stock selection. Stocks were then ranked from highest to lowest total score on each rebalance date.
Momentum
Price change over the last 63 trading days: 25 points at +20% or more, stepping down to 10 points for any gain.
Relative strength
63-day return minus SPY's 63-day return: 20 points at 15 points ahead or more, down to 8 points for any lead.
Near 52-week high
15 points within 2% of the 52-week high, falling to 4 points within 20% of it.
Volume
Today's volume against its 20-day average: 10 points at 1.5 times or more, 4 points at 1.0 times.
FROZEN TEST DESIGN
The rules were fixed before any performance was calculated.
Universe
Stocks in the S&P 500 on each rebalance date (point-in-time membership)
Eligible stock-dates
68,001
Rebalance
Every 20 trading days, 144 dates
Portfolios
Equal-weight Top 3 / Top 5 / Top 10
Holding period
20 trading days
Benchmark
SPY over the same 20 trading days
Trading costs
0, 5, 10 and 25 basis points per trade
Missing data
5.27% of stock-dates could not be verified and were excluded, not filled in
THE RESULT
The top-ranked stocks
trailed the market.
Every portfolio size lost to SPY on average, before any trading costs. The highest-scored stocks beat SPY in fewer than half of the 144 periods.
Adding realistic trading costs only widened the gap. At 10 basis points per trade, the Top 3 portfolio trailed SPY by an average of 0.51% every 20 trading days.
Average excess return per period
Beat SPY in 47% of periods · before costsAverage excess return per period
Beat SPY in 44% of periods · before costsAverage excess return per period
Beat SPY in 42% of periods · before costsEvery eligible stock-date from 2015 to 2026, grouped by its score, with the average return over the next 20 trading days minus SPY over the same days. Higher scores did not lead to better results. The lowest-scoring group was the only one ahead of SPY on average, and the top group (highlighted) was behind it.
!
HOW TO READ THIS
No relationship is not the same as a reverse relationship.
It is tempting to flip the score and buy the lowest-ranked stocks. That would be a new idea chosen after seeing the results, which is exactly the kind of data snooping this research is designed to avoid.
The differences between score groups are small compared with how much individual stock returns vary, and no statistical test was run on them. The fair reading is that the score did not help, not that low scores are a buy signal.
RANK TEST
Did higher scores line up with higher returns?
Mean Spearman rank correlation between the score and each stock's next 20-day return, measured within each rebalance date. Zero means no relationship.
EARLY2015–2018
−0.012Mean rank correlation
Slightly negative. Positive on 55% of dates.MIDDLE2019–2021
−0.020Mean rank correlation
Slightly negative. Positive on 50% of dates.LATER2022–2026
+0.002Mean rank correlation
Essentially zero. Positive on 51% of dates.THE DECISION
Close it.
Do not tune it.
Across the full sample the mean rank correlation was −0.009, and the correlation was positive on only 52% of dates. No period showed a useful positive relationship.
Adjusting the weights until the backtest looked good would have produced a model fitted to this history and nothing else. Instead, the score was frozen as a permanent baseline and any new idea had to be specified as a separate hypothesis before testing.
WHAT THIS DOES NOT SHOW
A failed blend,
not failed ingredients.
The result applies to this exact score, these weights, this 20-day horizon and this universe. It does not show that trend, momentum, relative strength, nearness to highs or volume carry no information under other definitions.
That is why the next two investigations each tested a single, pre-specified idea rather than another blend.
LIMITATIONS
Recorded with the result.
Membership history
Point-in-time S&P 500 membership ends July 13, 2026
Unverified observations
3,855 stock-dates (5.27%) excluded; the gaps may not be random
Prices
Adjusted daily prices; not actual trade fills
Costs
Fixed sensitivity assumptions, not realized trading costs
Later sample
2022–2026 was not an untouched out-of-sample test
Data sources
Daily prices from Yahoo Finance, with Tiingo for some delisted stocks