Price Study
When a price becomes
a judgment.
A trading experiment exploring what changes when continuous prices are translated into “bad,” “neutral,” and “good.”
A simpler signal.
A stronger suggestion?
A price is a number. Calling it “good” adds an interpretation. I wanted to examine whether that interpretation changes when someone trades, the price they accept, or their willingness to trade at all.
For a transaction-based platform, fewer trades could matter commercially. That is a motivation for the study—not an observed revenue effect.
Every evaluation is directional and relative to a defined reference. It is not a forecast of future returns.
Hold the market steady.
Change the representation.
Four displays separate the presence of a signal, its categorization, and the addition of evaluative language.
Price only
Market prices and quotes, without an additional rating.
Continuous deviation
A direction-adjusted numerical distance from the moving average.
Neutral categories
Five numbered price-position bands with neutral wording.
Evaluative categories
The same five bands, expressed from “very bad” to “very good.”
Interactive display preview. The same prices appear in every condition. Candles and quotes update each second; ratings update on a completed minute. Order entry is available in the research prototype.
Preview uses real semiconductor data. Neutral wording is condensed to numbered bands here; the research interface also shows the price-position description.
How the rating is calculated
The completed minute’s close is compared with the previous 20 completed closes. That difference is divided by a fixed residual standard deviation from the five preceding trading sessions. The sign is reversed for buying.
Direction-adjusted deviation → −1.5 / −0.5 / +0.5 / +1.5 σ → 5 bandsThe thresholds remain fixed across display conditions. The evolving candle does not change the rating until its minute closes.
A moving market.
A repeatable experience.
Live markets would expose people to different price paths. I built a historical replay so the same market sequence can be presented under each display condition.
Feedback raised two needs: personally relevant stocks and a chart that feels active. Participants can choose a sector before starting, then watch minute candles form from recorded trades.
Semiconductors
NVDA · INTC · MU
AVGO · MRVL · AMD
Consumer
F · TSLA · NIO
ABEV · AMZN · NKE
Six stocks per sector, ranked by prior 20-session average share volume within a disclosed candidate screen. Sector preference is recorded; comparisons must account for sector selection.
Only elapsed data appears. Best bid/ask is sampled once per second. The simulator does not reproduce full order-book depth, queue priority, or market impact.
Give traders an objective.
Leave room for judgment.
A pre-session briefing identifies the required buy and sell, target quantities, price limits, cash, and starting positions. The participant then decides when to act.
Required tasks make completion observable. Optional trades reveal whether a display encourages action or waiting when there is no obligation to trade.
Required trades
10 minutes
Required + optional trades
10 minutes
Required + optional trades
2 minutes
Prototype funding: $10M cash plus starting holdings. A shorter deadline changes the time available to trade; an AI pilot does not establish human cognitive pressure.
Observe the decision.
Not just the final trade.
Timing
Time to first order; waiting before submission; timing relative to grade changes.
Execution
Submitted and filled quantity, simulated execution price, and required-task completion.
Revision
Order edits, cancellations, and changes in limit price.
Discretion
Optional orders, traded notional, and decisions to remain inactive.
The replay and event logging are implemented. AI personas support technical pilot checks. Human participant evaluation is the next stage.
No participant findings or commercial effects are reported in this case study.