Why a strong technical setup can still underperform — and why it's mostly timing
Our healthiest tag lagged the market on a raw basis. The culprit wasn't the picks — it was the day you'd have to buy them.
Here is an uncomfortable number from our own data. Our healthiest technical tag — the one the scanner reserves for stocks in a clean uptrend passing every check — underperformed the market over the next three days, on a raw basis, by about a third of a percent. That is the opposite of what a "strong setup" label is supposed to do.
Rather than quietly drop the finding, we went looking for the cause. It turned out not to be the stock selection at all. It was the day you would have to buy them.
| Bucket | Horizon | Excess vs universe | t-stat |
|---|---|---|---|
| Strong-setup tag (raw) | 3 days | −0.31% | −3.1 |
| Same tag, after matching on same-day move | 3 days | −0.12% | −1.2 (ns) |
The mechanism: it can only fire on a good day
The strong-setup tag has a built-in condition — it will not flag a stock that closed down more than about a percent on the day. That rule exists for a sensible reason (you do not want to call a falling knife a healthy setup), but it has a side effect: the tag can only appear at the close of a day the stock was already up. By construction, it flags stocks that just had a strong session.
Now think about what happens next. A stock that jumped today tends, on average, to give a little of it back over the following one to five days — ordinary short-term mean reversion. A trader reading the tag after the close and buying at the next morning's open is buying after the pop, right into that small giveback. The setup did not pick badly. It flagged the stock on the single day it is statistically worst to chase.
How we proved it was timing, not selection
We matched every strong-setup stock against non-tagged stocks that had the same same-day move, then re-measured. If the tag were genuinely selecting worse stocks, the gap would survive. It did not: the underperformance collapsed from −0.31% to about −0.12%, and the statistical significance vanished (t fell from ~3.1 to ~1.2). Almost the entire effect was explained by one thing — these stocks had already moved.
How we measured this
Daily equal-weight excess over the scanned universe, statistics computed over the day series with Newey–West standard errors for overlapping windows. The "matched" row conditions on the stock's own same-day return before comparing, isolating selection from timing.
What it means for how you use a screen
A strong technical setup is a candidate to watch, not a market order for tomorrow's open. The technicals can be perfectly healthy and the timing still poor, simply because the flag tends to arrive on an up-day. The disciplined use is to note the name, then wait for a pullback or a defined entry — rather than buying the extended close the label was printed on.
Why we still show it
None of this means the tag is useless — it means the tag marks structure, not entry timing. Structure and timing are two different questions, and conflating them is exactly the mistake this data exposes. StockLearn shows the structure and leaves the timing to you, which is why it never tells you to buy anything.
Key takeaways
- The strong-setup tag underperformed by ~0.31% over three days on a raw basis.
- Almost all of that is timing: the tag can only fire after an up day, and up-day pops tend to give back.
- Matching on the same-day move erased the effect (t fell from ~3.1 to ~1.2, not significant).
- A strong setup is a candidate to watch, not a reason to chase the next open.
- Structure and entry timing are different questions — the tag answers the first, not the second.
See the scanner these numbers come from
StockLearn runs this read on ~2,000 NSE stocks every evening. Nifty 50 is free, no login.
Browse today's scan →This article reports measured technical results from StockLearn's own scan history over a specific, limited period. It is educational research, not investment advice or a recommendation to buy or sell any security. Past statistical tendencies do not predict future returns.