Taken on the same day, the American market and the French market place their medians almost at the same level - and diverge completely as soon as we move up the distribution.On August 30, 2026, 148 American ads gave a median of $9.99 and 42 French ads gave a median of €8.00. But the ratio between the maximum and the median is 42 on one side and 7 on the other, and the certified segment weighs 15 announcements against 2. The two markets are similar in the ordinary and have nothing in common in the exceptional.
A precaution is necessary before any figures. These two statements are denominated in two different currencies, and I do not convert them: an exchange rate applied to asking prices would add an approximation where there was none. Everything that follows therefore comparesrapportscalculated within each market — a ratio has no unit, and is compared from one market to another without conversion.
This constraint is not a hindrance, it is what makes the comparison honest. Comparing $9.99 to €8.00 would require a rate, a date, and the hypothesis that the exchange explains something. Comparing a gear of 42 to a gear of 7 requires none of that.
Deux relevés, quatre mesures
The two queries were made on August 30, 2026, on the same term, in the same category, a few minutes apart. On the American side, 148 advertisements retained after withdrawal of the lots; French side, 42. All values cited are sums that sellersdemandaientat this moment; no sales are observed, and two photographs taken on the same day say nothing of any developments.
First measurement, the center: median at $9.99 on one side, €8.00 on the other. Second measurement, the low dispersion: first quartile at 5.06 and 4.52. These two levels look similar, in their respective currencies, and at first glance suggest two comparable markets.
Third measurement, the high dispersion: third quartile at 27.99 against 17.39. Compared to the median, this gives 2.80 on one side and 2.17 on the other. The gap opens. Fourth measurement, the peak: maximum at 419.98 against 55.00, or 42 times the median against 7 times. There, the difference is no longer a nuance.
The center looks the same
Median at 9.99 and 8.00 in their currencies, first quartile at 5.06 and 4.52. The regular issue is trading at a comparable level on both sides.
The top diverges by a factor of six
Maximum 42 times the median on the American side, 7 times on the French side. The difference survives resampling: it cannot be explained by the size of the surveys alone.
Certification is not at the same stage
15 certified ads on one side, 2 on the other. Out of 148 and 42 ads, that's 10% compared to 5%: half as frequent, on a much narrower sample.
Certification bonus varies
Median of certified compared to that of raw: around 20 times on the American side, around 5 times on the French side. The gesture does not have the same weight.
These last two measures deserve immediate reservation. Two certified ads are a pittance: the median of €38.70 that we get is the average of two opinions, not a market position. I cite it because the order of magnitude differs markedly on the American side, where fifteen listings give a median of $199.00, but I don't give it any more weight than that.
The most useful conclusion from this comparison is negative: a median noted in one market does not translate to the other, even when the two medians are similar. They look similar because they measure the same thing – the current issue – and not because the two markets behave the same.
Anything above the third quartile is a different logic on each side. A collector who values an ordinary copy can look at both statements; a collector who values an unusual piece must look at his own market, and nothing else.
Why sample size explains part of the gap
We must resist the temptation to explain the difference in peaks by a difference in taste or wealth between the two markets. A simpler explanation precedes all the others: 148 announcements against 42. The more observations a statement contains, the more likely it is to contain an unusual item, and the maximum is precisely the most sensitive indicator for this effect.
The reasoning is verified on indicators which do not depend on size. The first quartile and the median, which describe the center, are similar: they are robust. The third quartile, already more exposed, deviates moderately — 2.80 compared to 2.17 compared to the median. The maximum, entirely governed by the highest line, deviates massively. The hierarchy of deviations exactly follows the hierarchy of sensitivities.
This hypothesis is tested, and I tested it rather than assuming it. I randomly drew 42 lines from the 148 in the American survey — to bring it exactly to the size of the French survey — and I repeated the operation two hundred times. The result deserves to be given in full, because it corrects two preconceived ideas at once.
First correction, at my expense: the median is not stable at this number. Over the two hundred prints, it oscillates between 7.80 and 17.99, a difference of 131% between the extreme cases. The third quartile ranges from 13.53 to 50.00, a gap of 270%. The maximum goes from 39.95 to 419.98, a difference of 951%. The hierarchy of sensitivities is indeed what we expected – the center moves less than the extremes – but none of these indicators is really stable over 42 observations.
Second, more important correction: sample sizen'explique pasthe gap observed between the two markets. In 99% of the two hundred draws, the maximum of the American subsample remains greater than 55.00, the maximum value of the French survey. In other words, reducing the American market to the size of the French market is almost never enough to produce such a low peak. The difference between the two distribution tops resists the test, and must therefore be taken as real.
I had written the opposite before doing the calculation, assuming that a workforce effect would be enough to explain everything. This is not the case, and the lesson goes beyond this particular case: the “it’s a sample effect” argument. is convenient, often correct, and it is verified in a few lines of calculation rather than by decreeing it.
The practical corollary comes down to two points. If the numbers of two surveys are very different, do not conclude on any indicator without having verified what the size explains — not even on the median, which here moves by 131%. And when the test shows that the gap survives resampling, as it does for these two highs, then the difference is in the markets themselves and worth taking seriously.
Six transposable lessons
Estimate on your own market
For an unusual piece, a foreign survey is not a benchmark. The vertices obey distinct logics, even when the medians coincide.
Compare ratios, not amounts
A max/median or q75/median ratio has no units. It crosses currencies without conversion, therefore without the approximation that an exchange rate would introduce.
Note the number next to each value
148 and 42 announcements do not produce indicators of the same quality. A median without its number cannot be read again.
Beware of small segments
Two certified ads do not make a segment. Any median calculated on a single-digit headcount is an opinion, not a position.
Read the quartiles before the limits
It is between the third quartile and the maximum that the two markets separate. The terminals alone would have hidden it; the quartiles show this.
Do not confuse size and structure
A larger survey contains more extreme values, mechanically. Before invoking a market difference, check what size already explains.
What these two statements do not establish
They do not establish any value relationship between the two markets. Currencies are not converted, voluntarily, and none of the comparisons made here relate to price levels. Anyone who would draw a conclusion like “it’s cheaper in France” would make exactly the mistake this article seeks to avoid.
Resampling itself has its limitations. Drawing at random from an existing statement assumes that the lines are interchangeable, which is not exactly true: a seller who publishes twenty ads makes them correlated with each other. The test therefore says something solid about the variability due to the number alone, and nothing about the other sources of variation.
They also say nothing about the actual availability of copies. Forty-two French announcements mean that forty-two offers were visible that day on a given query: neither the existing stock, nor what circulates outside the market places, nor what is sold between collectors.
Finally, they do not allow any projection. Two simultaneous readings give a comparison, not a trend. It would take repeated observations over several weeks to say anything about a movement, and I don't have that.
Translate this into a file
The instruction:attach to any estimate the market on which it was taken, as well as the date.A figure written down without its market becomes unusable as soon as we no longer remember where it came from — and this moment arrives faster than we think.
For an ordinary copy, the two markets are the same as a benchmark, and the more extensive of the two will give a more stable figure. For an unusual piece – a variant, a certified example, a sought-after number – this equivalence disappears: only the market on which you would actually sell or buy is the reference.
Also note, for each estimate, the relationship between the maximum and the median of the reading. This single number summarizes the shape of the distribution better than a range, it does not depend on any currency, and it will remind you at a glance whether the reading was homogeneous or unbalanced. Here, 42 on one side and 7 on the other are enough to know which of the two lends itself to a simple reading.
One final note on certification. With two announcements on one side and fifteen on the other, the only thing to honestly record is the headcount, not the median. Write “certified segment: 2 announcements” is accurate and useful information; write “certified value: €38.70” gives two isolated opinions the appearance of a market price.
For an ordinary issue, the two statements of August 30, 2026 give comparable centers: medians of 9.99 and 8.00 in their currencies, first quartiles of 5.06 and 4.52. For an unusual coin, no: the maximum is 42 times the median on one side and 7 times on the other. The two markets separate above the third quartile.
Because an exchange rate applied to asking prices adds an approximation where none existed, and shifts the discussion towards exchange rather than structure. The internal ratios for each market — maximum to median, third quartile to median — have no unit and are compared directly, without additional hypothesis.
Yes, once the test is done. I drew 42 lines at random two hundred times from the 148 American ones: in 99% of cases, the maximum of the subsample remains higher than the French maximum of 55.00. Bringing the large reading down to the size of the small one is therefore almost never enough to produce such a low peak. The gap persists, and it must be taken as real.
Nothing more than the average of two opinions. The French certified segment of the statement includes two advertisements: drawing a reference value from them would give two isolated opinions the appearance of a market price. The only honest thing to record in this case is the workforce itself, not the value derived from it.
The ratio between the maximum and the median. It is a number, does not depend on any currency, and immediately tells whether the distribution was homogeneous or unbalanced. Here, 42 on one side and 7 on the other: we know without further calculation which of the two readings lent itself to a direct reading.
An estimate without its market is worthless
My Comics Collection keeps your observations with their date, their number and their origin, so that a figure remains judgeable long after.