Couple of tabs

[undyingking]

Yay, I finally got some money from Predictify. This is the site where you make wisdom-of-crowds type predictions, and if they turn out to be correct, you get a share in the money put up by whoever it was wanted to know the answer. I've so far shared in about 35 payouts, the biggest being $12.86 for predicting Clinton's % share of the vote in the Iowa caucus, and I've made a total of $41.42 (over a period of about 6 months). Which puts me 125th in the rankings -- the most successful predictor has made over $200.
I must admit that I'm highly doubtful about the business model¹, and I was wondering whether they were ever going to be able to afford to actually pay out. But they had a venture capital round to raise the money to do so. Click here if you'd like to give it a go!

In sadder news, one of my favourite webcomics, Rice Boy, has finished. I've probably talked about this before, but if you missed that, it's a fabulously strange and beautiful thing. The Wikipedia entry seems rather baffled but gives a little idea. Anyway, read it, it's great.

---

1 Partly beause the "game" aspect of it is somewhat broken, and when players realize this, they stop behaving in a business-helpful way. I can go into more detail about this, if anyone's interested (don't all rush at once though).

Comments

[secondhand-rick.livejournal.com]

Presumably people gamble on winning 'big' by avoiding the crowd, thus skewing the results?

[undyingking.livejournal.com]

That is one problem. I do this myself on questions where basically I've got no idea what the likely correct answer is, so if I'm guessing anyway I might as well guess where the big payoff is likely to be.

[bateleur.livejournal.com]

I can go into more detail about this, if anyone's interested

Yes, predictably I am most interested!

(My guess would approximately match secondhand_rick's. Also there needs to be some kind of reliability-based bonus, otherwise there's too much incentive to spam questions about which you know nothing with complete guesses. Possibly a 10-cent buy-in or something.)

[thecesspit.livejournal.com]

Reading through the FAQ, there is an accuracy bonus. A Guru (someone who gets 75-100% of predictions in that area correct) can get a 300% bonus on their share of the cash.

[undyingking.livejournal.com]

I've identified two problems, but thinking about it and observing a bit might turn up more.

The main one I was alluding to above is like this. Assuming that you're involved for the money, rather than just the sheer joy of answering questions:

* Gaining high status is immensely valuable, as it greatly increases your payoffs.

* You jeopardize that status by answering any questions to which you're not very confident of your answer.

* So it's not worth risking answering non-payoff questions (which is almost all of the questions on the site) unless you're absolutely certain of your answer.

* So you don't take part in the overwhelming majority of non-payoff questions.

* The purpose of non-payoff questions (as far as the site's concerned) is that users pose them to each other for fun and practice, plus they serve as something to show prospective question-payers (customers) regarding the collective wisdom of the site, plus they show the site as busy and thriving. So the behaviour of skilled users as above undercuts all of these. Non-skilled or "fun" users may still answer everything in sight and chuckle when they get them wrong, but they're not what you want to be showing to your customers.

So for example I fairly quickly got my status up to Scholar in every field. I answer every payoff question that comes up, regardless of my confidence in my answer. If I get enough wrong that I lose my status in that field, then I pick a few non-payoff ones to answer to push me back over the line: then I go back to ignoring the non-payoff ones again.

This would all be mitigated if there were more payoff questions compared with non-payoff, but that won't happen unless customers can view the site with confidence, which... vicious circle.

The other problem is the one secondhandrick said: that in payoff questions where I have no idea of the answer, and where I don't think other people will really either, I guess one that's away from the hump of guesses. Assuming other people do this too, the distribution that the customer gets reported back is artificially flattened.

[bateleur.livejournal.com]

If I get enough wrong that I lose my status in that field, then I pick a few non-payoff ones to answer to push me back over the line: then I go back to ignoring the non-payoff ones again.

There is a problem there, but I don't think it's the one you describe above.

What seems to be going on here is that you find the non-payoff questions to be on average easier than the payoff questions. That's a problem if the reputation systems for the two are the same.

Lack of motivation to answer non-payoff questions isn't really a problem because - as you have proved - a player who believes themselves to be underrated will always want to answer questions, paid or otherwise.

[undyingking.livejournal.com]

The non-payoff questions aren't on average easier (I don't think), but there are so many more of them that it's simple to find plenty that are easier.

But this means I only answer a tiny proportion of the non-payoffs, and ignore many that I might think "fun". So although motivation certainly does remain, it's very narrowly targeted, and the large majority of non-payoff questions¹ will be ignored by experienced users. Maybe they have a way of spinning that to their customers.

---

1 One thing I haven't mentioned is that a lot of these questions are framed in ways that make it unlikely even a skilled contestant will have much chance of getting the right answer, so I avoid them automatically. Presumably they were hoping that the standard of framing (in the sense of people framing questions of the sorts that they themselves would like to answer) would rise as users became more experienced, but this doesn't seem to be happening.

[thecesspit.livejournal.com]

Taking a later risky answer seems very much as you suggest. They need to get answers in quickly.. and probably not by showing what others have already voted/predicted (which it seems to do at the moment, I can't see the guesses).

Some of the self-posted questions are pretty awful... partially bound, or vague.

Personally, I can't see anyone chipping out $500 to get a bunch of interweb users to guess the Length of the King of China's Nose.... even in cases (such as voting) where it can be useful. For a lot of things I can get the info for free by visiting Betfair/Oddschecker for an rough idea of the wisdom of the crowds (works well for politics/ents polls at least).

Course, these sites also useful for making a few quid out of the foolish occasionally (like the 4-1 odds I got on Bush winning the 2004 election at 1am UK time, based on one Al Gore state win, which he was always going to win).

[undyingking.livejournal.com]

Yes, opinions backed by real money are always likely to be more valuable!

I've been surprised too by how non-savvy Betfair users can be. You expect customers of terrestrial bookies to eg. over-favour England teams, but I was expecting that on Betfair that would always be balanced by savvy customers moving the other way.

[thecesspit.livejournal.com]

Interesting... not sure that gaming in the way mentioned would be the most profitable (as ruining your accuracy would appear to lower your rate of return).

[undyingking.livejournal.com]

Mm, that's the main problem really, people tend to very jealously guard their high status once they achieve it.