At the beginning of 2022, I decided to try the predictions thing. You can read my original predictions and commentary in this document. Now that 2022 is all done, it’s time to see how they stacked up.
There are 100 predictions in total, but a handful don’t resolve either way, so only 97 of them have a result. I had to Google some of these, and I was too lazy to document every page I looked at, so if you disagree with my assessment (some are a little subjective, but not eye-of-the-beholder subjective), leave a comment and maybe I’ll reevaluate.
Results
Format: Event — my prediction probability — outcome (TRUE/FALSE) and whether the side of 0.5 it was on matches the outcome (e.g. 0.3 and FALSE match, so would have ✅).
New Zealand politics
National has a change in leadership — 0.2 — FALSE ✅
Labour has a change in leadership — 0.1 — FALSE ✅
National leads Labour in at least two polls by different pollsters — 0.2 — TRUE ❌
TOP polls at least 2.5% in at least one poll during the year — 0.1 — TRUE ❌
Act polls above 8% in most polls between July and December — 0.2 — TRUE ❌
Greens poll above 8% in most polls between July and December — 0.7 — TRUE ✅
Jacinda Ardern leads every preferred prime minister poll, or all but one — 0.9 — TRUE ✅
An MP resigns from their party due to a conflict or scandal — 0.2 — TRUE ❌
An MP resigns from Parliament — 0.4 — TRUE ❌
At least one major party leader openly advocates for house prices to fall — 0.3 — FALSE ✅
Both major party leaders openly advocate for house prices to fall — 0.1 — FALSE ✅
Government's "immigration reset" is concrete and widely considered to be a genuine shift — 0.4 — FALSE ✅
Government introduces a bill implementing its hate speech proposals with no or only minor changes — 0.6 — FALSE ❌
Electric vehicle subsidy continues or resurfaces in some form after 31 March — 0.6 — TRUE ✅
International politics
Republicans gain control of both houses in the US midterms — 0.8 — FALSE ❌
There is a significant dispute over US federal or gubernatorial midterm election results — 0.4 — FALSE ✅
US Supreme Court overturns Roe v Wade — 0.3 — TRUE ❌
Joe Biden remains US president for the entire year — 0.9 — TRUE ✅
Donald Trump declares candidacy for 2024 presidential election — 0.2 — TRUE ❌
Emmanuel Macron wins the French presidential election — 0.6 — TRUE ✅
Labor wins the Australian federal election — 0.6 — TRUE ✅
Boris Johnson remains UK prime minister for the entire year — 0.3 — FALSE ✅
A snap election is called in the United Kingdom — 0.1 — FALSE ✅
Russia invades Ukraine — 0.6 — TRUE ✅
A country in Eastern or Northern Europe joins NATO — 0.1 — FALSE ✅
Basic Law of Hong Kong is formally made obsolete — 0.1 — FALSE ✅
There is a significant military conflict between PRC and ROC — 0.05 — FALSE ✅
Evidence surfaces in Xinjiang that sends the CCP into damage control internationally — 0.05 — FALSE ✅
United States passes major new regulation on social media platforms — 0.1 — FALSE ✅
New Zealand economics
QV House Price Index annual change to 1 Dec 2022 exceeds previous 12 months (28.42%) — 0.3 — FALSE ✅
… exceeds the 12 months to 1 Dec 2020 (13.27%) — 0.8 — FALSE ❌
Inflation in New Zealand in the 12 months to December 2022 is at least 5% — 0.6 — TRUE ✅
New Zealand official cash rate is at least 1.50% at some point during the year — 0.7 — TRUE ✅
Unemployment in New Zealand is under 5% in all four quarters — 0.7 — TRUE ✅
COVID-19, policy
MIQ stops being standard for vaccinated New Zealand citizens by 1 May — 0.8 — TRUE ✅
MIQ stops being standard for most travellers from rich countries by 1 October — 0.8 — TRUE ✅
No isolation requirements for most travellers from rich countries by 31 December — 0.5 — TRUE
A court declares some part of the MIQ system to be in breach of the NZ Bill of Rights Act — 0.3 — TRUE ❌
Foreigners no longer need a specific exemption to enter New Zealand — 0.7 — TRUE ✅
Rapid antigen tests are widely accessible in New Zealand by 1 May — 0.4 — TRUE ❌
New Zealand requires vaccine passes for restaurants all year — 0.8 — FALSE ❌
New Zealand authorities stop recommending mask use for the fully vaccinated — 0.2 — TRUE ❌
At least one major event in New Zealand is cancelled due to COVID-19 in Q4 — 0.7 — FALSE ❌
United States CDC stops recommending mask use for the fully vaccinated — 0.2 — FALSE ✅
A fourth jab is officially recommended and rolled out in the United States — 0.6 — TRUE ✅
San Francisco requires vaccine passes for restaurants all year — 0.7 — FALSE ❌
China abandons COVID-zero strategy — 0.5 — TRUE
France substantially relaxes its health pass system — 0.4 — TRUE ❌
A Pacific Island nation has a COVID-19 outbreak reaching 10% of its population — 0.6 — N/A
COVID-19 disrupts Winter Olympics enough to delay (or cancel) its completion — 0.4 — FALSE ✅
COVID-19, personal
I test positive for COVID-19 — 0.9 — TRUE ✅
I'm required to self-isolate as a close contact of someone who got COVID-19 — 0.9 — FALSE ❌
Someone I live with tests positive for COVID-19 — 0.9 — TRUE ✅
I am hospitalised for COVID-19 — 0.05 — FALSE ✅
[My employer] is a COVID-19 exposure location while at least [number] employees on site — 0.6 — TRUE ✅
[My fitness studio] is a COVID-19 exposure location (at any time) — 0.8 — N/A
[My gym] is a COVID-19 exposure location while I am there — 0.3 — FALSE ✅
[My WCS studio] is a COVID-19 exposure location — 0.5 — TRUE
[My former WCS studio] closes temporarily due to a COVID-19 outbreak among its dancers — 0.3 — FALSE ✅
Hobbies and personal
I renew my fitness studio membership — 0.5 — N/A
I attend solo dance classes for most of each of the four terms — 0.5 — TRUE
I enter a WCS Jack and Jill registry competition — 0.5 — TRUE
I run a half-marathon race — 0.3 — FALSE ✅
I run a sub-46-minute 10km race — 0.2 — FALSE ✅
I can dead hang for 2 min 10 s (30 s more than current PB) — 0.6 — TRUE ✅
I attend [redacted] for at least two months — 0.1 — FALSE ✅
I do [redacted] — 0.3 — FALSE ✅
I continue [redacted] — 0.9 — TRUE ✅
I attend commencement in person — 0.7 — TRUE ✅
I visit Australia — 0.7 — TRUE ✅
I buy a house (or sign something committing to buying a house) — 0.05 — FALSE ✅
I move out of my current house — 0.2 — FALSE ✅
I go on at least one second date — 0.1 — TRUE ❌
I enter a relationship — 0.05 — FALSE ✅
Career and work
I commit formally to a teacher training programme for 2023 — 0.8 — FALSE ❌
I commit formally to [specific training programme] for 2023 — 0.7 — FALSE ❌
[My employer] [redacted] — 0.9 — TRUE ✅
[My employer] [redacted] — 0.4 — TRUE ❌
[My employer and I] [redacted] — 0.2 — TRUE ❌
[My employer] [redacted] — 0.7 — TRUE ✅
[My employer] has [redacted] — 0.8 — FALSE ❌
I start a blog about maths education and publish at least two substantive posts — 0.5 — FALSE
I complete personal efforts to understand literature on growth mindset — 0.4 — FALSE ✅
Secondary teachers get average raises of at least 4% at start of new collective agreement on trained scale — 0.3 — FALSE ✅
Friends
[friend] moves to [country] before [date] — 0.7 — TRUE ✅
[friend] moves to [country] — 0.1 — FALSE ✅
[friend] moves to [country] — 0.3 — FALSE ✅
[friend] is pregnant (or gives birth) — 0.4 — FALSE ✅
[friend] is pregnant (or gives birth) — 0.3 — TRUE ❌
[friend] becomes eligible to [redacted] — 0.2 — TRUE ❌
[friend] earns at least one [redacted] — 0.9 — TRUE ✅
[friend] earns at least one [redacted] — 0.3 — TRUE ❌
[friends] succeed in [redacted] — 0.1 — FALSE ✅
[friends] get engaged — 0.2 — FALSE ✅
[friends] get engaged — 0.1 — FALSE ✅
[friend] finds a new job — 0.9 — TRUE ✅
[friend] defends [his/her] thesis — 0.8 — FALSE ❌
[friend] defends [his/her] thesis — 0.95 — TRUE ✅
[friend] defends [his/her] thesis — 0.95 — FALSE ❌
[friend] defends [his/her] thesis — 0.2 — FALSE ✅
Evaluation
This was a disaster.
The following chart converts probabilities < 0.5 to their inverse (e.g. 0.3 FALSE maps to 0.7 TRUE) to combine predictions of the same confidence level before aggregating. The blue line is the ideal calibration (it’s just a y = x line); the red line is how I did:
EDIT TO ADD: Here it is in tabular form, along with the size of each bin:
For some reason, 0.8s (including inverted 0.2s) went particularly badly: 8 true, 13 false, 1 unresolved, p ≈ 3 × 10⁻⁵. Maybe 0.8 is in some awkward land between “roughly not certain” and “roughly certain” and I end up overusing it. The 0.5s also went badly, though just at p ≈ 0.11 (5 from 6), so maybe it’s not too much to get hung up on.
There were quite a few calls where I was “wrong”, but that’s fine—the goal is set my confidence levels right, not to be an oracle. That said, I was probably overall too pessimistic about how quickly we would move on from COVID-19 restrictions. I don’t really have much else at this stage, go easy on me please, this was my first time trying this!
One issue with this evaluation method is that it assumes that events are independent, and a lot of them plainly aren’t. For example, half of the career and work section turned on one major career decision I knew I would have to make this year; once I decided the other way, half the section collapsed. For 2023, I might try a bit harder to avoid listing highly correlated events.
This year
I’m working on a predictions list for this year. The general idea will be similar, except:
There will be way fewer COVID-related statements.
For international politics, I’ll just use the list in Astral Codex Ten’s predictions contest here: