A Reasonable Approximation

Latest posts

My Bet Log

I occasionally make bets about future events. I think I probably should do that more often than I currently do. Given that, I think it's good practice to have a public record of the bets I've made and their outcomes, when available.

I'll try to keep this updated going forward, if I continue to make bets. Currently I'm down £72.46 with one win and one loss.

(This has nothing to do with the matched betting that I previously wrote about. There I was earning money from sign-up bonuses. Here I'm trying to earn money by being better at predicting the future than other people.)

In order of date I made the bet:

Undecided: NO on Trump leaving before the end of his first term.

I placed this bet on 20-Oct-2019, on Betfair, after a friend said on Facebook that the odds seemed very favorable to him. The odds at the time were 1.3, giving an implied probability of ${1.3 - 1 \over 1.3} = 30\%$ that he'd leave. I bet £50 against this, giving me £15 profit if he doesn't. (Minus commission, so £14.25.) As of 19-Mar-2020 the odds are 1.12, for an implied probability of 11% that he'll leave. (I could cash out now, taking my £50 back immediately and £7.56 minus commission when the market closes1. I'm not inclined to do that.)

I lost: NO on a large increase in working from home.

LessWrong user Liron said:

I also bet more than 50% chance that within 3 years at least one of {Google, Microsoft, Facebook, Amazon} will give more than 50% of their software engineers the ability to work from home for at least 80% of their workdays.

I took him up on this for $100 each. We confirmed on 02-Jan-2020. I don't think I'd heard the term "coronavirus" at the time.

This cost me $100 (£92.69 over Paypal, which I think included £2.99 transaction fee). But I got a bunch of LW karma for it. Swings and roundabouts.

Undecided: YES on someone contracting Covid-19.

In a Telegram group chat I'm in, I bet another member €30 each that at least one person in the chat would contract Covid-19 by 28-Feb-2021. (We made the bet on 28-Feb-2020.)

I wasn't confident of winning at the time, and I'm still not. Confirmed infection rates are lower than I'd anticipated - for example, Hubei has had just over 0.1% of its population confirmed as having had the disease at one point, and Italy just over 0.05%.

I think an ambiguous result is more likely than I used to, because people who get mild cases of the disease often aren't tested.

I won: DECLINE in Trump's odds of being re-elected.

On 04-Mar-2020, I noticed that the betting markets' odds on Trump's re-election had gone up since Covid-19 became a thing. (According to Election Betting Odds, rising from about 40% mid-november to about 56% at the time.) At the time, the number of officially diagnosed cases in the US was very small. I predicted that as the official number rose, and as it became more widely known that the CDC and FDA had basically fucked up, the odds would drop again.

The obvious objection was: "if you can predict that happening, so can anyone. Why haven't the odds changed already?" I didn't have an answer for that, but the stock markets had seemed to react slowly to the news2 so maybe it would happen here too. (And the next obvious objection was: "oh, so you're expecting the same thing to work twice in a row?" I didn't have an answer for that either.)

Less obvious objections that I realized/someone pointed out after I'd made the bet:

In any case, I went ahead. Again on Betfair, I bet £100 against Trump at odds of 1.76. The odds started to drop pretty soon after that. Two weeks later, on 18-Mar-2020, I cashed out (by betting £110.26 for Trump at odds of 2.1, but actually by pressing the "cash out" button). I've claimed my initial £100 back, and no matter who wins I'll earn £21.29. (£20.23 after commission, but nothing if the market is voided.)

Undecided: BIDEN to win the 2020 election.

On 05-Sep-2020, I put £30 on Biden at the same market as my previous Trump bets, at odds of 1.98 (implied probability ~51%). At the time, FiveThirtyEight, Metaculus and Good Judgment Open all gave Biden at least 65% to win (with different caveats). I have no strong opinion; my bet was on the basis that I think I trust those predictors more than I do the market. If the market gets more in line with what they're saying, I may cash out, whether I'm up or down at the time.

Note that Betfair settles the market according to projected Electoral College votes. A faithless elector, or an incumbent refusing to stand down, don't affect the result. My interpretation is that 538 and Metaculus are also trying to forecast that, so that's not the source of the difference. (I can't see any fine print on GJ Open.) I can think of a few other possible sources:

But I don't really expect that any of these explains it.

  1. I can't take any profit now, because if he gets assassinated the market is void. 

  2. This didn't do me any good. I hadn't sold any stocks at the time, and I still haven't. 

Posted on 19 March 2020

Comments elsewhere: LessWrong

comments powered by Disqus