Denny McLain, the AJ Score, and Predicting the 2017 Jeopardy! Tournament of Champions

[whoisalexjacob]

https://medium.com/@whoisalexjacob/denn ... d36ac0df01

[dhkendall]

https://medium.com/@whoisalexjacob/denn ... d36ac0df01

“AJ Score”. I see what you did there, Mr. Alex Jacob!

Nice work, I’m glad there’s an alternative to Coryat for measuring in these type of situations and I think it’ll get a fair bit of use over the coming years, especially from the stat obsessed among us.

[sarisson]

I like this a lot. For multiple reasons, but mainly because it's been an extremely accurate predictor of who makes it to the finals.
I came up with a "Jeopardy passer rating" a while ago, where DD's and Finals are analogous to touchdowns, and negs to interceptions, but the AJ Score is better. Can the Archivists add "AJ Score" to the stat pages of TOCers?

[seaborgium]

Can the Archivists add "AJ Score" to the stat pages of TOCers?

It's already there (for everyone, not just ToCers), as "DJ! score."

[Cat Hammarskjold]

I really enjoyed this article. The one thing that I thought was missing that I wasn't sure if it belonged or not was bottom-row clue conversion rate. Since the bottom row has a much lower conversion percentage, the expectation value per known clue is much higher, so I was wondering if any of the people who have done Jeopardy! sabermetrics so far has found out how important that is compared to other things.

[triviawayne]

I really enjoyed this article. The one thing that I thought was missing that I wasn't sure if it belonged or not was bottom-row clue conversion rate. Since the bottom row has a much lower conversion percentage, the expectation value per known clue is much higher, so I was wondering if any of the people who have done Jeopardy! sabermetrics so far has found out how important that is compared to other things.

With 60% of the value of the board there I tracked my percentage on the $1600 & $2000 clues for a while. Having that information has kept me from several incorrect guesses, so I think it’s important.

[AndyTheQuizzer]

I really enjoyed this article. The one thing that I thought was missing that I wasn't sure if it belonged or not was bottom-row clue conversion rate. Since the bottom row has a much lower conversion percentage, the expectation value per known clue is much higher, so I was wondering if any of the people who have done Jeopardy! sabermetrics so far has found out how important that is compared to other things.

I've tracked this for the ToCers as part of the "Low-Level Clues, Net" and "High-Level Clues, Net". "Low-Level" is $200, $400, and $600, and "High-Level" is $800 & $1000, my thought being that players who do better at the bottom of the board normally will do better in ToC-level competition. Also, I track it as a net average per game score, so a correct response earns +1 and an incorrect response earns -1.

That said:

Spoiler

Buzzy yesterday (net 4.6/g on high-level clues in his regular games) may have completely torpedoed that theory.