What does this
entry actually cost?

What it measures

Three things that are already true at the moment you look: how far this entry sits from the stock’s own recent prices, how far it would have to fall to reach a band where it actually traded, and what the round trip costs. All of it is arithmetic on two years of daily opens, highs, lows, closes and volume.

The ten measures

Stretch

How far above its own moving average it trades, in units of its own daily range — then read against its own two-year distribution of that same quantity. Eight per cent above the 50-day is nothing on a stock that moves six per cent a day and a great deal on one that moves one.

Position

Where in its own range this sits, and what share of the window’s volume changed hands below the price you would pay. The second half is the point: it counts how many holders are already in profit above their cost.

Heat

What the last few sessions added. The overbought family — RSI, stochastic, Williams %R, CCI, money flow — is five ways of asking one question, so they are each read against their own history and reduced to a single median rather than counted five times.

Room

The last N sessions binned by price into a volume shelf; the distance down to the nearest band that holds a real share of that volume, in per cent. A level where the stock genuinely changed hands, not one where it merely turned round once.

Volatility

Realised volatility from log returns, annualised, read against its own two-year range — and scaled to the window as a one-sigma move. Quoted as a range, because the estimate has a standard error of its own.

Friction

The only cost here that is certain. Dollar volume and how coarse one tick is against the price. The quoted spread is not in this feed, so it is not estimated — inferring it from the intraday range would be measuring volatility and calling it cost.

Locked out

The share of this stock’s movement that happens between the close and the next open, while the book you would trade in is shut — and, when it gaps up, how often the session hands it back before the bell. A name that makes most of its price overnight is one you cannot time an entry into and cannot exit when it moves. The two readings combine by taking the higher, never by adding: they correlate about 0.5, and adding them would let one fact vote twice.

Effort

How many shares it has taken lately to move this price up one per cent, against how many to move it down one. A high reading means every rise is being bought through supply while the falls need almost no selling — the expensive side to be entering on. The window’s own drift is removed before the comparison, because without that step the statistic simply reads the trend back to you as depth. This is a reading about the state of the tape over recent months and not a property of the company: it correlates only 0.13 with itself across separate windows. It is also not a liquidity measure — this feed carries no quote, no spread and no depth.

Passenger

The share of this stock’s daily movement that is simply its own sector index moving, fitted against the matching sector fund. A high reading says most of what you are buying is a basket that trades in its own wrapper, usually more cheaply. It drops, by name, for anything with no sector on the register — there is deliberately no fallback to the broad market, because that fallback caught a third of the names in testing and produced every wrong answer in the set.

Valuation

Price against what the company actually filed — earnings, sales and book value, from the SEC’s own XBRL company-facts API, which is the company’s figure and not an estimate. It is still a position measure like everything else here: where today’s multiple sits in this company’s own two-year range of that same multiple, taken as the median of the three so one expensive ratio cannot carry it. It is not a judgement about whether the business is worth the price. The trailing-twelve figure is rebuilt from the filings the way the company reports them, including deriving the fourth quarter from the annual less the nine-month year-to-date, because a fourth quarter is never filed on its own. It drops, by name, for anything that does not file with the SEC.

The second pass, and why it is not an average

Everything above is positional: it measures this entry against this stock’s own recent self, and it treats being extended as a cost. That is an assumption about how markets work, and on a tape that trends it is the wrong one. A page that scores only that way and calls it “the reading” is asserting a market model without ever testing it.

So the page reads the same entry a second time, over ten years rather than two, asking four questions the first pass cannot:

Persistence

A variance ratio over this stock’s own returns, at four horizons, with the heteroskedasticity-robust significance test. It asks whether multi-day moves in this series extend or come back — which is to say it tests the first pass’s central premise directly, on this name. Reported with its z, and a tilt that does not clear two sigma is called a lean rather than a finding.

Precedent

No model at all. Every past session where this stock stood the same distance from its own 200-day mean as it does today — measured in units of a month of its own volatility, so it is a stationary quantity and a stock that has trebled is not permanently “extended” — and what the next N sessions actually did from there. The sample counted is episodes, not observations: forty consecutive qualifying sessions are one event, not forty, and below four episodes it returns no reading.

Short-sale flow

FINRA’s daily consolidated-tape file: the share of each day’s printed volume that was sold short. This is short volume, not short interest — a flow, not a position — and much of it is market-maker hedging rather than anyone betting against the company. So it is never read as a level: every day is compared with the same day’s share across all twelve thousand symbols, and what is scored is the gap and how it has moved. Not available for listings off the US tape.

Filing calendar

From SEC submissions: how long since the last 10-Q or 10-K, and therefore roughly how far the next one is on the ordinary ninety-one-day cycle — an estimate from cadence, never an announced date, because submissions carries none. Plus how busy the 8-K tape has been against this filer’s own two-year rate. Buying eight days before a print is a different act from buying eight days after one, and no price series knows the difference.

The two numbers are never averaged. Averaging them would destroy the only thing they have to say. Two readings built on different questions that land within a point of each other is the strongest result this page can produce; two that disagree by three is a decision about which model you believe, and the page’s job is to show that it is a decision rather than to split the difference and hide it.

Where it refuses

A window with too little history returns no score rather than a number with a caveat. Seven days needs 120 sessions, thirty needs 160, ninety needs 260 — below that, the measures have no distribution to be read against.

A window also returns no score when too little of it answered. The test is coverage, not a count: if fewer than 70 of the 100 points of weight could be computed, there is no number. This replaced a rule that counted dropped measures, which stopped working the moment there were nine of them — an exchange-traded fund has no sector to be a passenger of, so a plain index lookup would have been refused for missing one input out of nine. Under the coverage rule a fund leaves 90 to 97 points and scores correctly; a listing with no volume at all leaves 60 to 69 and is refused, also correctly.

A plausible-looking number standing in for absent data is the one thing this page is built not to do.

What it is not

• Not a probability. A reading of 80 does not mean an eighty per cent chance of losing money. It means every measure here sits at the expensive end of its own range.
• Not a forecast. Nothing here states or implies what the price does next.
• Not a valuation. It does not know what the business is worth. It has no earnings.
• Not a judgement of the company. An excellent business can read 90. The reading is about the price and the window.
• Not fitted. No number in the weight table was chosen by testing which value would have made money. The weights are a stated opinion about how long each measure takes to matter, and they are printed in the tally so you can disagree with them.
• Not advice. This page does not know your position size, your cost basis, your other holdings or your actual fill, and it never tells you what to do.

One technical note worth keeping honest

Stretch, heat and volatility are computed on split- and dividend-adjusted closes, which is a total-return series — while what you actually pay is a price. On most names the difference is immaterial; on a high-yielding one it is real. The price shown at the top of the reading is the raw last print, not the adjusted one.