Definitions

Every number on this site, in plain language. All statistics use settled predictions only; unresolved settlements (halts, missing data) are excluded.

Quant

Short for quantitative analyst: someone who judges trading by measurement, not stories — sample size, verified records, recomputable edge. QuantRank500 borrows the discipline, not the gatekeeping: everyone is welcome, and every record is measured the same way.

Long (and short)

A long position profits when the price rises: buy first, sell later. Short is the reverse — selling borrowed shares to profit from a fall. QuantRank500 is long only: every prediction says “this stock goes up,” which keeps the rules simple and every record comparable.

Win Rate

The share of settled predictions that won. Breakevens count in the total but not as wins, so they lower win rate without being losses.

Profit Factor

Dollars won per dollar lost: total winning PnL ÷ total losing PnL. Above 1.0 is net-positive. A record with no losses shows “>X”, never infinity. The scoreboard ranks by a conservative lower bound of this number, so short lucky streaks can't outrank long consistent records.

Fill Rate

How often the market reached the predicted entry. Predictions are limit orders: if the session never trades at your entry, the prediction expires unfilled — recorded and counted here. A low fill rate means entries sit too far from the market.

Breakeven win rate

The win rate these win and loss sizes need just to break even: average loss ÷ (average win + average loss). The gap between this and the actual win rate is the edge. Big wins with small losses might need only 5%, so 6% is profitable; a high win rate below its own requirement still loses money.

Flat expiries

Predictions that ended their three-day window within ±1% of the fill price — neither win nor loss. Shown as “flat” on the site; recorded as “breakeven” in the ledger and CSV. The account still moves by the exit price.

Typical stop distance

How far below entry this predictor places stops, on average. Tight stops risk less but get hit more often; wide stops survive noise but lose more. Neither is correct — this is style, not skill.

Avg reward-to-risk

Planned target distance ÷ planned stop distance. At 3.0, the aim is to win $3 for every $1 risked — why a low win rate can still profit, and a high one can still lose.

Last 33 — win rate (“recent form”)

The same win rate over only the last 33 settled predictions. Shows current form; never affects ranking.

Eligible stocks

Any US stock at or above $3 and $1M daily dollar volume. The floors protect the record, not the predictor: very cheap, thinly traded stocks can be nudged to touch a price, and their prints vary more between venues — both make settlements less trustworthy. Both values are working defaults and may be tuned.

The $500 Account

Everyone starts at $500. Each fill commits one third of the balance at that moment; the balance compounds with every settlement, in order. Recomputed from the ledger every time — never stored, never editable.

Settled / Unfilled / Unresolved

Settled: the market decided — stop, target, or three-day expiry. Unfilled: the entry was never reached; recorded and counted in fill rate. Unresolved: the outcome could not be determined (halt, missing data); excluded from all statistics and visibly labeled.

Spread, commissions, and fill timing

Why a real trade would differ from a settled prediction. Spread: buying costs slightly more than selling pays; settlements use trade prints, which sit inside that gap. Commissions: brokers charge fees the simulation doesn't subtract. Fill timing: a real order stands in line — a touch of your price doesn't guarantee a fill, but the touch rule credits one. All three usually make real results slightly worse than the record, which is why the statistics measure prediction accuracy, not trading returns.

CSV export columns

The per-identity export (“Download record”) and the nightly full-ledger export share these columns. A mostly-blank row is not missing data — predictions export commitment-only until their 9:30 AM ET reveal. pnl is standardized at $500 notional, not the account impact.

Column	Meaning
id	The prediction's ledger number.
identity_id	The predictor's public UUID.
session_date	The trading session the prediction targeted.
status	Where the prediction is in its life: queued · working · active · unfilled · closed.
commitment	SHA-256 of the canonical payload + nonce, public since submission. Recompute it from the revealed fields to verify the prediction was locked before the open.
created_at	When the prediction was recorded (database clock, never the client's).
ticker	The predicted stock. Blank until reveal.
entry_price / stop_loss / target_price	The predicted plan. Blank until reveal.
commit_nonce	The random salt inside the commitment, published at reveal so anyone can re-hash. Blank until reveal — publishing it early would make the commitment guessable.
fill_price / fill_time	Where and when the limit order filled; blank if unfilled or still queued.
exit_price	The settlement price.
pnl	Standardized profit/loss: 500 × (exit − fill) ÷ fill. Recompute account balances by replaying settlements with ⅓-of-balance sizing.
result	How it ended: win · loss · breakeven · unresolved.
settled_by	What ended it: stop · target · expiry · halt · ambiguous.
settled_at	Market time of the settling print or bar.

A record of predictions, not investment advice. Statistics reflect prediction accuracy against exchange trade records — not actual trading returns.