"Yes Virginia, There Are Match Pairs", though it can be hard to see when you're starting out.
Data:
Start with a homogeneous data set. Look at many many sales or within the same subdivision, or the same complex (including rent rolls). The more data the better. Comb though the MLS (go back in time). Look up all of the past and recent transactions listed by the Assessor. Plot and label this data out on a map. In table or on the map, list the KEY economic characteristics like square footage, beds + baths, views, YOC, # of garage spaces, basement, no basement, finished basement, condition, net sale price, $/sf, and date of sale, etc.
The Analysis:
The marketplace isn't random. Just like in nature, the marketplace creates patterns based on symmetry, path of least resistance, and disturbances.
As you go through this you may start to notice ones that are very similiar but only slight different. Break apart the difference based on 1) total dollar difference, 2) $/sf difference, and 3) percentage difference.
Sometimes two sales are fairly similar but have two variables that are different. You can adjust one of the variables. For example you can adjust up/down for market conditions or you can add/subtract for the garage. You now have synthetically created a reasonably similar match-pair.
Sub-Adjustments to Adjustments:
Remember that the adjustments that you extract are affected by the other appraisal models, e.g., physical depreciation, functional depreciation, external obsolescence, cost-push inflation, fluctuating gross rent multipliers, vacancy rates, etc, etc. The can provide a basis of sub-adjustments. Terrible markets have external obsolescence. In office buildings/condos, I was routinely using a tenant improvement finish cost to make adjustments and then knocking off 25% external obsolescence found by extraction to that cost adjustment. It appears to reconcile near match-pairs appropriately.
As another example, a newly constructed residential condominium buildings in my area may command $20,000/floor level difference and about $30,000 for a garage space; these depreciate and it should reasonably reconcile with, in my area, with an older condo building's $2,000/floor level and $10,000 garage space.
More Analysis:
Familiarize yourself with cost books. Does that $30,000 garage space extracted from comps make sense when replicated in a mini-cost approach? Does the rate of depreciation between the $30,000 and $20,000 garage space make sense? Does the value of the extracted $10,000 garage space make sense? These then become additional tools to make adjustments and double-check the reasonablity of your adjustments.
Familiarize yourself with vacant land values in your many neighborhoods, districts, and sub-markets. You can "pick" up that garage and drop it in a higher or lower land value neighborhood.
Bit More Analysis:
Next start looking for other patterns. MS Excel's graphing is very convenient and once graphed you can easily drag the columns to display the independent variable (X-axis) that may be explaining dependent variable (Y-axis). Appraisal is preoccupied with making the dependent variable Value or Value per square foot. It doesn't always have to be on the value variable, though. Because we started out with a fairly homogeneous data set, you can do a simple statistically valid linear regression analysis.
Tad Advanced Analysis:
All of the appraisal models allow for cross extraction. A rent comp can be compared to a similar sale comp to determine a cap rate or GRM. If an old garage from a third comp rents for $75/month, the capitalized rent of $900/year x 9 GRM, equals a garage's value of $8100. It makes sense. That reconciles in alignment with the $10,000 as found by the sales comparison extraction in the mature central Denver market.
Reasonability and Reconciliation:
File your research for future reference. But bear in mind that it is a snapshot in a point in time. 2009's numbers may not relate in a 2013 market. The pre-established adjustment lists lose track of reality because they were extracted too long ago, and it'd be wrong to apply, for example, a $30,000 garage adjustment to an old building, and vice versa. Likewise, a middling $20,000 garage adjustment is just as incorrect when applied to both the older building or to the newer building.
Have fun!
Find a bunch of substantially similar sales in a tract development, say 10+ & plug in their sales prices into a spread sheet. Now, pick one as the subject & pretend you are appraising it. Apply SF adjustments at, say, $5 increments to the other sales. Start low.