Group scores mismatch from Top Groups page and individual Group's pages

Started by pauldunn

actiasluna Lv 1

It would seem to me that the Groups and Group page scores should match and these scores would pull data automatically rather than being hand-entered… is this not the case?

LociOiling Lv 1

I'm changing the topic to "server" to reflect where the problem is.

I screen-scraped the results for the top 10 groups into a spreadsheet. I did this today, 4 June 2019, outside of normal puzzle expiration times. I counted the scores for all puzzles in the last 120 days, puzzles 1630 to 1679 in the regular science puzzles, plus a few beginner puzzles that expired on 4 February 2019.

Of the top 10 groups, only Hold My Beer appears to have a correct score on the Groups page, with 503 global points.

I'm still checking my work, but I did the screen-scraping for Gargleblasters twice and got the same results.

I'll just show the results as a list for the moment:

group	total	Foldit	difference
Beta Folders	3442	3420	22
Anthropic Dreams	3411	3337	74
Gargleblasters	2975	1903	1072
Go Science	2638	2528	110
Contenders	2478	1567	911
Void Crushers	1120	1003	117
L'Alliance Francophone	1109	1106	3
Hun-Magyar Csapat	1083	512	571
Marvin's bunch	773	772	1
Hold My Beer	503	503	0

With a little luck, you can copy and paste these lines into your own spreadsheet. I'll make the full spreadsheet available at some point, please PM me if you want a copy in the mean time.

LociOiling Lv 1

I was able to copy the results from the previous post, and paste into OpenOffice Calc.

When pasting, I right-clicked and select Paste Special… -> Unformatted text. This option brings up an "Import Text" dialog, which has a preview of the rows and columns. I accepted it with the default option, which uses the tab character as the delimiter.

LociOiling Lv 1

OK, things got a little better, but overall, it's still a mess.

The good news is that the scores on the "Groups" page now match the scores on each individual group page, at least for the first 11 teams. A win for data integrity.

The Russian Team has now moved into the top 10.

The bad news is that the totals don't seem right. I've updated my spreadsheet to include the latest puzzles. According to my calculations, the 120 day window is 12 June 2019 to 12 February 2019, puzzles 1634 to 1682.

There are a couple of potential problems I've noted. The puzzle results pages on fold.it report a "last played" date. This is not necessarily the date the puzzle ended. The entries on the results pages seem to be in descending "last played" order. I've tried to take this into account, but still might be off by a puzzle or two in some cases.

I'm assuming the points actually get awarded when each puzzle expires. In other words, if you stopped playing puzzle 1633 on 10 February, you received the points on 12 February, and they didn't expire until 12 June.

Beginner puzzles are another potential issue. They run a long time, and may appear out of sequence due to the "last played" issue. I've tried to fix a few cases where it was obvious, but I may not have found them all. Also, a couple of beginner puzzles may have gotten points since the last update that aren't counted here. I'll try to figure this out for the next revision, but I think we're talking only a few points here and there.

Here are the current totals from my spreadsheet, along with the Foldit numbers from the groups page. I don't see a easy explanation for the large discrepancies on some of these. Once again, the table below can be copied and pasted into a spreadsheet for easier viewing. PM me if you want the full spreadsheet.

rank	team	total	Foldit total	difference
1	Beta Folders	3403	3403	0
2	Anthropic Dreams	3222	3193	29
3	Gargleblasters	2903	1485	1418
4	Go Science	2735	2648	87
5	Contenders	1673	2540	-867
6	Hun-Magyar Csapat	1086	519	567
7	Void Crushers	1076	1019	57
8	L'Alliance Francophone	1063	1097	-34
9	Russian Team	789	463	326
10	Marvin's bunch	780	801	-21
11	Hold My Beer	419	425	-6

(Edit: changed the puzzle range. Puzzle 1633 should have just expired when I collected my numbers, puzzle 1634 should be the oldest science puzzle in the tally.)

bkoep Staff Lv 1

Just a quick update: we still haven't figured out why the group pages are failing to update automatically. But we have found an easy way for site admins to flush some caches manually and cause the group pages to update. For the time being, we're trying to keep things up-to-date by doing this manually whenever a puzzle closes, but if you see them go out of sync, please send a PM to me or jflat06.

I'm definitely concerned about the discrepancies that Loci found from his calculations. Can you forward the full spreadsheet to me? (See your PMs)

bkoep Staff Lv 1

I think that Loci's discrepancies are due to the same issue, where group pages were not updating. I believe the values in the "Foldit total" column were drawn from the individual group pages, which were way out of sync with the actual values. These are liable to keep falling out of sync until we pin down the issue, although we'll do our best to keep them up-to-date.

In the meantime, the rankings on the Groups page seem to be updating correctly, and should always be accurate.

Thank you LociOiling for your hard work tracking this down! If anyone continues to see problems, please let us know!

LociOiling Lv 1

Here's an interim update, reflecting puzzle 1638 to 1684, 22 February 2019 to 22 June 2019. I debated whether to include puzzle 1637, which expired on 21 February at 23:00 GMT. I collected these results today at 18:20-19:20 GMT. Adding puzzle 1637 didn't seem to help the confusion, so I left it out, but I'll recheck in a couple of days to get a clearer window.

As before, the table below can be copied and pasted into the spreadsheet of your choice, which will improve the formatting. Full spreadsheet available on request.

The first three columns were taken from the Foldit Groups page, also known as "Top Groups".

The "Individual Groups Page" column was taken from the each group's page. Once again, there's no agreement between the two sources.

The "Screen Scraped Scores" column was taken from the "show_scores" page for each group. Once again, this was just an update to the previous spreadsheets. The table has been sorted in descending order on this column.

There's only one match in the entire table. The score shown for Hold My Beer on the Top Groups page matches my screen-scraped results.

A difference of 100 points or so might reflect one science puzzle being counted wrong. I'm having a harder time with the large differences of 300 points or more.

Rank	Group	Top Groups Page	Individual Group Page	Top – Individual	Screen Scraped Scores	Top – SSS	Individual – SSS
#1	Beta Folders	3208	3367	-159	3307	-99	60
#2	Anthropic Dreams	3140	3178	-38	3143	-3	35
#4	Gargleblasters	2732	2649	83	2980	-248	-331
#3	Go Science	2864	2660	204	2972	-108	-312
#5	Contenders	1139	1324	-185	1405	-266	-81
#7	Void Crushers	940	921	19	1044	-104	-123
#6	L'Alliance Francophone	942	1018	-76	939	3	79
#10	Hun-Magyar Csapat	450	522	-72	922	-472	-400
#9	Russian team	461	459	2	795	-334	-336
#8	Marvin's bunch	780	814	-34	781	-1	33
#11	Hold My Beer	406	416	-10	406	0	10

LociOiling Lv 1

Here's the latest, with numbers collected before 1687 expired.

Since the last update, bkoep got things updated, so the global score on the Groups page now matches what is shown on each individual group page. There's now a whole column of zeroes, long may it last.

Bkoep also noted that beginner puzzle points are not included in a group's global points. (Individual players do receive global points for beginner puzzles.) I removed all the beginner puzzles, and also any other puzzles that awarded zero points.

The screen-scraped scores below include 49 puzzles, from 1639 to 1686. Hold My Beer sat out puzzles 1684 and 1652, but the other groups played all 49. As before, my assumption is that puzzles that expired in the last 120 days are counted in global points totals.

Despite everything, the screen scrapings don't agree with what's shown on the group pages. Gargleblasters shows an especially large discrepancy, 207 points. I re-scraped the G-blasters score, but got the same results again.

As before, you can copy the table below and paste into a spreadsheet for better formatting.

Rank	Group	Global Score	Individual Group Page	Top – Individual	Screen Scraped Scores	Top – SSS	Individual – SSS
#1	Beta Folders	3208	3208	0	3207	1	1
#2	Anthropic Dreams	3140	3140	0	3085	55	55
#3	Go Science	2864	2864	0	2755	109	109
#4	Gargleblasters	2732	2732	0	2525	207	207
#5	Contenders	1139	1139	0	1139	0	0
#6	L'Alliance Francophone	942	942	0	942	0	0
#7	Void Crushers	940	940	0	907	33	33
#8	Marvin's bunch	780	780	0	778	2	2
#9	Russian team	461	461	0	460	1	1
#11	Hold My Beer	406	406	0	406	0	0
#10	Hun-Magyar Csapat	450	450	0	380	70	70

LociOiling Lv 1

Here are the latest results, gathered 7/16/19 starting at 2100 GMT, and finishing 7/17/19 at 0330 GMT.

Anthropic Dreams is now #1, despite a slight discrepancy between the total shown on the web site and my screen-scraped totals.

Beta Folders is second, based on my screen-scraped numbers, but shows as third place on fold.it.

There are large discrepancies for Go Science and Gargleblasters, comparing the web site totals to my screen scraping.

Once again, the global scores on the Groups page match what is shown on each individual group page.

As before, the numbers below will look better if copied and pasted into a spreadsheet. Complete spreadsheet available on request.

The table is sorted in descending order by the "Screen Scraped Scores" column.

Rank	Group	Global Score	Individual Group Page	Top – Individual	Screen Scraped Scores	Top – SSS	Individual – SSS
#1	Anthropic Dreams	3137	3137	0	3129	8	8
#3	Beta Folders	2995	2995	0	2994	1	1
#2	Go Science	3001	3001	0	2892	109	109
#4	Gargleblasters	2723	2723	0	2516	207	207
#5	Contenders	956	956	0	956	0	0
#6	L'Alliance Francophone	905	905	0	905	0	0
#7	Marvin's bunch	851	851	0	833	18	18
#7	Void Crushers	851	851	0	818	33	33
#10	Russian team	416	416	0	415	1	1
#9	Hun-Magyar Csapat	429	429	0	359	70	70
#11	Hold My Beer	339	339	0	335	4	4