What can we learn from L2 reviving legends data?
Introduction
I’ve always been intrigued by the introduction of reverse holo foil cards. Although this concept was introduced very early in the Western Pokemon world, with the infamous Legendary collection, and later on all e-series and ex-era sets, in Japan, the earliest set to introduce this concept is only L1 in the Legends era, towards the end of the year 2009.
The concept is rather simple, for (almost) every card of a given set, you can collect a rarer variant, which is shiny. Only one of those cards only is available per pack. This gives any common, uncommon, rare, and sometimes holo or even ultra rare card a rarer variant, and can make any card more exciting than its non-shiny counterpart. For us collectors, it’s an additional full set to complete, often more challenging than the non-reverse set, but one in which every single card feels prettier, making it very desirable.
But enough of an introduction, the one question I wanted to get you, that has been on my mind for a long time is: how are reverses distributed in a set? Each booster pack has several slots dedicated to commons, uncommons (unco), and usually one for the rare and/or holo/ultra rare card. This means by opening a booster pack (or box), you know you will have a fixed numbers of commons, uncommons, all the way to ultra rare (« quotas » or « hit rates » that exist in most any sets, and depend on language and sometimes time of printing). But the reverse slot introduces incertitude in those numbers, given the reverse can be of many rarities.
For instance, in the L2 set, the reverse slot can contain a common reverse, unco reverse, rare reverse, holo reverse, prime reverse, or even half of a legend card, which will be non reverse. But what dictates the odds that you will pull each of those rarities? Are you equally likely to pull a Rattata reverse and a Tyranitar prime reverse?
Is the reverse picked at random among all cards of the set (in the case of L2 one of 74 cards that can be reverse), without consideration of rarity?
To answer this question without inside information on how the packs and hit rates were designed, the best thing we can do is to look at a number of documented booster packs opening, if possible of unbiased origin (singles packs could be possibly weighed, mapped, and have consequently an altered distribution of reverses). To get unbiased data, I went to YouTube, which thankfully was already around at the time L2 was published. If the number of Japanese booster boxes was nothing compared to today’s sets, I could still find 7 full boxes being opened, constituting a data set of 140 booster packs. The 8th box was almost fully opened, but 2 packs were not. In total, this made up 158 datapoints to try and answer the question of the reverse distribution in the L2, Reviving Legend set. If this is nowhere near enough to be a representative sampling and to draw definitive and exact conclusions, it should still be enough to see trends and give a likeliness than one reverse holofoil distribution hypothesis is right or not.
Before I get into the numbers, I would like to formulate several plausible hypotheses about the reverse distribution, with a few facts about the L2 sets and hit rates that I’ve observed in all booster boxes being opened.
The L2 set is comprised of 81 single artworks, of which:
- 26 common
- 22 unco
- 12 rare
- 10 holos
- 6 half legend cards, making 3 sets of duo legends cards
- 4 prime
- 1 secret rare (lithograph)
Each booster pack contains 11 cards, with 4 common slots, 3 uncommon slots, 1 rare slot, 1 reverse slot, 1 holo slot, and 1 energy slot.
A booster box contains 20 packs.
The secret rare interestingly takes the last uncommon slot.
Primes are pulled on the holo slot.
Legend cards are always pulled in pair (contrary to our western booster packs), with one half being pulled on the reverse slot, and the other half (associated) on the holo slot.
All booster boxes I have seen being opened contained 2 pairs of legend cards (out of 3 in the set) and 3 prime cards (out of 4 in the set), meaning by opening a box you would only be missing a handful of cards. This leaves 18 (20-2 for the legend cards) spots for reverses, and 15 (20-3-2 for the prime and legend cards respectively) holo slots containing holo cards (as opposed to legend or prime cards).
Interestingly, there doesn’t seem to be a rule about holo distribution, ie although the set contains 12 different holo cards, most boxes don’t contain one of each + 3 duplicates holo as you could expect. Rather, one would get several duplicates or even triplicates holo cards, and consequently be missing a few holo cards. In statistic and probability terms, this mean the holo cards are determined following a « draw with replacement » (see Wikipedia urn problem), which simplifies a bit the situation. Indeed, pulling a specific holo card in one pack doesn’t influence the odds in the next pack, which wouldn’t be the case is there was a « quota » ensuring minimum duplicate cards.
The same fact could be observed for the reverse distribution, ie one isn’t guaranteed to get 18 different reverse cards in one box, in fact several of the boxes I’ve seen ripped contained one or more duplicate reverses.
An important assumption that I will make is that inside each rarity tier, all cards are equally likely (ie you have the same chances of pulling a Rattata reverse than an Eevee reverse since they are both a common card).
Finally and of little scientific interest although a fun fact: each time the secret rare (lithograph) was pulled, it announced an ultra rare (prime or legend pair) being present in the pack (remember, the secret is pulled from one of the uncommon slot - not from the holo slot, so there is still room for an ultra rare pull).
Reverse holofoil distribution hypotheses:
Hypothesis 1: All 74 possible reverse cards are equally likely to be drawn from a reverse slot.
This would be the simplest case: imagine a giant wheel being spun for each pack containing a reverse, with 74 sections of equal size representing each common, unco, rare, holo and prime reverse that can be pulled.
This system would make sense in terms of design, and also ensures that more premium pulls (holo or prime reverses) are less likely than the reverse being a « lesser » pull (say common or unco). Indeed, the set contains in total only 12 holos and 4 primes, against 26 commons and 22 unco, so a premium pull would be 3 times less likely than a lesser pull, purely by design of the number of each card in the set.
A first look at the data.
Let’s look at the actual data through the lens of this first hypothesis. I have put the grand total first, but will link all detailed pulls in a spoiler tab.
Grand Total:
Reverse holofoil have been pulled in 142 of the 158 packs being opened (16 of those packs contained a half legend card in the reverse slot).
In total, out of those 142 reverses were pulled:
52 common (or 36.6% of the total)
55 unco (or 38.7% of the total)
26 rare (or 18.3% of the total)
8 holo (or 5.6% of the total)
1 prime (or 0.7% of the total)
If the first hypothesis was true, we should have expected the following distribution:
26 common out of 74 possible reverse cards ie 35.1%
22 unco ie 29.7%
12 rare ie 16.2%
10 holo ie 13.5%
4 prime ie 5.4%
The model fits decently well with the distribution of common and uncommon reverse 37% vs 35% expected, 39% vs 30%). However, rare, holo, and particularly prime reverses would be severely underrepresented should the hypothesis be right (respectively 8% vs 16% expected, 6% vs 14% expected, less than 1% vs 5% expected).
The data being clearly skewed in one direction (« lesser » pulls being more common than « premium » pulls), it’s fair to say that the current hypothesis is wrong, and we need to look at a way to distribute the reverses more heavily towards the lesser pulls.
One way to do it is:
Hypothesis 2: The probability of each rarity being pulled is weighed by how common this pull is in a booster pack/box.
This one is less intuitive, but should skew the data in the direction we want. Basically, someone at TPC could have said « let’s make holo reverse as rare as a (non reverse) holo is in a pack ».
Another way to formulate is to consider that a pack contains 4 commons, 3 uncommons, but a single rare and a single holo card. The distribution could be engineered so that out of 9 reverses, 4 are commons, 3 are unco, 1 is rare, and 1 is holo or prime (with primes being 5 times as rare as holos considering a box contains 15 holo cards for 3 primes).
If this hypothesis is right, and given the different booster slots, we should expect the following distribution:
One booster box contains:
20x4 = 80 common cards
20x3 = 60 uncommon cards
20x1 = 20 rare cards
15x1 = 15 holo cards
3x1 = 3 prime cards
(and other cards that cannot be reverse, like the legends and lithographs, irrelevant for this point)
For a total of 178 cards (20 packs of 9 cards that can be reverse, minus the 2 legend cards).
This would give the following distribution:
Common: 44.9%
Uncommon: 33.7%
Rare: 11.6%
Holo: 8.4%
Prime: 1.7%
Looking back at the observed distribution, it is a slightly better fit, and we do see the desired trend towards premium pulls being significantly rarer. However, a few things don’t match, for instance, the observed distribution gives a very significant difference between pulling a rare and holo reverse (18.3% vs 5.6%, or 3 times rarer) - which is not the case here. Also, the common reverse would be, according to this second hypothesis, making up for almost 50% of the pulls, and being quite a bit more likely than uncommon reverse (which is not what I’ve seen in the sample of packs being examined, the opposite actually).
This leads me to believe the reverse holofoil distribution has been engineered specifically by TPC, to differentiate between 4 types of pull:
Hypothesis 3: This one doesn’t follow a distribution linked to the set design (number of commons uncommons etc in the set, or their relative abundance), but rather an « artificial » rarity that is designed in. In practice, TPC could have said « let’s print 10.000 sheets of commons and uncommons, 2.500 sheets of rares, 1.000 sheets of holo, and 150 sheets of prime reverses, and distribute them into the packs » for instance.
Common/uncommon: the data seems to show common rare and uncommon rare are roughly as likely as each other. Most packs (in the sample I’ve observed around 3 out of 4 packs) contained a lesser pull, with only one in 4 containing a rare or better reverse. In this 1 in 4 « hit pack », the distribution is quite heavily biased towards a rare reverse (3 out of 4 « hit packs »), where the top 25% (so 1 in 16 packs in total) will contain a holo or a prime reverses (let’s call it « super hit pack » for the sake of it). Funnily enough, it’s quite possible that just one in 4 of those « super hit packs » will contain the prime reverse (so 1 in 64 packs overall, roughly one every 3 booster boxes - it would fit the data quite well, especially if there had been a second prime reverse rather than just one in the sample I’ve observed).
So in conclusion, my best guess is the following: the set was designed by printing a given number of sheets/cards of each rarity, and by distributing them evenly.
I believe they printed 64 sheets:
48 of commons and uncommons reverses
12 of rare reverses
3 of holo reverses
1 of prime reverses
And distributed the lot evenly.
This would have the following distribution:
Commons and uncommons combined: 75%
Rare: 18.75%
Holo: 4.7%
Prime: 1.6%
This represents I think quite a good fit to the observed sample from YouTube which had:
Commons and uncommons combined: 75.3%
Rare: 18.3%
Holo: 5.6%
Prime: 0.7%
This is both satisfying the data measured, and it is also a simple enough system that can be designed, with 4 « tiers » of reverses, each being 4 times rarer than the previous tier (tier 4: common/uncommon reverse, tier 3: rare reverse, tier 2: holo reverse, tier 1: prime reverse).
Digression 0:
L2 and L3 are incredibly similar in terms of set: same theme, same number of cards, same structure and number of ultra rares, and in fact of each rarity, L3 has just one fewer uncommon and one more common.
I think everything applied here to L2 can be applied to L3. I’ll let someone watch all the openings of clash of the summit if they want to confirm!
Digression 1:
How bad is the sample distribution from the 8 videos I have based all this discussion around?
The short answer is: 158 is pretty small as far as sample size goes. However, it is the most reliable, verifiable and unbiased data I can get.
As for the long answer: there is one other way to look at a verified number, which is looking at the PSA population of the cards. It introduces a whole lot of possible bias, the main one being that people grade cards of Pokemon they treasure (in a few cases), or cards that can fetch a higher monetary value (in most cases). If 100 L2 boxes were opened today, I bet we would see far more prime, legends and holo being graded than common cards like our poor Rattata.
One way to erase the propensity of people to grade certain cards rather than other is to look at cards that will get graded whether reverse or not. Usually there are only a few of those, but luckily for us, L2 contains a lot of eeveelutions, importantly in several different rarities (Vaporeon Jolteon Flareon being Unco, while Espeon and Umbreon are Holo). There is also one Tyranitar, a very popular card.
Let’s look at the data for those cards, in first edition (important note: data from February 2025 pop reports):
Unco:
Flareon: 126 non reverse graded, 147 reverse graded
Vaporeon: 119 non reverse, 130 reverse
Jolteon: 101 non reverse, 130 reverse
Holo:
Espeon: 986 holo (non reverse), 83 reverse
Umbreon: 1485 holo (non reverse), 103 reverse
Note: 2 471 / 2 654 is around 93.1% of holos vs reverse holos.
In my data: 120 vs 8: 93.8% - excellent fit.
Prime:
Kingdra: 568 non reverse, 94 reverse
Lanturn: 528 non reverse, 77 reverse
Tyranitar: 868 non reverse, 142 reverse
Steelix: 530 non reverse, 89 reverse
Note: 2494/2896 is around 86.1% of prime vs reverse prime.
In my data: 24 vs 1: 96%, bad fit. I believe for 2 reasons: people have a bias towards grading prime reverses when they get their hands on one, making them appear proportionally more in the pop report, and also my sample has an unusually low number of prime reverses (only 1 seen, I believe we could have expected between 2 and 3 on average).
The most interesting piece of data here is in the holo repartition: for each Umbreon reverse graded, there were 14.4 Umbreon holo (non reverse) graded.
The ratio is reasonably similar for Espeon, with 11.9 non reverse graded for each reverse. Let’s take 13 as an average.
If we take for assumption that people would grade a Espeon or Umbreon holo or their reverse counterpart with equal chance (which is a big one, but it’s the best approximation we can make), then we can deduce the following: you have roughly 13 times more simple holo (non reverse) than you have holo reverse. Given that we know you get exactly 15 holos in a box, it would give a number of 1.15 holos reverse on average per box, or (for the 18 packs containing a reverse rather than a legend), a 6.4% chance that a pack contains a holo reverse.
In the data from the YouTube openings, we have observed a 5.6% rate of holo, I would say that’s very closely matching the number we got from the PSA population.
Back to the question about relative rarity of uncommon reverse vs holo reverse:
For uncommons, unfortunately I believe too few people took the decision to grade a non reverse uncommon card, even a gorgeous Vaporeon for instance, so I wouldn’t try and observe something meaningful in the same way. It is however clear than more of those uncommon reverse were graded (407 for those 3 eeveelutions out of 22 uncommons, giving a potential number of 2985 unco reverse that could have been graded if all uncommons were as popular as the eeveelutions), compared to just 186 Umbreon and Espeon reverse. Those 2 holos were themselves 2 out of 12 holos, so an expected number of 1116 holos reverses if all holos were as popular as Espeon and Umbreon. Overall this gives holo reverses around 2.7x rarer than uncommon reverses: this is clearly a large deviation from what I observed in the samples from YouTube (5.6% vs 38.7%, or 6.9x rarer). I believe this observed deviation is due to the fact Espeon and Umbreon were heavily graded in reverse, even in comparaison with Flareon, Jolteon and Vaporeon.
Digression 2:
The rarity paradox, or why can a holo be almost as rare as a prime.
When evaluating hypothesis 2, we established that one box contained:
80 common cards
60 uncommon cards
20 rare cards
15 holo cards
3 prime cards (holo slot)
4 legends cards (2 pairs, on the reverse and holos slots)
0.5 lithograph (one every two boxes, on an uncommon slot, meaning the “60 uncommon cards” above should technically read “59.5 uncommon cards”).
18 reverses
20 energy cards
But of course each of those rarities has a different set of possibilities: 10 different holos exist, but only 4 different prime cards for instance.
This means that the probability of getting at least one SPECIFIC holo (say Umbreon for the sake of it) in a box is 1-(9/10)^15 = 0.794or 79%. This is the probability of at least one Umbreon, of course the average number of Umbreons holos per box is 15/10 or 1.5.
Meanwhile, getting one SPECIFIC prime (say Tyranitar for instance) is 3/4 or 75% (because no duplicate primes are allowed in a box, as opposed to holos which are truely random). The average number of Tyranitar per box is of course 0.75.
So although primes are taken together 5 times rarer than holos, you have only about twice as many umbreons (non reverse) than Tyranitar in circulation (and this reflects in the PSA population report, 1485 Umbreon holo vs 868 Tyranitar prime).
Adjusting for their rarity, there is, on average, per box:
80/26 = 3.0769 SPECIFIC common
59.5/22 = 2.705 SPECIFIC uncommon
20/12 = 1.667 SPECIFIC rare
15/10 = 1.5 SPECIFIC holo
3/4 = 0.75 SPECIFIC prime
4/6 = 0.667 SPECIFIC legend card
0.5/1 = 0.5 SPECIFIC lithograph
20/8 = 2.5 SPECIFIC energy cards
This scale tames down the scarcity of the rarer rarities.
Also of note, by opening a box, you were in a pretty good position for a full set, missing only 3 to 4 ultra rares, likely a holo or rare card or 2 (because of duplicates) and a good shot at the full common/uncommon set. Of course because of the reverses, you would still be miles off a master set though, especially given the rarity distribution that was the original point of this post.
Disgression 3:
How rare are the unlimited prints of L2? L3? Are they the same?
If I look at the PSA data, and had to give a number on how much rarer is unlimited compared to 1st Ed, looking at ultra popular Pokémon to try and avoid any grading bias (ie people might be tempted to grade a rattata unlimited reverse because it is unlimited reverse, where the 1st Ed reverse is much less graded).
Data from July 2026 pop reports:
Espeon holo 1st vs 1st reverse vs unl vs unl reverse: 1506 - 83 - 124 - 7
Umbreon holo 1st vs 1st reverse vs unl vs unl reverse: 2464 - 197 - 189 - 12
Tyranitar prime 1st vs 1st reverse vs unl vs unl reverse: 1199 - 81 - 193 - 17
Celebi prime 1st vs 1st reverse vs unl vs unl reverse: 864 - 36 - 129 - 9
Celebi holo 1st vs 1st reverse vs unl vs unl reverse: 418 - 27 - 41 - 2
To the question: how much rarer is L2 1st Ed vs L2 unl, based on Espeon + Umbreon:
95.6% of 1st Ed overall - which would mean unl is about 20-25 times rarer than 1st Ed - I find that believable
To the question: how much rarer is L2 1st Ed vs L2 unl, based on Tyranitar:
93.4% of 1st Ed overall - small difference with the holos. Could it be due to PSA mislabeling unlimited? The 1st Ed sign is tiny, and on those dark type cards, very very easy to miss.
To the question: how much rarer is L3 1st Ed vs L3 unl, based on Celebi Holo:
94.1% of 1st Ed overall - seems consistent
To the question: how much rarer is L3 1st Ed vs L3 unl, based on Celebi prime:
95.7% of 1st Ed overall - sensibly the same ratio as L2
To the question is the ratio of unlimited the same for L2 and L3: I would say yes, see previous point. The only thing that makes me doubt it is looking at Tyranitar prime (L2) vs Celebi prime (L3), the difference of proportion of unl is respectively 6.6% vs 4.3% - sounds just enough to be real. But at the same time it’s the opposite if you compare the holos.
To the question: how much rarer is L3 vs L2:
Total graded cards: L2 (24 502 cards) / L3 (12 846 cards) would suggest about twice as rare - I found that hard to believe. I think there is a bias where L2 contains many more popular cards (eeveelutions, Tyranitar and Steelix in the primes being arguably more fan-favorites than Celebi and Machamp), meaning L2 was more graded.
Also, coming out just 5 months apart, I would expect similar run numbers in order of magnitude.
Assuming Tyranitar and Celebi are about as popular (I guess not quite, but that’s probably an OK ish assumption to make), and looking at regular 1st Ed prime graded, we would have about 39% surplus of L2 compared to L3 (1199 vs 864).
My guess is that Tyranitar is a bit more popular than Celebi so I would tame down this 39% to around 25-30%.
I would say L2 was printed to about 1.25-1.30 times more than L3 (purely an educated guess, nothing more).
Conclusion:
tldr: the data point towards a 4-tier system for the distribution of reverses (tier 4: common/uncommon reverse, tier 3: rare reverse, tier 2: holo reverse, tier 1: prime reverse), with each tier before 4 times rarer than the tier below.
In practice:
1 prime reverse every 64 packs
3 holo reverse every 64 packs
12 rare reverse every 64 packs
48 common or uncommon reverse every 64 packs
Final point: prime reverses are excessively rare (roughly one every 3 booster boxes, kinda similar to gold stars in western ex era booster boxes), holo reverses are pretty much one per box, the rest is a bit easier to pull.
finally thanks for reading, and I would be happy to take any comment or question on the process leading to these conclusions! 
