As has been discussed, in:
https://collegereadiness.collegeboard.org/pdf/2015-psat-nmsqt-understanding-scores.pdf
the concordance tables and percentile tables don’t tend to agree. If you just look up the percentile associated with your SI, you get one number. If you concord all your sections and add that up, then look at last year’s percentile table, you will get a much lower percentile. This makes no sense, as concordance tables are defined such that the percentiles should be equal. (And the difference is more than a single point, so it can’t be explained by the small change in the way they’ve defined percentiles.)
One observation is that a given SI can (obviously) correspond to different subscores. Points can be exchanged one-for-one between reading and writing, and you can trade 2 math points for 1 reading (or writing) point. The strange thing is that concording different subscore alternatives for the same SI can yield vastly different results.
Here is the table I generated:
SI 200 concords to between 192 (32R/33W/35.0M) and 200 (31R/38W/31.0M)
SI 202 concords to between 194 (32R/34W/35.0M) and 203 (38R/33W/30.0M)
SI 204 concords to between 196 (33R/34W/35.0M) and 205 (38R/33W/31.0M)
SI 206 concords to between 198 (34R/34W/35.0M) and 208 (38R/35W/30.0M)
SI 208 concords to between 201 (33R/36W/35.0M) and 210 (36R/38W/30.0M)
SI 210 concords to between 203 (34R/36W/35.0M) and 214 (37R/38W/30.0M)
SI 212 concords to between 206 (34R/36W/36.0M) and 218 (38R/38W/30.0M)
SI 214 concords to between 209 (34R/36W/37.0M) and 220 (38R/38W/31.0M)
SI 216 concords to between 212 (35R/36W/37.0M) and 222 (38R/38W/32.0M)
SI 218 concords to between 216 (35R/36W/38.0M) and 224 (38R/38W/33.0M)
SI 220 concords to between 220 (35R/37W/38.0M) and 225 (38R/38W/34.0M)
SI 222 concords to between 224 (35R/38W/38.0M) and 226 (38R/35W/38.0M)
SI 224 concords to between 228 (36R/38W/38.0M) and 229 (38R/38W/36.0M)
SI 226 concords to between 232 (37R/38W/38.0M) and 232 (37R/38W/38.0M)
SI 228 concords to between 236 (38R/38W/38.0M) and 236 (38R/38W/38.0M)
Let’s say that you have a 214. Maybe you have 35R / 36W / 36M. If you concord each subscore separately, you get a total concorded score of only 209. It went down… 
But note that someone who scored 38R/38W/30M would have the exact same 214 SI. Concording their individual subscores would yield a 220 (!) SI.
Let’s call the set of best possible concorded scores you can get for an equivalent SI (e.g. 200 -> 200, 210 -> 214, 220 -> 225) the “max concordance” table.
I will note that the “max concordance” table yields percentiles which are closer to agreeing with data from the previous year. (e.g. 214 (the first 99+ percentile) has a max concordance of 220, which is close® to the 223 (which is the first 99+ from the previous year after adjusting for the percentile definition change). For lower SI’s, they don’t change much / go up a little - but that would make more sense than them going down a lot…
It implies that a 218 this year is equal to a 224 last year - so maybe 218 will be around the (high state) cutoff. For a state which had a 208 cutoff last year, maybe it will be 206 this year.
Anyway, I have no idea why you would use this max equivalent concordance value, or why it would match percentile data better. But maybe it’s something to think about. It at least has the advantage that it makes “in between” guesses - between the percentile table and the raw concordance table.
