Status: pre-registered on 23 September 2026, data collection open. Hypotheses, definitions and the analysis plan below were fixed before the first response and will not be changed. Part A (market) runs on DAX, NQ and Dow, M1 CFD data 2015–2026, net of spread and slippage; its result is held back until the questionnaire closes so it cannot steer participants. Part B (perception) collects answers on the questionnaire page: twelve situations, four states, about ten minutes. Collection closes on 31 October 2026 or at 50 complete participants, whichever comes first. No trading recommendation.
1. The observation
An observation from my own trading, September 2026. When I have been in a move from the start, the pullback looks like a place where I might have to pay twice. I re-enter later, at a deeper level, or not at all, but I add to the position I already have. When I have missed the same move, the same pullback looks like a gift, and I am in earlier and at a shallower level. Same chart, two decisions.
That is not surprising. It is still worth measuring, because one of the two versions of me is wrong about the money, and I do not know which. The price path is identical in both cases. If the decision differs, the difference lives in the trader, not in the market.
2. Two questions, kept apart
The study separates what the market does from what the state does.
The market question. After a directional move, what is a pullback entry worth, conditional on what can be observed at that moment: size of the move, its age, depth of the pullback, time of day? And is adding to a running position worth anything different from a new entry at the same price and stop? The market cannot see the difference, so the honest expectation is: no.
The perception question. Does knowing that you were in the move, or that you were stopped out of it, shift the threshold at which you re-enter, and in which direction? And once the forward path is known: does the shifted threshold make or lose money against the unshifted one?
Only the second question needs participants. The first needs data and a definition fixed in advance.
3. Hypotheses
Each hypothesis has a prediction that can fail.
| # | Hypothesis | Prediction that would confirm it | What would refute it |
|---|---|---|---|
| H1 | Having been in the move raises the re-entry threshold. | In the "in the move" framing the entry rate at the decision point is lower and the accepted pullback depth is deeper than in the "flat" framing. | No difference, or the reverse. |
| H2 | A break-even stop-out raises it further. | The "stopped at break-even" framing shows the lowest entry rate and the highest share of "not in this move any more". | Entry rate equal to or above the flat framing. |
| H3 | Missing the move lowers the threshold. | In the "flat" framing participants enter earlier, at shallower depths, and more often report the move as "missed". | Entry rate in flat is not above the other framings. |
| H4 | Objectively, participation does not matter. | In part A, adding to a runner and a fresh entry at the same price and stop earn the same R per unit of new risk, and the pullback-entry expectation depends only on observable state. | A measurable difference, which would mean the definitions leak information. |
| H5 (exploratory) | The size of the open profit changes the re-entry, not just its existence. | The "+5 R" state differs from the "+1.5 R" state in entry rate or size: either looser adds on house money, or fewer, because the move is felt to be spent. The direction is left open on purpose. | No difference between +1.5 R and +5 R. |
H4 is the anchor. If it holds, every difference found in H1 to H3 is a bias in one direction, and part A tells us which direction costs money. H5 is registered as exploratory: with three situations per state and participant it needs about 40 participants before a difference means anything, and it is reported as a pilot below that.
4. Part A: the market
Definition of a move and a decision point, identical to the one that selects the stimuli. Reference range is the median of the last twenty sessions' high-to-low range. Long case: the session low L, later a high H with H − L of at least 0.6 reference ranges and at least 30 minutes between them. The decision point is the first 5-minute close after H whose pullback depth (H − close) / (H − L) lies between 35 and 55 percent, at most 90 minutes after H and at least 60 minutes before the session close. Short case mirrored. One event per session and market.
Measurements. Entry at the decision close, stop at the move origin L and, as a second variant, at the 62 percent level; exit with the trailing stop used across this site (break-even at 0.5 R, trail 1.0 R, step 0.5) and, as a second variant, at the session close. Net of spread and slippage. Expectation in R and hit rate against the random-walk yardstick with the measured intraday drift, by pullback depth (35–45, 45–55), move age, time of day and market. Standard errors clustered by day. The time-of-day grid serves as the plausibility benchmark introduced in the volume-gate study.
Add versus new entry. On the same events, the position that entered at L with the same trailing stop, once with a unit added at the decision point and once without, against a fresh unit at the decision point. R per unit of new risk, paired.
Prior evidence that bears on the question. Late breakout setups run better after an early whipsaw (study), the second break of a range is negative everywhere (study), measured moves do not exist (study), and 50 percent retracements are geometry, not signals. None of those measured a pullback entry conditional on a completed move; that is the gap.
5. Part B: the perception experiment
Design. Within subject. Twelve real chart situations, drawn mechanically from the part A event pool with a fixed seed (four per market, six long and six short, at most two per calendar year), shown as 5-minute candles from the cash open to the decision point, with market and date hidden. Every participant sees every situation once, in random order, in one of four states assigned at random and balanced three, three, three, three:
- flat: "You are flat. You have not traded this move."
- in the move, +1.5 R: "You have been long since the start of the move. Your open profit is about +1.5 R."
- in the move, +5 R: "You have been long since the start of the move. Your open profit is about +5 R."
- stopped: "You were long from the start and were stopped out at break-even on this pullback."
Per situation: enter now, yes or no; if no, at which depth or not at all; size relative to normal risk; confidence 1 to 7; and whether the move feels "like mine", "like missed" or neutral. Time per situation is recorded. Before the charts, five belief items and a short profile (experience, discretionary or systematic).
Analysis. Primary outcome: entry rate at the decision point, flat against the two in-the-move states pooled and against the stopped state, as within-participant differences (each participant supplies all four states). Secondary contrast for H5: +5 R against +1.5 R. Secondary: accepted depth, size, the "missed" and "mine" ratings as a manipulation check. Significance convention |t| ≥ 2 with participants as clusters; n and standard errors reported for every cell. Then each answer is scored against the forward path of its situation, and the realised R by framing is compared: the price of the state in R per situation.
Sample. Twelve situations per participant, three per state. For a difference in entry rate of 15 percentage points between flat and the pooled in-the-move states to reach |t| ≥ 2 with the variance a pilot of one showed, about 30 participants are needed; the H5 contrast needs about 40. Fewer than 20 will be reported as a pilot, not a result.
6. Part C: reconciliation
Part A says what a late pullback entry was worth in the data. Part B says what the state did to the decision. The difference, in R per situation and per framing, is the cost or the value of having been in the move. If the reluctant version of me is right, the study will say so. If the version that missed the move and jumped in is right, it will say that.
7. What is published, and what this cannot show
Both parts are published in full when collection closes, including a null result. What the study cannot show: participants are visitors of a research site and readers of a trading forum, not a random sample of traders; twelve situations from one mechanical definition are a narrow slice of "moves"; a decision on a static chart lacks the pressure of an open position; and the "+1.5 R" and "+5 R" in the states are constants that real trades do not offer. The origin of the question is one trader's self-observation, which is the reason for the study, not evidence for its answer.
Take part: twelve situations, four states, ten minutes.
Disclaimer: Historical statistics are no guarantee of future market behaviour. This study is not investment advice. Trading involves risk of loss up to total loss.