I covered this in another post. I can restate here. It's really simple actually. Beckett has a standard method of calculating Beckett values and also forces us to use common sense once we see what the process is.
First they collect a random sample from online and in person sales. They employ buyers to hit card shows to purchase lower cards and survey buyers at shows to see how much their purchases totaled. This is old school tactics that contribute slightly to the overall pricing estimates. In the last decade, e-Bay (et. al) has provided a lot of insight as well. Furthermore, in conjunction with the Beckett Marketplace, they log the sales made through their website. Hence they try to census e-Bay (et. al) and online sales but only random sample the shows and dealers. BE CAREFUL: Shipping is included in online sales. This is part of the basis for the "over 50 dollars will include free shipping." on Beckett marketplace.
After a suitable number has been sampled (typically 20-50 sales) the robust measures are used to cite the high and low numbers. Typically the 90th percentile marks the high value and the 10th percentile marks the low value. This is to minimize outliers and reduce the impact of shipping costs.
Now as cards get older, interest in the cards subside. Therefore it becomes very difficult for the minimum number of cards to be sold for a value to change. This is why we rarely see older cards ever fluctuate.
In more technical terms, if there are N total cards made, they require the sample to be representative say with size n to be at least 1% (I don't personally know this representative number) of the cards made.
From there, cards are typically binned into specific values and processed into their cataloguing system.
Hope this helps.