Grading stats and crossover analysis

gemtone65

I couldn't locate the original thread from DH, but my recollection is he based his support for the low PCGS crossover rate in part on data purporting to show that PCGS grades more strictly than NGC. I also seem to recall that the statistics he used were examples showing the percent of all MS graded coins of a certain type and date graded at selected levels by PCGS and NGC. Unfortunately, such data may show more about the quality of coins submitted than the rigor of the grading standards.

In particular, these data are skewed by PCGS having more submissions of lower quality coins. When one adjusts for this disparity, e.g., considers only coins graded MS64 and better, a different picture emerges.

My data is admittedly old, from the mid-1990's, but it contains information on the first 100,000 1881-s Morgans graded by PCGS, and the first 50,000 such pieces by NGC. At grades MS64 and above, there was no statistically significant difference in the distribution of grades, given that the coin grades MS64 or higher.

That being the case, and if this finding is generally representative, it should follow that if the PCGS grading is unbiased with respect to crossovers, the failure rate of crossovers should be comparable to the likelihood of taking a previously graded PCGS 1881-s dollar graded MS65, resubmitting it, and having it come back MS64 or worse. Clearly this is not the case, so I am prepared to conclude that PCGS is not unbiased about its grading standards on crossovers.

coinguy1

gemtone, you said, in part "I couldn't locate the original thread from DH, but my recollection is he based his support for the low PCGS crossover rate in part on data purporting to show that PCGS grades more strictly than NGC. I also seem to recall that the statistics he used were examples showing the percent of all MS graded coins of a certain type and date graded at selected levels by PCGS and NGC. Unfortunately, such data may show more about the quality of coins submitted than the rigor of the grading standards"

I believe that the last sentence in your above paragraph is key, so I will repeat it (again) -

"Unfortunately, such data may show more about the quality of coins submitted than the rigor of the grading standards".

That points out but one major flaw in using statistics in this case. Unless the identical coins are used to submit to both PCGS and NGC, the grade distribution included in the reported statistics is virtually meaningless. And, obviously, the submissions of all of the coins being discussed, were nowhere near identical.

For example, if NGC has certified a higher percentage of PR65 and higher Seated Liberty Dollars than PCGS has, based on the number of all proof Seated Liberty Dollars submitted to each service, it probably means one of two things : Either 1) identical quality coins were submitted to both grading companies and PCGS was stricter or 2) the average quality of those submitted to NGC was higher and they graded them accordingly. The second scenario is as likely, if not more likely, than the first.

I will repeat (and then shut up for a while) - unless identical coins were submitted to both NGC and PCGS, to arrive at the reported statistics (and everyone knows the submissions were far from identical), the statistics are insignificant at best and more likely, nearly meaningless.

roadrunner

Good points Coinguy. It's also true that the data from the 1990 days is totally useless today. We already know that the standards employed by both major services (and probably ANACS too) back then WERE very comparable over the US coin spectrum. I can also state for a fact that PCGS was far tougher on smaller coins like 3c silvers and half dimes than PCGS was. I had a lot of fun finding PQ PCGS pieces and getting them upgraded at NGC. It was also true that PCGS tended to be tougher on Indians and large coppers.

Today, many things have changed. The standards have shifted all over the place. You can't lump all the old date in with the new and expect to get anything useful.

roadrunner

roadrunner

Thanks to Coin Guy for catching the error in my last post. PCGS graded tougher than "NGC" on those coins. Actually, I knew of several series that NGC was easier to get the next grade on, but NO series where PCGS was easier as a rule. I think even back in 1990 PCGS coins were slightly favored but not by much.

While it is normally true that the coin sells itself and not the holder, there are many lazy (and opportunisitic) dealers and collectors out there who will offer much less for a coin just based on it not being in a PCGS holder. It doesn't matter how nice the coin is either. They'll try to rip it as just an average coin. The auction floor appears to be the only place where holder bias is negligible. A PQ coin will bring a PQ price regardless of the holder.

roadrunner

gemtone65

Coinguy 1: Obviously, my thread takes the position that the data as used by DH was indeed almost meaningless, and certainly misleading. However, with the adjustments I made by considering only coins graded MS64 and higher, I don't think it was a statistical quirk that the distribution of grades by level across grading service was almost identical. Here are the results for 114,000 coins graded between MS64 and MS66:

Grade PCGS NGC

64 55.8% 58.2%

65 36.2% 34.8%

66 8.0% 7.0%
------------- ----------
100.0% 100.0%

The average grade for these pieces was almost exactly the same --MS64.5.

BowAxe

<< <i>That being the case, and if this finding is generally representative, it should follow that if the PCGS grading is unbiased with respect to crossovers, the failure rate of crossovers should be comparable to the likelihood of taking a previously graded PCGS 1881-s dollar graded MS65, resubmitting it, and having it come back MS64 or worse. Clearly this is not the case, so I am prepared to conclude that PCGS is not unbiased about its grading standards on crossovers. >>



Gemtone, I am skeptical of your conclusion concerning crossover bias, which appears to be based upon the ASSUMPTION that if a statistically significant sample of PCGS-slabbed MS65's was cracked out and resubmitted to PCGS, the percentage that would come back graded MS64 or lower would be dissimilar to that of MS65 NGC crossovers that would come back as MS64 or lower. You say, "Clearly this is not the case, . . ." but how do you know? Has anyone actually tested this proposition by cracking out and resubmitting a quantity of PCGS MS65 1881-S dollars (or any other coins) and comparing it to a similar number of NGS MS65 crossovers? If so, show us the data.

Dell

IrishMike

BowAxe that is the point that no one has tested this. What he said and it is logical, is that based on submissions to both companies, on average the grades were the same. The ony way you could argue this is to postulate that out of 150,000 submissions to both companies NGC got a statiscally more sigificant batch of lower qualities submissions. Give the numbers quoted that would be highly unlikely. Anyway its an interesting hypothosis.

gemtone65

Bow Axe: The failure rate of crossovers is currently 74%. Could anyone seriously believe that if you cracked out a random sample of PCGS 65's and resubmitted them to PCGS, 74% or more would downgrade. I don't think so. For if this were true, 1) no one would ever break out a coin and resubmit, and 2) PCGS grading would be a complete sham, and would have been exposed as such long ago. Neither scenario is plausible.

Thus, while we don't have the precise empirical data that you would like to see, we do have other information upon which to make a very sound conclusion.

BowAxe

There is another way to test your hypothesis. So that the results mirror PCGS's current grading standards, take two groups of NCG MS64 coins, say 20-30 coins in each group. Submit one group to PCGS for crossover and crack out the other group to be submitted raw. A comparison of the results should unmask any bias in the crossover group.

I should think that there are crackout artists amongst us who are meticulous enough in their record-keeping to have accumulated meaningful data on crackouts of NGC coins to PCGS. If so, it would be interesting to see such data.

Meanwhile, I will reiterate: Conclusions grounded on assumptions, regardless of the apparent merit of those assumptions, hold very little water.

EVillageProwler

I think we should have 3 groups, with the third group being cracked out PCGS-slabbed coins.

If both groups of NGC-slabbed coins did poorly, but the PCGS group did well, then that's more evidence.

EVP

sinin1

3 groups would be interesting - but we would need to do many series wouldn't we? Are PCGS standards vs NGC standards different for Morgans/ModernCommemoratives/Just about any series?

Taking population reports to prove anything is misleading and vague at best - especially when their accuracy is in question with resubmitted crackouts and coins/slabs out there of coins with population ZERO!!!

It comes back to opinions and inferences with no one willing to spend the thousands in grading fees to scientifically prove anything.

gemtone65

Bow axe; My conclusions are based on logic, not assumptions. Sound conclusions can be made without empirical data. See Einstein; see Keynes.

But, since you want data, try this. I submitted one NGC coin ( a full gem monster toned 1882 MS66 liberty seated 1882 half) to PCGS for crossover; it failed. In contrast, I have submitted 100 PCGS coins for regrade; none were downgraded.

So, if PCGS treated crossovers the same as regrading their own coins, what would be the likelihood that they would not downgrade 100 consecutive coins that I submitted? The answer is, the likelihood is the same as winning 4 consecutive lotteries at odds 20 million to one; or, equal to placing a winning single number winning roulette bet 38 times in a row.