Branching this from The 2023 Texas States thread.
A couple things:
What software or formula(s) are you using to produce your forecasts?
Possibly a mistake in your hole 1 forecast which UDisc shows as a 717 ft Par 4?
Starting to include more OB penalties in course design is similar to when the 3-pt shot was introduced in basketball. Incorporating lots of OB areas changes the game from traditional golf scoring (mostly penalty free at the elite level*) to a new not-yet-labeled golf scoring scheme where some throws cost the player two strokes instead of one. However, unlike 3-pointers which are scored based on skill, there are many OB penalty strokes which are randomly "earned" (fluky) versus earning them strictly from very "unskillful" throws.
Point being that there are two separate games being played on tour depending on the amount of penalty elements - OB, hazard, mandos, 2-meter - actively in play within a course design and their stats should be separated, especially ratings. For those who feel ratings have become inflated, OB penalties padding propagator scores is a key factor.
*People might be surprised to know that the top 200 PGA tour players average just 1 penalty of any kind every 2.5 rounds which is 0.4 per round. The average penalties per player per round ranges from 3 to 6 on our DGPT courses heavy with OB. In a 4-round event, a PGA player might be expected to get 2-3 penalties total. On our DGPT OB courses, our players can get anywhere from 12 to 24 penalties total and that doesn't account for additional hidden strokes from losing distance on some penalties.
If we're going to have courses with OB and other penalties, do them on every hole, like using gutters in bowling, and ideally design their positions well to minimize random penalties. Or use/design traditional courses to expect few or no penalties. Separate each course type's rating stats and tour stats.
TL;DR: I overanalyzed your question and have a short, reasonable answer in the last paragraph.
As a programmer, my brain can't help but think about this. I think to have an accurate algorithm for this, we'd need to first gather a lot of data.
Specifically, we'd first need to graph where players land on what types of holes. This would allow us to see the expected drive length, range of errors, and distribution of errors for each hole. We'd need to do this for at least three types of shots: drives from the tee, putts, and anything else(approaches, second shots on par 5s, pitch outs, etc.). We have enough of this data for putts (UDisc's rough approximations), but we don't have it for drives and the "anything else" category.
Second, we'd need to map out holes a lot more accurately. Tee signs/caddie books are a great start, but they're sometimes not accurate, and I've yet to see one that is as accurate as is needed. They'd need to be drawn to scale and be 3d to some degree. They'd also need to know where the gaps are, where the trees are, etc.
Lastly, we'd need another algorithm to basically determine how a pro is going to play the hole. Take Winthrop's hole 10. Some pros play for the birdie 3 by taking the simple hole out to the left, and some go for the eagle. Determining how many would take the birdie route and how many would take the eagle route could be done by writing an algorithm that compares similar holes and estimates.
That's probably far more than what you're looking for. I think a fairly accurate algorithm could be written based off of the data we have now, plus a little bit more. If someone were to have access to the UDisc score data and perhaps PDGA ratings, they could add a few data pieces to each hole and get a decent predictor of scores. UDisc provides hole length, so we'd need to add (at least) elevation change and density (something like 1 meaning no trees, 10 being the most wooded hole ever), we could have a decent predictor of scores. It wouldn't account for weather or the holes nuances (forced layups) but it'd be a lot better than what we have now (nothing). I could write the code for most of that, but I don't have the time/desire to.