Re-run: Midnight Madness: Ryan Patch and Jon Seale

Ryan Patch and Jon Seale run Midnight Madness, an overnight New York City puzzle hunt for the financial industry that grew out of Goldman Sachs. Held every other year and sometimes called Compass, the event was announced to return on October 10. Patch and Seale took it over without having played it themselves, then apprenticed with its previous creators for four years to learn the tradition. The hunt traces back to designer Alicia Weasel, who ran it for about a decade before stepping down; a winning team then operated it for several years under a new name.

The organizers treat the hunt as an immersive quest: puzzles unlock the next stage, often by sending teams to another location. They distinguish puzzles, which demand lateral thinking, from tasks, which ask players to do something. A strong puzzle, in their view, makes players connect two kinds of information through a logical leap and rewards the insight without frustrating them into quitting. Unexpected New York locations and high production values are part of the experience, not just decoration.

Difficulty is difficult to calibrate, so the team relies on timing, experience, extensive playtesting, and feedback from trusted collaborators. Patch and Seale said they had been designing hunts for about 11 years and could argue through 12 rounds before agreeing on a puzzle’s difficulty; operations lead Teresa pushed for thorough testing. They recalled that the 2009 Great Gotham Challenge was intended to last four hours, but participants were only a quarter of the way through after that time and nobody finished. The example shows the risk of designing for the creators’ sense of difficulty rather than the players’ actual pace.

Hints help keep teams moving while preserving the challenge, and can matter greatly to completion times: the organizers said times could vary by 200 percent without a hint system. Midnight Madness uses a personalized controller to send hints, sometimes deliberately tormenting teams. Great Gotham Challenge instead uses prewritten hints whose time penalties shrink the longer a team has spent on a puzzle. The organizers favor well-placed clues, while describing Midnight Madness as more willing to require difficult leaps, hint-taking, and endurance.

Midnight Madness is planned roughly a year ahead, usually for early October. The intended schedule starts around 1–2 p.m. and lasts 12–14 hours, with the aim of finishing shortly after midnight, although a past event ran from about 4 p.m. until 2 p.m. the next day. The hunt raises money for Good Shepherd Services through firms sponsoring teams rather than through a conventional fundraising dinner. Teams can include up to six people, sponsorship costs about $40,000, and Goldman accounts for roughly 30–40 percent of teams, with 15–20 firms represented overall. The organizers said more of the fee now goes to charity rather than production and described an upcoming competition with 25 teams.

Past experiences illustrate the event’s production ambitions: teams have used a real Enigma machine aboard the Intrepid, changed the Empire State Building’s lights, played the F.A.O. Schwarz big piano, and viewed the city from the Port Authority bus terminal roof. A 2021 Central Park puzzle used an orb that flashed patterns based on participants’ locations. For the 2023 “lost cosmonauts” theme, the organizers considered a $10,000 inflatable pool obstacle course but could not secure a venue.

Patch and Seale say finance is a strong market for their work partly because of the overlap between finance and technology and the presence of former MIT Mystery Hunt participants in the industry. They also run corporate team-building events, usually three to four hours long, for groups ranging from four to roughly 300–400 people; Bridgewater has allowed them to say it has hired them many times, sometimes several times a year. Since the pandemic, they have seen fewer tech-company clients and smaller tech groups—often 5–12 people rather than the pre-pandemic 50–80—alongside more hedge funds and AI companies. They have informally noticed that tech clients’ performance in their games seems to correlate with company performance in the market, but do not claim that relationship is proven.