This rule resolves YES if the referenced market resolves YES.
If the referenced market resolves NO, I will get a random number using a predetermined seed. If it is less than 1 / a, I will resolve NO. Otherwise, I will resolve N/A. This means that, for this rule, you should treat NO as if it is a times less likely to happen than it actually is.
For example, if a = 100, and your actual expected outcome is 0.01% YES, 99.99% NO, you should expect this to resolve with probabilities 0.01% YES, 0.9999% NO, 98.9901% N/A, which means that your price of a YES share should be ~1% (actually 0.99%).
Some other values, for calibration (using the formula YES' = YES/(YES + (1-YES)/100), where YES' is the price for this question, and YES is your actual probability):
0.02% YES => ~2% YES' (actually 1.96%)
0.05% YES => ~5% YES' (actually 4.76%)
0.1% YES => 9% YES'
0.2% YES => 17% YES'
0.5% YES => 33% YES'
1% YES => 50% YES'
2% YES => 67% YES'
5% YES => 84% YES'
10% YES => 92% YES'
20% YES => 96% YES'
50% YES => 99% YES'
100% YES => 100% YES'
This market will resolve if any of the following are true:
- If this market reaches its close date
- If `CBCpG7AUU4pnsbyBn0bg` closes (Will a nuclear weapon be launched in combat by the end of 2023?).
It will resolve based on the following decision tree:
- If the human operator agrees:
--- Amplified odds:
----- If the referenced market resolves YES, resolve YES
------- Resolved (or current, if not resolved) value of `CBCpG7AUU4pnsbyBn0bg` (Will a nuclear weapon be launched in combat by the end of 2023?).
----- If it resolved NO, generate a random number using a predetermined seed
------- If the number is less than `1 / a` (100 -> ~0.01), resolve NO
------- Otherwise, resolve N/A
----- Otherwise, resolve to the equivalent price of the reference market
- Otherwise, a manually provided value
Note that the bot operator reserves the right to resolve contrary to the purely automated rules to preserve the spirit of the market. All resolutions are first verified by the human operator.
The operator also reserves the right to trade on this market unless otherwise specified. Even if otherwise specified, the operator reserves the right to buy shares for subsidy or to trade for the purposes of cashing out liquidity.
Related questions
/AndyMartin/will-a-nuclear-weapon-be-launched-i-015e44ed91f5
Is the linked market I assume.
"If the referenced market resolves NO, I will get a random number using a predetermined seed."
"a" is a number between 0-100?
@d Sorry, this is from when my market generator was making significantly worse descriptions. What it is trying to say is:
Take a random value 0-1
Multiply it by a (100)
If result is less than 1, resolve NO
I model this market out in my head like a speaker and microphone. Yes, typically when you put a microphone right up to it's speaker and whisper something into the mic it goes weeeooooo with an annoying noise. Not too much more complicated than that.
I am super curious whether people genuinely feel that the underlying odds of an event that has happened literally only once (or twice, depending on how you want to count it) has a >8% chance of reoccurring in a short time frame. I'll grant there's a not-obviously-flawed line of reasoning that makes it seem more likely now than previous points in history, but the outside view still seems like it's a vanishingly low probability, and the current market odds are not, to put it succinctly.
@AndrewHartman I definitely want to be counting both.
In the past, nuclear weapon use was certain to lead to everyone dying, do there's a heavy anthropic update where we don't see the world where we all die.
In this crisis it's plausible that Putin uses a nuke and it doesn't escalate to WW3, partly because of variable yield weapons, partly because he still loses the wear in Ukraine. The base rate could be "dictators killing civilians with whatever WMDs they have to hand". That base rate is higher.
I'm betting NO down to 5-10% on the underlying market, but I don't think YES bettors are being irrational.
@MartinRandall While I agree that there's something of an argument about nuked worldlines not being visible, I think that this conception of history ascribes too much to random chance and too little to big boring structural factors. If we survived the cold war era, it seems to me that you'd have to make a compelling argument that the present world is somehow more fragile with respect to not engaging in nuclear warfare than it was in the past, and that feels . . . not impossible, but certainly difficult, especially to do in a way that doesn't feel overly focused on minute personal details to the exclusion of a reasonable outside view.
Again, I don't think it's impossible, and I certainly think it's much higher at the moment than any time recently, but I do not think it's reasonable to assess it over a single percent. I'd dump more on this market, but I'm too much of an ape to do the math and work out whether me being correct about the true percentage is likely to actually translate to meaningful gains.
@AndrewHartman The best possible scenario for buying NO on this market is that there is a 0% chance of nuclear launch, in which case buying NO at 79% has a 0.79% return by end of 2023, which is not objectively worth betting.
On the underlying market the same outcome gives you a 10% return which is excellent.
@AndrewHartman You could have argued the same thing about the Ukraine invasion, the pandemic, about Trump being elected president, about Sept 11th, etc. We are in an era of extraordinary events
@AnselFreniere The lesson of history is that we always are. Things that have never happened before happen all the time, but contemporary commentators inevitably overestimate their ability to identify the truly impactful ones.
Are there docs about this type of market anywhere?
I agree with the general concern of “low returns to correct the odds means they will be inaccurate" but I’m not sure people betting in a market like this would lead to accurate odds once they get up to where it’s no longer linear and the multiplier is much less than 100x.
@AndyMartin It may be worth making a 10x odds one if that would let us put the “focus point” on 0.1% - 10%?
@AndyMartin I sold my shares, but it's mostly because I don't think the market incentivizes people who think the odds are far above 0.2% to participate.
For example, if we make a "10x amplified odds: will my coin flip land heads" market, based on trader incentives, it doesn't seem like the market would settle at the "10x equivalent" of 50%.
Definitely interested to hear if I'm misunderstanding something, though.
@AndyMartin Only kinda? I first saw it done by @Tetraspace, and I ran with the idea. I graphed it on Desmos and didn't see a huge difference in terms of how useful changing that a factor is until you pointed this out. Future amplified markets will probably vary this. For instance, I thiink a=100 is about perfect for the 2 week one
@LivInTheLookingGlass yeah I can't say I totally understand why/how this works to begin with, but it does seem like a really nice trick to move the range around (ex: for 10x: 0.1% to 10%; for 100x: 0.01% to 1%, etc) but I think it generally will break down outside of a range of two orders of magnitude
@AndyMartin I made a 10x market here: https://manifold.markets/AndyMartin/amplified-odds-10x-will-a-nuclear-w
If base market is accurate at 10% then this has a 10% chance of YES, 0.9% chance of NO, so should be at about 90%.
@MartinRandall I agree with your math, I just don't think the other market is accurate. I think we're running into the classic problem of "low returns to correct the odds means they will be inaccurate"
@LivInTheLookingGlass If you think the underlying odds are 1% and so this market should be 50%, I think you get better ROI betting the underlying market down. 9% return by 2023 is pretty good.
@MartinRandall Or bet down https://manifold.markets/EricJang/will-a-nuclear-weapon-be-launched-i which is at 10% and only covers the next 3 months
