Assume that a database has all of the Pokemon sale data, such as the timestamp, the card’s properties, and the price in US dollars. You can be sure that the card will always be in pristine condition. Every card may be offered for sale at various times and rates.
Given the card’s attribute, what kind of time-series statistical model would be suitable for estimating the worth of any particular card?
For my part, I would begin with linear regression and observe its performance. Since there are several variables you may use to help model the price with regression, I’m not sure time series makes as much sense in this case. Whether or not you believe the data should be handled as IID is the main determining factor.
Depending on how opaque the market is, I would anticipate that past card sales could have an impact on future prices.
It’s accurate. 
Doing an OLS is also my first pass idea. Perhaps the card should have any value, and the time series component only contributes a small portion of the explainable volatility.
Indeed, it might be somewhat beneficial to have an AR or MA component, but I doubt If you have adequate explanatory variables, even that section is superfluous.
Therefore, whether you believe there is a noticeable pattern component or seasonality swings involved is crucial.
It would be easy and effective to use the average of the previous ten sales or x sales.
Only qualities with a high volume of recent sales would benefit from it; extremely rare traits would not.
There are several instances covered by this one-minute solution. Really nothing to laugh at. In any case, it will be difficult to predict really unusual ones.