Rating Issue

[Micah_Smith]

@wintdoan

From what I can tell, the on-line tournament should have been initialized at 1669/5 from the FIDE rating. That looks like that was done. The recent tournament should be a blend of the 1633 OL rating (game weight around 4) with the converted FIDE 1669 (game weight 5).

I’m not sure I understand the 1700/5 (not that it matters much). It’s not like 20+1.02 x FIDE is that hard a calculation.

That formula is only used if the FIDE rating is greater than 2000. The player in question had a FIDE rating of 1617 so it uses the formula 180 + 0.94 × FIDE.

[wintdoan]

@nolan

1699.98/5 was the initialized rating (using just the FIDE rating) according to my log files for August of last year.

The event hasn’t been rerated yet, as it was submitted for rating after today’s rerate completed, so finding the initial rate is a bit more work, but I suspect the manual regular-OTB rating correction overrode the blended formula.

Sorry, grabbed the wrong branch. Yes, so 1700/5 would be correct if it weren’t for the OL tournament.

[wintdoan]

@nolan

As I understand it, here’s how the current blending formula should apply here. Tom, feel free to correct me as needed.

His first event (online regular rated) should be initialized using his FIDE rating, re-scaled to US Chess. This would produce a pre-event rating based on 5 games since his FIDE rating is under 2100.

For his first OTB event, the initialization would be blended using the post-event online regular rating AND his FIDE rating, again re-scaled. I haven’t done the math but I suspect this would produce a pre-event rating based on 10 games.

I’ll advise the ratings department, there has been some turnover there and not everyone may be familiar with how the blending formula works, they’ve only been around for a little over a year.

It gets more complicated with Canadian ratings, because we don’t have an online time-series history of them like we do for FIDE and US Chess ratings.

I’m a bit confused. Why would there be any need to “manually” add the FIDE initialization? Wouldn’t the correct procedure be to put the FIDE ID link in and let the re-rate cycle take care of everything?

[nolan]

Yes, that is the proper procedure these days, hopefully this thread will help train newer staff and remind them all of the right way to handle things.

There are times when a manual correction is necessary, this wasn’t one of them.

[Micah_Smith]

@wintdoan

@nolan

As I understand it, here’s how the current blending formula should apply here. Tom, feel free to correct me as needed.

His first event (online regular rated) should be initialized using his FIDE rating, re-scaled to US Chess. This would produce a pre-event rating based on 5 games since his FIDE rating is under 2100.

For his first OTB event, the initialization would be blended using the post-event online regular rating AND his FIDE rating, again re-scaled. I haven’t done the math but I suspect this would produce a pre-event rating based on 10 games.

I’ll advise the ratings department, there has been some turnover there and not everyone may be familiar with how the blending formula works, they’ve only been around for a little over a year.

It gets more complicated with Canadian ratings, because we don’t have an online time-series history of them like we do for FIDE and US Chess ratings.

I’m a bit confused. Why would there be any need to “manually” add the FIDE initialization? Wouldn’t the correct procedure be to put the FIDE ID link in and let the re-rate cycle take care of everything?

Yes, adding in the FIDE rating manually was a mistake made by someone in the US Chess office. This is what caused his initial US Chess OTB regular rating to be initialized only on his FIDE rating instead of a combination of his FIDE and US Chess online regular rating.

Hopefully this is a one-off mistake and the office hasn’t made this mistake in other instances.

[nolan]

I see other manually entered ratings from FIDE in the past 18 months, but it will be up to the ratings staff to review them and decide if those were handled correctly or not. If the only information we have on a player’s strength before his first US Chess event is a FIDE rating, a manual adjustment might have the same effect as the blending formula, it’s just more work than necessary these days.

In the case of this player, the manually computed rating of 1700/5 was very close to the blending formula computed one of 1699.98/5, but because the player had both a previous US Chess online event and a FIDE event before his first OTB-regular event, the manual corrected OTB-regular entry wasn’t the right way to handle it.

Although handling initialization issues properly according to current policies is something staff should be trained to do, in most cases a few points, even a few dozen points, won’t have a material impact on ratings. A player’s rating is only an approximation of his strength, the precision is probably no better than +/- 50 points for established players, the variance is likely much more for young and improving players.

[Micah_Smith]

@nolan

I see other manually entered ratings from FIDE in the past 18 months, but it will be up to the ratings staff to review them and decide if those were handled correctly or not. If the only information we have on a player’s strength before his first US Chess event is a FIDE rating, a manual adjustment might have the same effect as the blending formula, it’s just more work than necessary these days.

In the case of this player, the manually computed rating of 1700/5 was very close to the blending formula computed one of 1699.98/5, but because the player had both a previous US Chess online event and a FIDE event before his first OTB-regular event, the manual corrected OTB-regular entry wasn’t the right way to handle it.

I don’t think the 1699.98/5 is his initial rating using the blended formula. That was his initial rating using just his converted FIDE rating.

[wintdoan]

The “blending” formula can have just one input. Which it would if the only known rating were FIDE.

[nolan]

I don’t think the 1699.98/5 is his initial rating using the blended formula.

The log files for the August 2021 events say it was 1699.98/5, and that’s what was used to initialize his ONL-Reg rating. After the September 2022 event is rerated (once it has been corrected), I can report what the blended rate is for the September 2022 OTB-Regular event. (I’m not sure of the relative weighting of the FIDE rating and the August 2021 post-event ONL-Reg rating, the latter does appear to be correct.)

BTW, there will be a maintenance window on the ratings system on Sep 13-14, and the rerate for next week will be done after the maintenance window.

[Micah_Smith]

I emailed the office about this issue (and a duplicate ID issue) last Tuesday and have still gotten no response. Is there some other process I should be using instead to get these issues fixed?

[wintdoan]

@Micah Smith

I emailed the office about this issue (and a duplicate ID issue) last Tuesday and have still gotten no response. Is there some other process I should be using instead to get these issues fixed?

uschess.org/forums/viewtopi … =5&t=26101

[nolan]

I’ve sent a reminder notice about this, and also about the issue with missing records in MSA, which appears to be one our developers will need to address.

Tom Doan has raised a potential minor issue with the blended initialization formula, I will be looking into it during this week’s ratings maintenance window.

[nolan]

We’re redoing this week’s rerate after fixing a minor issue in the blending formula.

[nolan]

It looks like the blended pre-event regular rating will be 1677/9 and the post-rerate rating for ID 30255425 will be 1709/15.

Thanks to Tom Doan for catching the error in computing the blended number of games.

[Micah_Smith]

Thanks Mike for your help in getting all of this resolved.

[Micah_Smith]

Is this weeks re-rate finished?

The post-event rating from my second to last online blitz event doesn’t match the pre-event rating from my last online blitz event.

uschess.org/msa/MbrDtlTnmtHst.php?12762225

[nolan]

Internally the rerate finished around 4PM yesterday, the crosstable database on MSA appears to be up to date, the latest refresh cycle had just a few events in it, probably all recently submitted events.

This could be a caching issue, though, that’s outside of my control. They look consistent to me.

[Smythe_Dakota]

@Micah Smith

The post-event rating from my second to last online blitz event doesn’t match the pre-event rating from my last online blitz event.

They look OK to me now. All your online blitz (going back 7 or 8 of them), and your recent OTB regular, all seem to match tongue in groove, perfectly.

Bill Smythe