Long-term trend
SMA200 is calculated from the current session and the previous 199 security sessions.
MARKET DATA HUNTER
RESEARCH INVESTIGATION 001
Trend Quality
PROSPECTIVE TRACKING
THE QUESTION
Does persistent strength above a long-term trend predict future returns?
Trend Quality investigates whether stocks that remain consistently above their long-term trend subsequently outperform the market.
Historical testing produced promising evidence, but the effect was not uniform through time. The methodology has been frozen and the surviving hypothesis has moved into prospective validation.
Historical Development
COMPLETEStatistical Challenge
COMPLETERules Frozen
COMPLETE04
Prospective Evidence
AWAITING OBSERVATIONSINVESTIGATION OVERVIEW
Start with an idea.
Then try to break it.
A historical pattern is not the finish line. It is a hypothesis that needs increasingly difficult tests.
This investigation separates historical development from statistical validation and, ultimately, from evidence gathered after the research rules were fixed.
HOW TREND QUALITY WORKS
Measure strength.
Reward persistence.
Trend Quality does not ask only whether a stock is above its long-term trend on one particular day. It measures how far above that trend the stock has remained across an entire 63-session window.
THE SIGNAL
Trend Quality 63d
For each of the most recent 63 security sessions, measure the stock's adjusted closing price relative to its 200-session moving average. Then take the arithmetic mean.
Trend Quality63=
mean63[
Adjusted CloseSMA200
− 1]Persistence
The signal averages the stock's distance from SMA200 across exactly 63 security sessions rather than relying on a single observation.
Trend confirmation
A stock is eligible only when Adjusted Close is above SMA200 and SMA50 is also above SMA200.
FROM SIGNAL TO TEST
The historical test followed a fixed sequence.
01
Universe
S&P 50002
Screen
Trend + liquidity03
Rank
Highest signal first04
Portfolios
Top 3 / 5 / 1005
Hold
20 sessions06
Compare
Versus SPYFROZEN SPECIFICATION
The rules that define the historical signal.
Signal
Trend_Quality_63d
Signal window
Exactly 63 security sessions
SMA200
Current session + previous 199 sessions
SMA50
Current session + previous 49 sessions
Trend filter
Adjusted Close > SMA200 and SMA50 > SMA200
Ranking
Signal descending; ticker ascending for ties
Portfolio
Equal-weight Top 3 / Top 5 / Top 10
Holding period
20 security sessions
Benchmark
SPY
Direction
Long-only; no rebalancing within the holding period
HISTORICAL EVIDENCE
The pattern looked promising.
Then we tried to break it.
Historical testing asked whether stocks with higher Trend Quality subsequently outperformed the market over the next 20 security sessions.
The initial results were encouraging. But a backtest can only describe what happened in the sample used to develop the idea. The next question was whether the evidence could survive harder tests.
36,688Security observations
144Evaluation dates
20Session forward horizon
SPYMarket benchmark
PORTFOLIO RESULTS
Average historical excess return versus SPY
Equal-weight portfolios formed from the highest-ranked eligible stocks produced positive average excess returns in the historical development sample.
Average excess return
vs. SPY · historical development sampleAverage excess return
vs. SPY · historical development sampleAverage excess return
vs. SPY · historical development sample!
THE IMPORTANT PART
The effect was not uniform through time.
A strong full-sample average can hide instability. When the historical record was divided into fixed time periods, the relationship was weak or negative in the earlier portions of the sample and substantially stronger later.
That does not erase the historical result. It changes what the result can reasonably tell us.
TEMPORAL STABILITY
One average. Three very different periods.
Mean rank correlation between Trend Quality and subsequent 20-session returns.
EARLY2015-2018
-0.003Mean Spearman rank correlation
Little evidence of a positive relationship.MIDDLE2019-2021
-0.021Mean Spearman rank correlation
The relationship was negative in this period.LATER2022-2026
+0.043Mean Spearman rank correlation
The positive relationship was concentrated here.WHAT WE LEARNED
Promising evidence
is not the same as proof.
The historical study produced a pattern worth continuing to investigate, but it also exposed an important weakness: temporal instability.
Instead of adjusting the rules to improve the historical record, the methodology was frozen. The surviving hypothesis would have to face new observations collected only after the research rules were frozen.
STATISTICAL CHALLENGE
Could chance have
produced the pattern?
Historical performance alone cannot answer that question. So the signal was subjected to a randomization test designed around a simple null hypothesis.
Within each evaluation date, the relationship between Trend Quality and subsequent returns was repeatedly broken by random permutation while preserving the date-level structure of the data.
01
THE PRIMARY NULL
Trend Quality contains no positive predictive information about subsequent 20-session returns.
If that null were true, randomly rearranging the signal within each evaluation date should routinely produce results comparable to the observed relationship.
RANDOMIZATION TEST
100,000permutations
Each permutation created a version of the historical sample in which the within-date link between signal rank and future return was intentionally broken.
01
Keep each date intact
Preserve the historical cross-section and evaluation-date structure.
02
Break the signal-return link
Randomly permute the signal within each evaluation date.
03
Recalculate the statistic
Measure the rank relationship produced under the null.
04
Repeat 100,000 times
Compare the observed result with the randomized distribution.
PRIMARY RANK TEST
The observed relationship sat near the edge of the null.
The primary statistic was the mean Spearman rank correlation between Trend Quality and subsequent 20-session security returns across the historical evaluation dates.
Historical development sample
Frozen positive-direction null
Reported as additional context
INTERPRETATION
Statistical evidence is not the same as stability.
Under the frozen one-sided test, the observed mean rank relationship was uncommon enough under the randomized null to cross the pre-specified threshold.
But that result does not show that the relationship was equally present throughout history, nor does it establish that the pattern will continue in future data.
TWO DIFFERENT QUESTIONS
Both matter.
A
Could this result arise by chance?
The permutation test challenged the statistical relationship under a frozen null model.
One-sided test crossed threshold
B
Was the relationship stable through time?
The fixed temporal split showed that the positive relationship was concentrated in the later historical period.
Concern remained
THE DECISION
Do not optimize it.
Freeze it.
The statistical challenge gave us a reason to keep investigating, not a reason to rewrite the rules around the historical result.
The signal definition, filters, ranking rules, portfolio construction, holding period, and benchmark convention were frozen. From that point forward, the next meaningful evidence would have to come from genuinely new observations.
PROSPECTIVE VALIDATION
Now the rules
cannot move.
Historical research can reveal a pattern, but it cannot recreate the conditions of a genuinely unseen test.
From the moment the methodology was frozen, a qualifying observation would count only if the required inputs and research snapshot existed before the outcome became known.
CURRENT RESEARCH STATUS
0
Pristine prospective observations
Current validated sequenceAWAITING OBSERVATIONS
No post-freeze observation has yet satisfied the full prospective evidence requirements.
WHAT COUNTS AS PROSPECTIVE?
The evidence must exist before the answer does.
The prospective policy is intentionally strict. A result cannot be promoted into the live evidence sequence simply because the necessary historical data can be reconstructed later.
Future observations only
Retrospective reconstruction does not become prospective evidence.
No backfilling
A missed scheduled observation remains missed and is never inserted later.
Snapshot before outcome
The research inputs must be captured and preserved before the forward return is known.
Rules stay frozen
A material methodology change ends the original pristine sequence rather than rewriting it.
MISSED OBSERVATIONS
We do not rewrite the clock.
Three scheduled dates passed before a qualifying pristine snapshot was captured. They remain part of the audit trail, but they do not count as prospective evidence.
MISSED - NOT PROSPECTIVE
MISSED - NOT PROSPECTIVE
MISSED - NOT PROSPECTIVE
THE RESEARCH BOUNDARY
Before the freezeHistorical evidence
RULES FROZEN
WHAT COMES NEXT
After the freezeProspective evidence only
WHERE THE INVESTIGATION STANDS
The honest result
right now is:
we do not know yet.
Trend Quality produced historically interesting evidence and survived a pre-specified one-sided statistical challenge. It also showed meaningful instability across historical periods.
The methodology is now frozen. Until genuinely prospective observations accumulate, the historical result remains a hypothesis under investigation rather than a prospective finding.
AWAITING OBSERVATIONSProspective evidence: 0 pristine observations