A recently registered forecaster writes:
If you have similar sentiments, read on for the response.
The market forecast is equivalent to a price… the current value/cost of a share in the “Yes” outcome.
The same basic action can be thought about, or talked about, different ways:
- “Spending points” = “trading” = “changing the market forecast”
- Or we might just call it “forecasting in the market.”
You can think about owning shares in an outcome “Yes” or “No” if that’s helpful to you.
You start each round in the markets with a fresh budget of points. Read more about it at this link.
When Replication Markets concludes, we will learn which claims actually were the subject of a replication effort. This will only occur for 5-10% of the claims available for forecasting. For these claims points/shares invested in the correct outcome will payout cash prizes (in U.S. dollars). See
the Explanation of Payouts for info on the size of those prizes.
For more info see here:
https://replicationmarkets.zendesk.com/hc/en-us
A couple of quick tips for the markets.
- Do invest all your points by spreading your bets over multiple claims. You’re likely to win more, and help our research more.
- If you bet, and then find a different claim that is, in your opinion, “more wrong,” you can withdraw points from the former claim so you can address the latter.
I’d love some feedback on this. Without dialogue, it’s sometimes hard to see what’s really being asked. Rereading our answer, I think we’re assuming too much. Perhaps some of the tutorial material here would be helpful: https://replicationmarkets.zendesk.com/hc/en-us. Give us some feedback on the video tutorials esp.
How do you forecast in the markets? You pick questions you have reason to think are far off their true value, nudge the value towards what you think it should be, wait to see how the market reacts, and repeat. You should return periodically to review and adjust. We suggest daily. More enthusiastic traders come more often, but you need not, esp. at first.
You _can_ treat it like a survey: answer questions until you run out of points, and never return, but (a) know that while you may do well, you are unlikely to do as well as active traders, and (b)
make small changes at first or you will run out of points very quickly. You are given enough points to make a few large changes), dozens of moderate ones, or hundreds of small one.
EACH TRADE IS A BET: YOU INVEST SOME OF YOUR POINTS TOWARDS AN OUTCOME, AND WILL GAIN IF THAT OUTCOME HAPPENS. The key is to realize you only have so many points, so you should invest them wisely. Robin Hanson suggests, “Edit out of outrage!” Meaning: the best you can do for us (and you) is to find places where the market estimate is WAY wrong, and correct it. But you probably want to temper that with building a portfolio, esp. because very few of our questions will resolve. So…
PAY ATTENTION TO THE COST CURVES and the details below them! They tell you how much each trade will cost you, and how much you would gain if correct. Our market revolves around the LMSR equation, the most widely-used and possibly best-understood market maker. **Costs and gains depend on RATIOS.** If you _double_ the chance of a correct outcome, you gain 100 points. If you _halve_ it, you _lose_ 100 points. So, moving “Yes” from 98% ➛ 99% will gain very little, because 99/98 is small. But it will **cost** 100 points, because you have also changed “No” from 2% ➛ 1%, cutting it in half, and we charge you now for what you might lose. (Otherwise you would take dozens of very risky bets, and leave us holding the risk.)
So… don’t push “uphill” to the extremes unless you are very sure, and there are no other claims offering better gain per cost.
Doing *really* well involves both knowing the subject area – which claims are un/likely to replicate, and understanding how to maximize your points. The amazing thing is that if cared *only* about playing the game and maximizing your points, your moves would still maximize our overall expected accuracy, at least as you see it.
You may not be a technical trader – that’s fine. There is room for expert “value traders” who give their best answers once or twice, and leave the maximizing to the gamers. But know that even their “gaming” is helping pool your insights to maximize our *overall* accuracy on hundreds of claims.
Hope that helps!
I am impressed my question made a blog post – thanks for the pretty thorough response.
I also think the linked tutorial material is pretty good, albeit hard to find (at least I did not find it initially while kind-of-seriously looking for it).
I think one of things that lead to my confusion is the UI design choice and the framing of investing as changing the probability back and forth. I understand that this is likely a product of a conscious decision to simplify the UI, but I feel like it is more complicated to understand than trading with binary options and leads to confusion, especially because people tend to perceive them initially as working in the same way as the surveys.
I think I am definitely getting a better feel for how the markets work and started to *actually* participate, which I didn’t do in earlier rounds because I felt like I didn’t understand the markets well enough.
Thanks for the question and reply, and glad to hear you’re getting more into the markets!
We thought the Zendesk help widget would be more salient, but it seems lots of people have missed it. I’ll ask for a tutorial link in the header bar.
Probability vs. binary options: if we get to Phase 2, we plan to implement a buy/sell shares UI and let people choose the one they want. We couldn’t quite fit it when making the new UI.