The results are in! Here are my thoughts on the non-COVID set of my predictions. (For detailed thoughts on the COVID predictions see Zvi's analysis)

Questions, predictions, results and analysis below. Predictions are given as (Scott:Zvi:Me). Questions are strike-through if they didn't happen.

My analysis is about whether my predictions were reasonable given what I knew. It's hard to know whether Zvi & Scott could have done better as we didn't have the same information as each other. One obvious place this shows up is in UK/US questions. I'm in the UK so should have higher confidence on UK stuff and lower on US stuff. 

Given the corrected result on q35 (see below), my score on this set of questions was marginally higher than Zvi who was marginally higher than Scott (S=1.53, Z=1.80, B=2.07) using Scott's scoring system. Unsurprisingly, I got a better log score on UK and worse on US that Zvi and Scott. The results are close enough that it doesn't really tell us that much but I'm at least happy that it wasn't a repeat of the COVID result!

(Note: I didn't predict on all of the questions)

POLITICS

21. Democrats nominate Biden, and he remains nominee on Election Day: (90:90:90)

Still seems reasonable

26. Trump is re-elected President: (50:50:60)

I outside-viewed the US election predictions. 

I probably over-relied on my reference class (recent incumbents) here and could have adjusted better for Trump’s poor approval ratings. I adjusted slightly for this but not enough.

One commentor on Zvi's prediction post suggested that this should have been even lower than 50% which seems reasonable in hindsight.

27. Democrats keep the House: (70:70:60) &

28. Republicans keep the Senate: (50:60:60) (judged inconclusive)

Not something I knew much about so just assumed slightly more likely to keep than to switch. I presumably could have done better looking at the markets but that seems like cheating.

29. Trump approval rating higher than 43% on June 1: (30:40:50)

I was too high here, I think I was a bit lazy at looking at his ratings – I kinda went “oh he’s about 43% now, I guess he’ll be about the same. However he was on about 42% so I should have predicted under 50% really.

30. Biden polling higher than Trump on June 1: (70:80:70)

Still seems reasonable, Zvi was maybe justified in being higher but I don't think I could have put more than 70%

… 

33. Boris still UK PM: (90:80:90)

I was torn between 90 and 95% on this one which I still think was the right range to be in given his recent strong election win and popularity in his party. Health was the main unknown but he’s only mid-50s.

34. No new state leaves EU: (90:90:90)

Seems reasonable.

35. UK, EU extend “transition” trade deal: (80:80:30)

Scott marked this one as true but I’m marking as false. The “transition” deal was a specific thing which wasn’t extended and the new deal which was put in place was something else (for instance, the UK left the free market and customs union on Dec 31st).

Here I had the upper hand due to being UK based and this is relatively my best prediction. The government were going to be unlikely to want to extend – their 2019 election slogan was “Get Brexit Done” which won them a sizeable majority.

36. Kim Jong-Un alive and in power: (60:80:80)

At the time there were reports of ill-health due to him not appearing in public for a while. I wrote them off as probably just speculation and he turned up not long after apparently fine. I think 80% was about right for me.

ECON AND TECH:

37. Dow is above 25,000: (70:60:70)

38. …above 30,000: (20:30:10)

38 was definitely my worst prediction.

I was torn between 10 and 20% for Dow >30,000 but even that was too low - 30% seems reasonable.

Part of the problem here was that I don’t really know much about stock market movements which isn’t too bad in and of itself. The much bigger problem was that I predicted far too confidently than I should have with this lack of knowledge. Probably there were few sensible claims I could have made about the stock market where I could have given 10% confidence. I almost completely failed to think outside the argument of what a naïve model would suggest.

Worse, this was one of the questions that I thought my probability was better than Scott’s (see my predictions post)! Of all the questions I could have put on this list, this was probably the worst choice.

I did realise my mistake when I read Zvi’s analysis of this pair of questions but by then I had already been a complete idiot.

39. Bitcoin is above $5,000: (70:60:70)

40. …above $10,000: 20%: (20:20:20) 

Still seem fairly reasonable predictions. Possibly could have been higher on the "above $10,000" option but not by much.

42. Crew Dragon reaches orbit: (80:70:90)

43. Starship reaches orbit: (40:30:20)

My predictions still seem reasonable to me and both turned out as I predicted. It is something I follow fairly closely so unless Scott and Zvi are similar they were potentially right to be less confident.

New to LessWrong?

New Comment