Cuts Like a Knife - Part 2
This blog entry is Part 2 of my look at card trimming. In Part 1 I investigated card manufacturing processes, explained standard deviations and normal distributions, and provided some data associated with the size of 2018-19 Prizm Basketball base cards. The following post contains a lot of statistical discussion, but leads to some insights about Prizm and plans for analysis of another popular set.
I estimate that the print run for 2018-19 Prizm Basketball base cards is somewhere on the order of 60,000 cards per player. That’s a lot of cards. About 20,000 of those cards come from hobby boxes and 40,000 from retail. There is nothing rare about them. That doesn’t mean the base cards aren’t valuable or won’t be valuable in the future. 2018-19 Prizm Luka Doncic base rookies graded PSA 10 are currently commanding over $100. That price is holding at that level, despite the most current PSA population report showing 340 of the 370 Luka Doncic Prizm base rookies have graded gem mint. That means 91.9% of these cards graded by PSA are gem mint or better. Beckett shows 329 out of 563 graded gem mint and 12 out of 563 graded pristine. The percentage for BGS is 60.6%. BGS 9.5 copies have also been selling for around $80. Ungraded Doncic base rookies can be had today for $20 or less. To me, the ratio of gem mint to total graded seem too high to warrant a $60 - $80 premium on a $20 card. That said, if you’re into grading and you believe the prices will hold, it’s like printing money. Go forth and grade.
The base cards aren’t the main attraction when it comes to Prizm. When you get down to it, everyone would prefer a serial numbered color Prizm. They command the highest prices. Who wouldn’t want a gold serial numbered to 10? So, I rounded up a sampling of serial numbered color Prizms and started measuring the height and width on those cards.
I can’t say that I had paid too much attention to the details of Prizm cards. I suspect most often the most desirable cards quickly go from pack to penny sleeve without being given a second thought. But, the exercise of measuring the cards gave me a reason to pause and really look at the cards. Upon closer examination, I noticed something substantially different between the serial numbered color Prizms and the base cards. The edges of the serial numbered color Prizm cards are different as compared to the base cards.
Go find a serial numbered 2018-19 Prizm basketball color Prizm. Hold it in your hand. Now feel the edges of the card on the front. Be careful, it’s sharp. Compare that to how the edge feels on any other non-serial numbered card from the same set. The serial numbered cards are different from the non-serial numbered base cards, silver Prizms, or even purple wave Prizms. That sharp edge is called a burr and the direction of the burr tells us something about how the cards are produced. The direction of this sharp edge leads me to believe that instead of cutting the cards with the blade entering from the front of the card, as we saw in the Panini production video in Part 1, the serial numbered cards are cut with the cutting implement first contacting the back of the card.
My hypothesis is that the serial numbering on the back of these cards changed the manufacturing process chosen by Panini. Is this an entirely different process or piece of equipment? I would guess that it is. A different manufacturing process also means there could be a difference in the mean and standard deviation of the height and width measurements. I measured 12 different serial numbered cards from the set. Here is a summary of the data.
How do these measurements compare to those of the base cards? We can employ more statistical methods to understand if there is a statistically significant difference between the cards measured. The first statistical test I’ll use is a F-test. The F-test will compare the variances observed for the serial numbered cards and the base cards. The variance is related to the standard deviation. Statistical tests start with what is known as a null hypothesis. A null hypothesis is simply a statement declaring what we are assuming to be true, unless the statistical test suggests otherwise. In this case, the null hypothesis is that the variance is the same for the base cards and the serial numbered cards. A null hypothesis should almost always assume there is no difference between whatever is being tested. When performing an F-test, two values of interest will be calculated. The first is what is called the F statistic (Fstat). The other is called F critical (Fcrit). When Fstat is greater than Fcrit, then we reject the null hypothesis. In this case, rejecting the null hypothesis would mean the variances are statistically different. If Fstat is less than Fcrit then statisticians would call it ‘failing to reject the null hypothesis’. For the purposes of this blog post, we would just say the variances are not different.
The second statistical test I will use is a two-sample t-test. The t-test is used much like the F-test but it compares the averages of measurements for each type of card in order to determine if there is a statistical difference between the averages. Much like the F-test, the t-test has a t statistic (tstat) and t critical (tcrit) value. The principles used in the F-test example apply to the t-test evaluation.
Let’s see what the statistics say.
The statistics confirm that the variances are different between the base cards and the serial numbered cards. There appears to be no difference in the average height or width. This data can therefore be interpreted as confirmation that the manufacturing process is in fact different and that the difference in variation observed is real. There is more variation in the size of serial numbered Prizms than there is for base cards. This analysis also confirms that we can’t use a ‘one-size fits all’ approach to evaluating card dimensions. The manufacturing process is important and if we want to have guidelines for the minimum size requirements for a given card, we should have data for cards from that particular set in order to define probability bounds.
All of this statistical analysis got me to thinking, should I extend this type of analysis to other sets? What about the most iconic basketball card set of all time, 1986-87 Fleer? It stands to reason that I could obtain some cards from this set that I feel confident are not altered and then take measurements. There are more 1986-87 Fleer basketball cards residing in PSA holders than any other basketball set produced, so obviously knowing that a card is unaltered is important to collectors. Perhaps this data would be of value to collectors that want to have confidence that their cards, whether raw or in PSA or BGS holders, are not trimmed. Honestly, I hope that trimmed cards never end up in graded card holders, as it would just further support the dark side of our hobby.
I’ll share details of my findings in Part 3 of this series.
Till next time - Jeff