Do you bet on questions based on your subjective probability, or to maximize rewards based on those bets?
32
Jan 1
Subjective Probability
Maximize Rewards
Other / See Results

There is a utility maximizing strategy that deviates from betting based on subjective credence, when the worlds where a market resolves YES and the worlds where a market resolves NO have different utility for mana, or for correct predictions, or when the direction of the prediction, in itself, has an impact on utility.

As a simple example, if you don't care about mana when you're dead, but like it a lot when you stay alive forever, you might bet YES on a market like "Will I stay alive for 1000 years" even though your subjective credence of this is very low. If you're wrong, you wouldn't have access to the mana anyway, and in the worlds where you're right, you value the mana infinitely more.

However, a based response to these incentives is to say "Nah, I'm predicting things honestly and ignoring profit-maximizing considerations". For example,

@Nikola says:

I don't condition on humanity still existing when buying questions. My policy is to buy questions in the direction of my subjective probability, even if this will predictably lead to me losing mana on extinction-related markets in worlds where we live. I want to maximize how reasonable my guesses are to future historians (who look back and condition on the information we had available today), not how much mana I have in worlds where I survive.

So which side are you on?

Get
Ṁ1,000
and
S3.00
Sort by:

E.g. Will Manifold stay operational next month. I should always vote yes because winning a no vote is worthless.

@RanaG and even if I want to vote no, everyone else is expected to vote yes so my vote will not profit. The market will be pushed to 99% until it suddenly resolves no at the event.

Game theoretically, you know others will move the market to maximise utility, so you're punished in the medium term if your vote deviates from the market pricing.

Based on my objective probability. I try to avoid subjective probabilities in most cases.

@Lion that which is objectively your subjective probability..? 😅 jk but i'm not sure what you mean. i take subjective probability to mean personal credence

@Bayesian No, I have a subjective probability, an objective probability, and a personal credence. Each differs from the others. (I'm anonymizing it because I'm politically unbiased on Manifold, and this is just a theoretical example for the presidential election.)

I believe candidate A will be the elected president. That is my personal credence.

Then I have my subjective probability, which is based on my personal views of the world and reasons that give candidates an advantage based on their opinions on various topics. For instance, if I agree that their point of view aligns with the population's consensus – i.e., candidate A has a 60% approval rating in my assessment, while candidate B has only 50% – this would favor candidate A over candidate B and result in my subjective probability.

Lastly, I have an objective probability, which is purely mathematical and relies on polls and the probabilities of events.

Returning to your example, I would place a bet based on an realistic objective probability (<1%). My subjective one (~10-30%) would likely be too high because I believe it's possible due to various innovations and by connecting them. To be fair, I'm stretching the 'stay alive' concept quite a bit at this point. It's probably not what most people think of when considering this question. My subjective credence and hope is a resounding YES.

Hope that makes any sense 😂

@Lion Thanks for taking time to explain your perspective. I still am pretty confused by what you mean precisely.
Your personal credence makes sense; it's your belief about the world; how likely you expect the event to happen. Presumably, though, it takes into account poll results, approval ratings, mathematical considerations, etc., basically any kind of stuff that can provide you information so that you can make a more accurate estimate of the correct subjective probability for the event; the best probability you can come up with based on what you know.

However, subjective probability more or less sounds like the same again, where you're selectively picking some pieces of evidence rather than others, and I'm confused what you mean. You say "this would favor candidate A over B and result in my subjective probability". How do you weigh these approval ratings to get to your subjective probability? I don't get what's different about the subjective prob compared to the personal credence.

Lastly, I think everything in the universe is either purely mathematical or approximatable with arbitrary precision to something that is. Polls are sometimes flawed and provide imperfect information, as do your emotions and feelings, as do facts you know about the world like that a coin flip lands around half the time on heads. I don't understand how the objective probability would deviate from the subjective one. I guess I have guesses but they are probably wrong.

"realistic objective probability (<1%) and subjective one (10-30%)" is confusing to me as a claim about your own epistemic state

@Bayesian Sorry, I didn't expect this to turn into a full-blown philosophical discussion. I'll try to remember to respond after Super Tuesday.

@Lion LOL all good

i think i mostly spend mana in limit orders at prices that are very different from my credence

I picked Maximize Rewards but there are some caveats/exceptions. Spending mana on memes, personal goal markets, etc are an obvious exception. I also bet closer to my credibility in general than is optimal, simply because I find that more fun. But when it comes to "mana is worthless to me in this scenario, so I'll bet against it" and similar choices I bet to maximize actual reward

@Tumbles makes sense, I feel similarly

@Tumbles twins!

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules