See @Obgeoff’s reply. Plus you also need to subtract the average skill this week (0.16). So Scottie is 2.78 better than the average player (2.94-0.16), and the scoring average is predicted to be -0.7. So -0.7-2.78 = -3.48 strokes.
where do you get avg skill? i assume you have to download the csv table
Yeah, you do. Didn’t intentionally make this difficult, lol.
is it sg_catagory on that sheet???
No, you need to actually take the average of the “final_prediction” column in the csv (“final” column on the page).
When there is a morning round advantage, should we account for that in the stroke projections? For example, lets say that Scheffler is projected for 66.929 strokes in his next round. He has the morning round and the advantage is 0.75. Do we subtract 0.75 from Scheffler? Do we add 0.75 to the afternoon wave?
1 Like
Yes, you should, but you should cut the projected split in half. So for tomorrow, the morning wave would get a 0.375 boost and afternoon a -0.375 boost (which adds up to the 0.75 stroke split). So for projecting raw scores, you would subtract the 1/2 the wave advantage value for the morning wave guys (so Scheffler, 66.9 - 0.375 = 66.525).
I should say that when projecting R2-R4, ideally you would be using Scheffler’s updated skill (based on his performances in R1, R2, etc.). But those aren’t published on the site anywhere.
Can someone please break this down into dummy language, I subscribed to the site and simply want to use it for stroke projections for apps such as prizepicks and underdog. I have read each reply in the thread, and I simply don’t get it. Any help would be highly appreciated.
Matt, if you get the chance, could you message me on twitter @TXMeskan. Thanks, in advance.
This is an awesome thread, and I know this topic has been beaten to death. I’m trying to confirm, though, that I’m calculating a player’s scoring average correctly without access to the old tournament data referenced in previous posts. To calculate Scottie’s projected score for the first round of the Open this week, I came up with 69.455. 2.77 [Final prediction from player skill decomp page] minus (-.075) [avg player skill] minus 1.5 [scoring average from DFS page] plus .2 [half of Thursday morning wave adv] equals 1.545. I then subtracted the 1.545 from the course’s par of 71 to get 69.455. Is this correct? Thanks in advance.
Yes seems correct… to confirm: field projected scoring average is 72.5; Scottie’s skill is +2.85 relative to avg player in field, and with a 0.4 wave split Scottie in the morning wave gets a 0.2 bump; so we get 72.5 - 2.85 - 0.2 = 69.45.
Probably already been said multiple times in this thread, but I wouldn’t trust these numbers that much. Especially this week our first round scoring average projection is basically a guess (although it’s probably a decent guess… based on past Open data plus some intuition). And then obviously this week there aren’t really wave splits as tee times are continuous (so ideally you would have player-specific adjustments, which we do have in our model). Anyways, apart from those disclaimers, your number looks good.
And yes this has been beaten to death.
Thanks and disclaimers noted.
Yes, it has been beaten to death, but I think these markets could be less efficient. How much do you reckon a half stroke is worth in American odds?
hi matt, how does DG calculate the scoring average to par per round? what types of inputs go into it as it seems like one of the most important inputs on projected scores.
For example, recently DG’s prediction for the Cazoo Open de France Thursday round on 9/21/23 was predicted at 1.9 and the actual was 0.22. The early and late wave bump was also predicted to be 0.3 with the morning group having an advantage, but in fact we saw that the later group had an advantage of nearly 1.8. (+1.08 early, -0.68 afternoon).
Curious what would cause such a difference in the predicted versus actuals, any insight would be appreciated! Thank you
Historical scores at the course and weather forecasts, see FAQ.
You aren’t ever going to predict that big of a wave split (unless a hurricane is forecasted to roll in for one of the waves). As with predicting golf scores, most of the variation in course conditions can’t be predicted, so naturally some days actual scoring averages will differ a lot from predicted. Our predictions for R2-R4 are probably more accurate as R1 tells us a lot about how the course is playing that week, which can differ substantially from past years.
I see thank you, I appreciate the insight!
Also how come player decomposition / SG isnt updated after R1 or subsequent rounds? I feel like there must a lot of predictive power after seeing the first X rounds. If i were to want to approximate that do you have a simple method that might work? likely limiting the bump to be like half a stroke at maximum but i think it might be reliable enough to use. what do you think?
We do update skill levels once the tournament starts, it’s just not reflected on the skill decomposition page. That is partly because we don’t want to hand sportbooks the answers (although at the moment they are very in line with our R2-R4 matchups anyways…), and partly because we just didn’t design the page with R2-R4 updates in mind. We’ve written about the within-tournament adjustments on our methodology page.
Hey matt, just wondering if the live updated skill levels are also reflected in the rd 2,3,4 showdown draftking/fantasy predictions?
Yes, the projections for R2-R4 use live skill levels.
Ok sweet, I have below the formula I am currently using to convert Rd showdown fantasy scores to rd scores. This formula was not created by myself so I wanted to check if it was the correct way for converting the numbers.
(Par score + rd score to par - (Specific player round fantasy score - field average round score)/4.1) +(0.5* wave adjustement) (dependant on which wave the player is scheduled).
Cheers Phill