Thanks for everyone’s feedback on the Quick SNITCH (or “Quitch” as coined by Templar here, thanks Templar!).
I’ve now added an initial set of Blog Solvers for this. Blog Solvers’ times are taken from the details they provide in the TfTT blog comments. These times are aggregated and reported (as a separate type of solver) alongside the times of the Reference and Tracked Solvers from the Crossword Club.
If you’ve been marked as a Blog Solver but don’t want to be, please let me know. Conversely, if you’d like to be tracked as a Blog Solver, also let me know. You will need at least 15 error-free times to have been listed in the blog to begin.
Thanks again for your input.
So how are you able to pull the times out of the comments? No only that, you would need some primitive sort of AI to be able to tell whether the commenter completed the puzzle successfully or not. I know that some of our users post their times at the beginning or the end of their comment, but most do not, and some give their times and then say they had an error.
A good question … it’s not 100% reliable but works in most cases. I look for a time in various formats (15:23, 15.23, 15m23s, 15 minutes, 15’23”, …) ideally at the start of the blog. Then I look for indications of errors (an/one error, pink square, DNF, …). There’s a hierarchy of sorts (a match at the beginning takes precedence over a match at the end, etc; if i read “one error” or “no errors”, I don’t look for “pink squares”, etc). I won’t bore you with all the details 🙂
If people want to be included as Blog Solvers, I recommend that they put their time up front, and be explicit about the number of errors (e.g. 15:23, no errors). This is pretty reliable.
I don’t expect a time for every comment, of course, and I don’t use any of the blog times for the SNITCH calculations.
I always put my time first if no errors or aids but otherwise I put DNF. So I guess my results would be useful data for you.
I’ll start posting my times at the start, as I do for the 15×15. Thanks for all your work on this!
I still can’t help but feel this doesn’t quite represent what’s going on accurately. Discounting DNFs removes a huge part of assessing difficulty.
Prior to today, my successful times over the past 10 have been … 00:20:16 00:21:25 00:19:26 00:11:05 00:14:30 00:24:36 00:18:12 00:07:56 00:14:25 00:18:04
But interspersed in there are some DNFs of 28:38, 56:58, 23:16 – three of the worst four times.
Now today I took an hour for a successful completion on Izetti so that can contribute to showing it’s hard. But if I’d quit or made a mistake it wouldn’t be included and it’s discounted.
Hi L-Plates, thanks for your comments and ongoing interest in the scoring.
A question: for your DNFs with high times, how did the SNITCH rating for those puzzles compare to the ones you finished and did more quickly?
If the SNITCH calculation shows that these puzzles were generally harder even without your data being included, then I’d say it’s working okay. The chances are quite good that if you took longer and/or had a DNF, then those who completed the crossword might also have taken longer and the SNITCH rating would be higher as a result.
Hi starstruck_au
Thought it might be easier if I give you my data? Maybe that will also allow you to add me as blog reference solver as there 26 completions listed.
The comma-delimited columns are QC#, Snitch value, Solve,Time(HH:MM:SS), Setter
2510,106,No,00:28:38,Mara
2511,119,No,00:56:58,Orpheus
2512,124,Yes,00:20:16,Wurm
2513,104,Yes,00:21:25,Teazel
2514,99,Yes,00:19:26,Hurley
2515,69,Yes,00:11:05,Trelawney
2516,91,Yes,00:14:30,Wurm
2517,95,Yes,00:24:36,Pedro
2518,107,No,00:23:16,Alex
2519,141,Yes,00:18:12,Breadman
2520,66,Yes,00:07:56,Joker
2521,92,Yes,00:14:25,Teazel
2522,91,Yes,00:18:04,Mara
2523,137,Yes,01:00:00,Izetti
2524,101,Yes,00:20:39,Felix
2525,83,Yes,00:23:20,Oink
2526,98,Yes,00:34:06,Joker
2527,82,Yes,00:12:43,Hurley
2528,97,No,00:21:38,Beck
2529,103,Yes,00:40:00,Wurm
2530,103,Yes,00:38:47,Izetti
2531,114,Yes,00:47:25,Orpheus
2532,103,Yes,00:17:08,Breadman
2533,118,Yes,00:20:40,Peridot
2534,95,Yes,00:31:30,Hurley
2535,96,Yes,00:44:07,Teazel
2536,104,Yes,00:24:01,Orpheus
2537,85,Yes,00:20:53,Mara
2538,91,Yes,00:34:51,Felix
2539,113,Yes,00:46:21,Rongo
Where I DNFed the time includes correcting it to a solve – usually only a few seconds extra.
There’s a couple of estimated times Izetti’s 2523 and Wurm’s 2529. The latter I actually took a long break – so it would be hours if taken on elapsed!
A quick note to let Blog Solvers know that the following keywords will trigger the result being counted with an error (and therefore your time won’t show up in the SNITCH detailed results):
• A non-zero digit followed by “error” or “errors”
• pink square
• gave up | giving up
• typo
• DNF | dnf
If you want to use any of these terms, putting “0 errors” after your time will override anything else.
Sorry that this is not very smart – maybe generative AI will give some better options.
Starstruck, just to stake my claim – I coined the neologism “The Quitch” (in my blog entry for QC No 2489 by Mara on 22 September 2023)!
Thanks again for creating it, it’s just fantastic.
Templar
Thanks, Templar! Now duly noted in the edited post 🙂