As I recall, you got 11 bonus points, but without the actual step 4 values, you can’t really do step 5. Using ratings estimator computed values is an improvement, as it got closer to the actual value.
One of the players saw a 46 point drop to their floor.
We’ve had some Original Life Masters who had a 2200 floor but whose actual playing strength was in the mid -1900’s.
In the document, it says to “use pre-event ratings” to calculate the bonus. This means that the three games are included in this case. The 11 bonus points are relatively low because they are based on Jonathan’s pre-event rating and his opponents’ pre-event ratings. That’s why I suspect the re-entered player may not have been counted somehow.
I’m 100% positive all 5 players were included, because the expected score was just over 2, otherwise it would have probably been somewhere around 1.6 or 1.7 for a 4 game event.
We tested the new software against over 300,000 sections, and it matched the existing program (except for some formula errors in the old programming) to within .000001 in nearly every case.
The closest I can come to duplicating the step 5 computation is to take the expected score of 2.08, subtract that from the actual score of 3 and multiply that by your K (36.3)
So 36.3 X .92 = 33.3 points. You gained 44 points, and as I noted above, our records show your bonus was 11. That’s within the range I’d expect due to rounding and Floating Point to Integer computations.
Sorry for so many questions and very grateful for your clear explanation, but I’m still a bit confused about the expected score and bonus, which are supposed to use pre-event ratings and should be straightforward to estimate locally.
I ran a local calculation using the initial rating 1532 against five players rated 1608, 1685, 1746, 1600, 1691, applying the pre-event ratings and the formula from the document. My expected score comes out to around 1.6–1.7, so I’m not sure why the system reports 2.
The bonus should be applied for the three wins, based on the pre-event ratings, K-factor, and threshold. My estimate is around 25–27, but the system shows 11 bonus points.
This makes me wonder if the re-entered player might be counted as unrated somehow(the UI doesn’t show it, which could explain the discrepancy). For example, if the Round 5 opponent is treated as unrated, it would result in a higher expected score but a lower bonus, which matches what I’m seeing.
Step 4 uses the pre-event rating (or the step 3 rating for newer players) when computing expected score, step 5 uses YOUR pre-event rating and your opponents’ step 4 rating.
I’ve worked with the formula for over 20 years, and was one or two courses short of a minor in math in college. Many of the members of the ratings committee have PhDs in math or physics.
Step 4 calculates the intermediate ratings, so of course pre-event ratings will be used there in calculating bonus points for the intermediate ratings. Step 5 uses the opponents’ intermediate ratings and your son’s initial rating, so the bonus that is applied at this 5th step depends on those factors. In the formula on pg 12:
Rs = R0 + K(S − E) + max(0, K(S − E) − B√m)
the “max” term is the bonus, calculated at the same time as the rest of the rating.
[Credentials: BS and MS in Math]
Thanks for the explanation. I have no doubts about the formula or the calculation steps, but I think there may be an issue with the input data when system is running, this might be edge cases.
In this case, you can see that Edward Y C Tan appears twice in the list (#63 and #88). The #63 entry has both its initial and post-event ratings empty (as shown in the UI). When the system tries to use this player’s rating to calculate other player’s expected score and bonus, it may first pick the #63 record. Since that record defaults to 0 or 100, the result ends up giving a higher expected score but a lower bonus in Jonathan’s case.
I also ran several ratings using the estimator for this section. The large discrepancy only appears when a player faces this re-entered record. In all other cases, differences are minor, usually only 1–3 points.
That’s not the way it works, I am 100% confident that players with more than one pairing number are being handled properly, we tested several thousand sections where that was the case.
That’s a display error, and only a display error.
In Jonathan’s case, his pre-event rating is 1533, 5 players: 1608, 1685, 1746, 1600, 1691
I run the calculation the expected score is around 1.6. The systems gives over 2. This impacts huge in follow steps, and improves the input data for calculating the expected score is wrong.
Round 1: 1 / (1 + 10^((1608-1532)/400)) ≈ 0.392
Round 2: 1 / (1 + 10^((1685-1532)/400)) ≈ 0.293
Round 3: 1 / (1 + 10^((1746-1532)/400)) ≈ 0.226
Round 4: 1 / (1 + 10^((1600-1532)/400)) ≈ 0.403
Round 5: 1 / (1 + 10^((1691-1532)/400)) ≈ 0.286
E_total = 0.392 + 0.293 + 0.226 + 0.403 + 0.286 ≈ 1.600
2.08 is the expected score for step 5, what you computed was the expected score for step 4. The expected score for step 5 would be higher than the one for step 4 because at least 4 of the 5 opponents have a lower rating coming out of step 4.
correct, I am not saying the step 5. I mean the bonus. it should use step 4’s expected score(1.6) to calculate the bonus not step 5(2.08) based on the doc:
E = 1.6
K(S−E)=36⋅(3−1.600)=36⋅1.400=50.4
Bm′=10⋅5≈22.36
Bonus=max(0,50.4−22.36)≈28.0
but systems gives 11.
That’s not what it says for step 5, note the phrase I highlighted in bold:
Repeat the calculations from Step 4 for every player, again using a player’s pre
event rating (or the assigned ratings from Step 1 for unrated players) to perform the
calculation, but using the results of Step 4 for the opponents’ ratings.
The bonus section of the regular formula does not override the instructions for step 5.
Thanks, that makes sense. Could you please help check the intermediate rating used in Step 4 from logs?
Thank you for your patience and professional response.
Thank you all. I also learned something as well; and my PhD was not on mathematics either.
Was February supplement generated?
Generated, files not posted yet.
I don’t have access to the logs on the new server and don’t even know if the step4 data is being logged. All I have is some step 5 data (K, expected score, new rating, bonus). I don’t know if there’s a way for me to access them and I don’t really want to ask the developers to take time to pull that data when I’m pretty sure I know what happened.
I can infer a lot of things after 20 years of experience, but I can’t confirm them.
Thank you very much for taking the time to look into Jonathan’s rating and for explaining the calculation so clearly. We truly appreciate your professionalism and patience throughout this process.
We especially want to thank you for the extra time you spent on this late last night. We recognize that this went beyond your standard responsibilities, and we are deeply grateful for your willingness to help us understand the situation.
For context, we have submitted a formal inquiry through official channels solely to seek further systemic clarification. That process is entirely separate and is in no way a reflection of your efforts here, which have been exceptional. We sincerely appreciate your support and help！
Thank you again for your time and kindness！
Wow! I sure learned something too. It doesn’t affect me personally, but may impact my opponents. I am a floored life master and, if I have a really bad tournament (like the one referenced in this thread), then my opponents earn rating points from a 2175 (say) and not a 2200. The drop would be greater for players with a larger K than me.
Michael Aigner, B.S. Math, M.S. Engineering
Since this thread is about the Golden State Open, I have another question.
Why is the February 2026 USCF regular rating for GM Nick DeFirmian still 2508 and not his post tournament rating of 2485? He has not played any other events since October.
For example, the tournament winner GM Mark Heimann saw his February 2026 USCF regular rating drop to 2590, which not coincidentally is his post tournament rating after Golden State. Other players that I checked also saw their February rating match the post tournament rating.
I compiled the NorCal Top lists and wondered why DeFirmian’s rating stayed the same despite suffering two major upset losses?
Michael Aigner
I don’t know and won’t speculate about possibilities until I can get some more facts.