We've just launched Predictle, a new daily prediction market game.
Play at https://manifold.markets/predictle!
You can also find a link to Predictle on the Home page. Have fun!
Predictle #20
β β
β β
β β
ββ
ββ
Ngl Predictle is getting lame,
thereβs so many markets on here and it keeps asking the same ones as yesterday, bruh, we need variety
I donβt want to care about Greenland and Tesla and Chinas rebar rate (lol apparently that was even asked 2 weeks ago)
Just give random markets (maybe a minimum quality filter) cause rn it basically asks about the same top 5 markets every day
The goal of Predictle should be to get people interested in markets with low liquidity to improve the information space of this website rather than forcing everyone to look at the same top markets
@ChinmayTheMathGuy if you want to keep it the way it is, itβs probably better to make it weekly and have like 5 of them available for engagement.
Cause the odds / liquidity donβt change that much day to day
@ChinmayTheMathGuy yeah Iβm assuming whatever AI selection itβs doing it is too restrictive
Is the UI awkward for anyone else? On my phone, dragging the tiles often results in dragging the page and refreshing. Arrows to change tile positions could be an alternative.
How fast are results in predictle supposed to reflect the live odds? #3 and #4 seem to have switched places recently enough that I got it right (per live odds) but the results show different odds that are a couple hours stale.
I got 3 on 2 and 3
2 was much more straightforward, 3 was kind of ambiguous
Also whyβd I think it was Predictile?
Predictle #3
ββ β
βββ
ββ β
β β β
βββ
Interesting
with naive guessing
States:
5/5: 1 of 120, Expected Remaining steps = 0
3/5: 10 (5C2) of 120, Expected Remaining steps = 1
2/5: 20 (2*5C3) of 120, Expected Remaining steps = 1 + 1/2 (0) + 1/2 (1) = 1.5
1/5: 45 (9*5C1) of 120, Expected Remaining steps = 1 + 1/9 (0) +
0/5: 44 of 120, Expected Remaining steps =
5/5: 1
1 in 120 of 1 guess (5)
3/5:
1 in 12 of 2 guesses (3, 5)
1 in 12 of 2 guesses (2, 5)
1 in 12 of 3 guesses (2, 3, 5)
3/8
idk why Iβm struggling
but
1 guess: 1 in 120
2 guesses = 26 (10 + 10 + 5 + 1) in 120
3 guesses = 40 (10 + 25 + 5) in 120
4 guesses = 40 (15 + 25) in 120
5 guesses = 13 in 120
Where
1 = 1
10 = 10
20 = 10 + 10
45 = 5 + 25 + 15
44 = 1 + 5 + 25 + 13
Now how to update this when the knowledge is not naive
so the odds of getting it in 1 go up from
1/5 x 1/4 x 1/3 x 1/2
to around
2/6 x 2/5 x 2/4 x 2/3 = 2/45?
or in general
n/n+4 x n/n+3 x n/n+2 x n/n+1
supposing a constant knowledge factor of n
Yeah Iβm yapping
anyways I was trying to parameterize the distribution of scores as function of a single skill parameter
for unskilled we showed the distribution was
1 : 26 : 40 : 40 : 13 => average 3.317
In reality, the distribution is probably closer to
16 : 204 : 160 : 80 : 13 => 2.725
which matches up with
idk maybe the bayes factor isnβt 2^n (corresponds to 2/3 chance), probably closer to 1.5^n (3/5)
yeah I need to get better at math, this should be simple combinatorics.
One interpretation of the game is that if you disagree with the ranking, you can potentially be rewarded by helping to change it.
Non-manifold friends are annoyed at today's. The vagueness of "solving philosophy", "resolves to poll", and "better off" is ruffling some feathers.
Also 2 pairs of markets being basically tied.
Damn, #3 is rough, turns out two pairs of questions with identical %s makes it a lot harder
Perhaps enforcing a rule that the markets need at least a 5% split would be good? Having two markets with identical %s means it's effectively a coin flip whether you sort them correctly
Yeah, I wondered how the markets were selected. If it's random them a rule enforcing some minimum split would be good. If they're hand picked then having two ties seems a bit mean!
Love it. Could we get personal performance statistics over time somewhere on the profile
@CalibratedNeutral I really want a βyou did better than x% of people todayβ. Bonus points for distribution
@DanzoAlerantos Spoilers! But drag the solar one above the Trump one. If you drag it below it'll snap back.
@Xizted Shouldn't have bragged.
Predictle #3
ββββ
ββββ
ββ β β
βββ β
ββ β β
Predictle is fantastic. Kudos to everyone involved!
Predictle #2
β
β
β
β
β
Very nice job! Making this my landing page for Manifold. This feels very shareable. In my teacher education program a TON of people play a wide variety of wordle-ish games during class. Once I have a few days of experience with Predictle I plan to start shilling it.
Okay that was fun! And tricky. I rely on intrinsic valuation (what I think the likelihood is) but this forced me to do belief modeling (what others think the likelihood is).
How will this handle markets that move significantly during the day?
Is there a new meta of paradoxical markets? - "Given it is chosen for a predictile, will this market be anything but 1st place?"
this is awesome!
not sure it's an easy fix, but there's some wonky behavior between platforms. i played first on desktop and got all 5, then on mobile it took two tries, and now on browser it shows something in between, two tries except that my first guess was correct. nbd if it ends up being platform-specific but it's a bit funky

@SG great, it's working now! (when I first checked on desktop after doing it on mobile it was showing some wonky results but after refreshing the page it seems to be correct now)
Predictle #1
β β
β β
β β
ββ
ββ
I had no idea what "Anthropic flips OpenAI" means. Would be helpful if we could see the market description.
I'm washed but this is very cool.
Predictle #1
β β β β
βββ β
ββ β β
ββββ
ββββ
Predictle #1
β β β
βββ
ββ β
ββ β
βββ
Play at https://manifold.markets/predictle.
Predictle #1
β β
β β
ββ
ββ
ββ
Fun, but I have no one to share this with. π
TALK ABOUT OUR LOAN REDUCTION, I KNOW YOU HEAR US @SG
Iβm just joking, I donβt care a ton but a bit of forewarning wouldβve been nice
I do like it though. I wonder if it would be more fun NYT Flashback style, where the markets come one at a time and you have to iteratively place them on the timeline/stack (but you get the probs for prior markets).
@JamesGrugett I like the last idea there because it really helps you build an intuition for what the probability is. The game would be a little different but probably more grounded.
@JamesGrugett Personally I think being given the probabilities would be less fun and imply some precision that might not be there for some markets, but either one sounds like an improvement.
Without a link to the market, the game shouldn't include markets that are ambiguous in meaning.
"Anthropic flips OpenAI"? what does this mean? How would I know? It sort of ruins the whole game...
Right in 1 club, post in here:
Predictle #1
β
β
β
β
β
Play at https://manifold.markets/predictle?r=RWxpemE
---
BTW I tried yesterday's edition and it took 3 tries. I think I just got lucky on this one.
@marvingardens I had already played it several hours ago before that prediction, which was about my future behavior.
A lot of fun! A little more of a metaculus kind of exercise than markets exercise. I request it shows you the actual market prices at the end
@FergusArgyll The prices? It shows the %, what would showing the prices instead look like

