Mar 13-17, 2023: How to Succeed at Jeopardy! Without Daily Doubles
Stephen Webb's run concludes with five DD-less games
Attempt Watch
Total attempts came in 113.4 per game this week, an increase from 104.8 per game in the last full week of regular play (the week ending February 17). This was driven by increases from both the leading and median attempter in each game. Sometimes Stephen Webb, who spent the entire week as defending champion, was that leading attempter and sometimes not, as he broke both above and below 40 attempts over the course of the week. Attempts continue to stay elevated compared to Season 38, but remain lower than the peaks of this season.
This Week’s Champions
Stephen Webb
Average buzz score: $18156 (8th out of 28)
Average attempt value: $35359 (11th)
Average conversion value percent: 51.5% (8th)
Average buzz value percent: 64.7% (8th)
Average accuracy value percent: 80.5% (10th)
Average DD score: $2244 (17th)
Average DD+: -0.85 (27th)
Average end DJ score: $20400 (9th)
As the week went on, Ken kept continual track of Stephen’s lack of Daily Doubles, and yeah, that was really something.
Stephen didn’t find a single Daily Double for an entire week, and only found five over nine games. Among 4+ game champions, his low DD+ is trailed only by Tyler Rhode’s, who found two DDs in six games. (Fun fact: One of the first examples of me actually seeing someone else using J!ometry statistics was when Tyler looked up his DD+ on my phone to confirm it for himself.)
Without Daily Doubles to carry him, it’s all the more impressive that Stephen took leads into Final Jeopardy! in all but one of his games, with runaways in five. That turned out to be essential, as he only went 3/9 on FJ in his streak.
Stephen’s streak was also long enough that I’ve published his cumulative stats three times now — at three games, five games, and nine games — and the general trend was that he got slightly better with buzzer timing but slightly worse with accuracy. I’m inching closer to some descriptive metrics for how these balance, and if worse timing might be better for some players by preventing wrong answers. Although I think that will apply more to challengers, I’ll be looking at Stephen’s run when I get there.
Stephen’s run came to an end on Friday in what was definitely an off-day for him, as he put in his lowest attempts and attempt values, as well tying his highest number of incorrect buzzes and having a low performance in accuracy. He doesn’t show a trend towards that over the course of a longer-than-normal tape day, but it is inescapable that it was the sixth game he had that day and I have to wonder if he hit a wall. It didn’t affect his timing, but maybe the recall wasn’t as crisp, or maybe the categories were just bad for him. Despite that, he did still grab the lead going into Final but couldn’t pull the correct response and lost the lead and the game.
Amazingly, despite Stephen’s difficulties in DD hunting this week, he does not rank last among this season’s champions in DD+. But that’s only because…
Kelly Barry
Average buzz score: $11600 (24th out of 28)
Average attempt value: $25836 (26th)
Average conversion value percent: 44.9% (15th)
Average buzz value percent: 72.8% (1st)
Average accuracy value percent: 61.7% (26th)
Average DD score: $0 (25th)
Average DD+: -1.02 (28th)
Average end DJ score: $11600 (26th)
…Kelly Barry was also shut out of Daily Doubles in Friday’s match when Mark Bernstein found all three. Despite also having no DDs, Kelly was able to stay close behind Stephen at the end of Double Jeopardy and then overtake him with a correct response in Final.
I wouldn’t read much into the idea that Kelly’s raw scoring totals rank low. Often the days that champions perform their worst in actual scores are their first day and their last — which are, after all, the days that they’re playing other champions. (Zach Newkirk notwithstanding.) Kelly could very well bound upwards in Monday’s match, playing against players who are not Stephen. However, the attempt numbers are less subject to competition variation and are pretty low, and the accuracy is a concern. Kelly had the timing down extremely well for a first-time player, though, and will likely continue to hold that advantage in her next game.
Programming Notes
I made a lot of progress this weekend on the next iteration of allocating attempt probabilities across buzzing opportunities and not against clues themselves. That’s the key to next iteration of metrics overall, so I’m really pleased with it. Nothing ready to go into production yet, but forward progress is better than no progress.