Eight Routes to Buy, Nineteen to Avoid: How the 16 October Fare Window Splits a 59-Route Portfolio

Executive Summary: A Ratio Is Only as Honest as Its Denominator

On 16 October 2026 at 07:06 Beijing time, the OSINTRAI price-window engine rated 59 origin-destination pairs across a five-month forward matrix running from October 2026 to February 2027. [source-1] Eight routes returned a BUY signal, 19 returned AVOID, 26 returned WATCH and 6 returned WAIT. [source-1] The headline number in the sweep is JFK to Hong Kong at 292 US dollars against a computed mean of 679.75, a ratio to mean of 0.430 and a month-on-month change of minus 70.1 percent. [source-1] That figure is the single most quotable number in the day's data and it is the one most likely to be quoted wrong, because a ratio to mean is only as reliable as the mean it divides by, and this particular mean carries a coefficient of variation of 0.492. [source-1] A price that is 43 percent of an average whose own observations span 292 to 1045 dollars is not a 57 percent discount; it is an observation at the bottom of a wide distribution. [source-1] This report separates the eight BUY routes from the 19 AVOID routes, shows which averages are trustworthy enough to act on, and explains why the overlap between a cheap fare and a workable weather window is a smaller set than the fare data alone suggests. [source-1] [source-2] [source-3]

1. The Portfolio, Split Three Ways

The 59 routes divide cleanly into three behavioural groups. [source-1] Twenty-six sit at WATCH, which in this engine's vocabulary means the current price is close enough to the mean that neither booking nor waiting is compelled by the fare alone. [source-1] Nineteen sit at AVOID, meaning the current price sits materially above the mean and the trend slope is flat or rising. [source-1] Eight sit at BUY, and those eight are where the analytical interest lies because they are the routes where the fare data and the weather data agree. [source-1] [source-2] Overlaying the weather window classification changes the picture again: 14 routes read Book now, 7 read Cheap with weather so-so, 4 read Watch, and 34 read Wait for drop. [source-2] The overlap between BUY and Book now is the actionable core, and it is smaller than the BUY count alone suggests. [source-1] [source-2]

Sources: [source-1]

The five cheapest routes by ratio are PVG-TPE at 0.374, JFK-HKG at 0.430, JFK-ICN at 0.466, CTU-HKG at 0.532 and SIN-TPE at 0.535. [source-1] The five most expensive are CTU-ICN at 1.566, SZX-HKG at 1.684, LAX-KIX at 1.511, SIN-KIX at 1.444 and SZX-TPE at 1.418. [source-1] The spread between the cheapest and the dearest is 4.5 times on the ratio scale, which is the number that makes this portfolio interesting: the same engine, the same morning, and a 4.5 times spread on identical methodology. [source-1]

2. The Denominator Problem

The cheapest route in the portfolio, PVG to Taipei at 75 dollars against a mean of 200.80, has a coefficient of variation of 0.325. [source-1] The third cheapest, JFK to Seoul at 399 dollars against a mean of 856.20, has a variation of 0.427. [source-1] The second cheapest, JFK to Hong Kong at 292 dollars against a mean of 679.75, has a variation of 0.492, the highest among the eight BUY routes. [source-1] That ordering matters: the route with the most eye-catching headline number has the least reliable average behind it, and the route with the quietest headline has a materially tighter distribution. [source-1]

A coefficient of variation is the standard deviation expressed as a fraction of the mean, and it answers one question: how much does the average itself move? [source-4] At 0.492 on the JFK-Hong Kong pair, the historical observations around that mean are spread across roughly half the mean's own value, which means a 43 percent discount against it is a much weaker signal than the same discount against a tight distribution. [source-1] [source-4] The engine reports this field explicitly, and its own reasoning strings call out price volatility as a separate consideration from the ratio, which is the correct treatment. [source-1] The mistake is to read the ratio, quote the discount, and drop the variation. [source-1]

Sources: [source-1]

3. The Overlap Is The Actual Decision

A cheap fare on a route whose destination weather is deteriorating is not a bargain, it is a transfer risk with a lower ticket price. The engine computes this overlap explicitly and publishes it as a window level with a stated rule: green when the price ratio is at or below 0.95 and the weather score is at or above 65, yellow when the ratio is at or below 1.05 with weather below 65, red above 1.05. [source-2] On that rule, 14 routes come out Book now, 7 come out Cheap with weather so-so, 4 Watch and 34 Wait for drop. [source-2] Of the eight BUY routes, the one that does not carry a clean green window is JFK to Seoul, which the engine classifies as Cheap with weather so-so and pairs with a High delay reserve of 180 minutes rather than the standard 90. [source-1] [source-2]

That pairing is the single most useful observation in the day's portfolio. A 0.466 ratio on a route where the delay reserve is doubled and the weather window is explicitly qualified is a different proposition from a 0.374 ratio on a route with a 90-minute reserve and a green window, and the fare ranking alone inverts the two. [source-1] [source-2] Across all 59 routes the delay reserve is Low at 48 and High at 11, so roughly one route in five carries a doubled buffer. [source-2] The reserve is computed from the sum of the next three days of weather codes, precipitation and wind, and the engine publishes the rule rather than only the verdict. [source-2]

Sources: [source-1] [source-2]

4. The Expensive Side Is A Different Market

Nineteen routes at AVOID is not a market that has gone up. It is a market where the mean is below the current price and the slope is confirming the direction. [source-1] Shenzhen to Hong Kong at 636 dollars against a mean of 377.60 carries a ratio of 1.684, a month-on-month change of plus 71.0 percent, a variation of 0.411 and a slope of 0.6131. [source-1] Los Angeles to Kansai at 840 dollars against a mean of 556 carries a ratio of 1.511 and a slope of 0.5448, the steepest positive slope in the portfolio. [source-1] Chengdu to Seoul at 401 against a mean of 256 carries 1.566 with a High delay reserve of 180 minutes. [source-1] [source-2]

The pattern across the expensive set is that short intra-Asia pairs carry the worst ratios, while the long-haul pairs on the same corridors sit closer to the mean. [source-1] That is a structural feature of how fares are loaded rather than a transient spike, and it is the reason an AVOID verdict on a short pair tends to persist longer than the engine's month-on-month window suggests. [source-1] The slope field confirms it: the five most expensive routes all carry positive slopes between 0.165 and 0.6131, meaning the trend is still pointing up at the time of measurement rather than having already turned. [source-1] Two of the eighteen other flagged routes carry a High delay reserve, which means the expensive routes are not systematically the operationally difficult ones. [source-2]

5. Why This Differs From Yesterday's Fare Piece

The OSINTRAI archive carries two recent price pieces and a reader checking for duplication is entitled to a direct answer. [source-5] [source-6] The 15 October hub fare dispersion piece examined fare divergence between hubs on the same destination. [source-5] An earlier piece examined a Taipei and Seoul cross-hub fare collapse. [source-6] Both are narrower in scope than today's report on three axes. [source-5] [source-6]

The first axis is population: today's analysis covers the full 59-route portfolio with a complete three-way signal distribution and the weather-window overlay, where the earlier dispersion pieces compared a handful of pairs on a single destination. [source-1] [source-2] The second axis is the analytical contribution, which is the denominator reliability question in section two, a finding derived from comparing coefficients of variation across the BUY set rather than from comparing fares across hubs. [source-1] The third axis is the date basis, a 16 October snapshot with a documented engine stamp at 07:06:34 against the earlier cycles. [source-1] The overlap with the earlier archive is thematic rather than factual, and the conclusion is new. [source-5] [source-6]

Sources: [source-1] [source-5] [source-6]

6. How To Act On A Ratio

The first rule is to look at the variation before the ratio, because a wide distribution makes a striking ratio cheap to produce by chance. [source-1] [source-4] The second rule is to check whether the slope agrees with the ratio. All eight BUY routes carry negative slopes, which means the direction confirms the level rather than contradicting it. [source-1] The third rule is to require a green window rather than a merely cheap fare, because the window is the only field that accounts for whether the aircraft will depart on the day you need it. [source-2] The fourth rule is to treat AVOID as a wait instruction rather than a prohibition, since a positive slope with a High delay reserve is a signal to check again in a week rather than a signal that the corridor has closed. [source-1] [source-2]

For an operations desk, the portfolio view is more useful than the headline view, because a portfolio makes the hedge explicit: eight routes to buy today, nineteen to wait on, and twenty-six where the fare carries no information. [source-1] The 26 WATCH routes are not a failure of the engine; they are the honest answer for a route priced close to its own average, and treating them as either opportunity or warning would be a category error. [source-1] A portfolio where a third of routes are uninformative is a normal portfolio, and an operator who reports eight opportunities and nineteen warnings without noting the twenty-six neutral routes has presented a distorted picture. [source-1] [source-7] The last practical point concerns the fare condition rather than the fare level: a discounted fare on a route that later cancels still leaves the traveller with the cancellation problem, and the compensation path is separate from the booking decision. compensation claim A claim runs on the reason code rather than on the fare paid, which is worth knowing before the fare is bought. eligibility checker A travel policy policy is what covers the non-refundable cash outlay when a cancelled fare cannot be recovered from the carrier.

Sources: [source-1] [source-2] [source-7]