Adding incentive model


#1

Here is a suggestion on how we can achieve the following goals:

  • Build the system to measure “good signals”
  • Incentive for the “makers”

Short definition

Makers (in this context)
They press the button “express your opinion”

Good signal
This is the signal/prediction which goes beyond the statistical most probable price distribution in the next Y hours

We assume that for the maker the good sensitive would be if he/she can get 50% reward in 24 hours.

Imagine the following system.

At any given time we have the statistical distribution of the possible price in the next 24 hours.
To define it - we look back the price actions in the last X hours.
We apply a model/function f(x) to get the upper (a) and lower (b) boundaries for the next 24 hours.

Signal creator have three choice:

  • for the next 24 hours the price will stay in the boundaries {a;b}
  • in the next 24 hours price will at least once go above a
  • in the next 24 hours price will go at least once below b

As an example, signal creator places 100 SANs as a collateral.
If the signal creator is right, she/he gets 150 SANs back.

These three pools are initially maintained by SAN team/treasury.
Each pool has (as an example) 10000 SANs initially.

(The detailed settings have to be checked and defined more precise. Some mathematic/statistic is needed, @tzanko ).

I do believe (intuitively) that in such a system we have all needed incentives (for the signal/prediction creator and for SAN platform) in place.

For the signal/prediction creator:

  • high potential reward (50 % in 24 hours is very hard to get)
  • clear outcome (someone will bet against you as long as reward pool isn’t committed)
  • you can’t win by just gambling (3 choices)
  • you can use it as a hedge in any market condition

For the SAN platform:

  • we get the standardised way/framework for measuring the signal quality
  • any of our own signals can be measured by the very same system (with few modifications like changing the X hours - up to 7, 14 days)
  • once the right gaming parameters and UX/UI are built, we get high SAN velocity/usage
  • as soon as we get high numbers of betting players, we get additional source of “crowd sentiment” information (that is rather distant effect, but could be interesting too)

#2

Do I understand correctly that we will be posting regularly upper and lower bounds for the price and the players will bet against them?

Such a task is possible but it requires us to create models for our price predictions. It is important how good the models are. In the given notation above a bad model would give a bigger range between a and b and this would make it less attractive for players to bet that the price would go above a or below b. A better model would give a smaller range where the price is expected to be.

I have a rough idea how to start looking for such models, but it will be an iterative process and I don’t know how good of a model we can eventually find.


#3

Correct.

I also agree and see it in the same war re: iterative approach.

The model should be good enough to start.

It’s not a big problem if some player will win against us.