algostable != overcollateralized stable. Being an overcollateralized stable means it's actually backed by something, even if it's USDC. An algostable is backed by the sister token (eg: LUNA), and relies on faith and confidence in the sister token.
DAI is not algorithmic at all, but it is controlled by smart contracts. For $1.00 of DAI to be minted, a user has to post somewhere between $1.25-$1.50 of collateral. The act of creating DAI is literally that of taking out a loan. Over time, your debt to the system accrues interest. If your collateral falls under a certain price threshold, the issuer of DAI (MakerDAO), will seize your collateral to close your position. You keep the DAI, and MakerDAO liquidating you keeps the system above water and DAI at the dollar peg.
Algostables are not collateralized at all. They're more akin to ancient gold coins. Think of Luna as a small piece of gold. When UST is created, the Luna is "melted" to form the UST. At any point, the UST can then be "melted down" to get the Luna back out of it. From a certain point of view, UST literally is Luna.
> Yes, based on literally thousands of hours of video footage.
Well that's fantastic! Probabilities can be established for all sorts of in-game occurrences and the probability of this dude's game being legit or not.
Not only that, but the gamer ability can then be separated from the randomness of the game on any particular play! Are all the high scores the result of aberrations in the game, or because the player is just that great???
Speedrunners are very well aware of all the probability based events of the games they play, as well as the odds of those events and any circumstances that can change them [0]. It's called 'luck manipulation'. Usually that's only really useful to tool assisted runs, but on occasion luck manip can be performed by human players.
Regardless, while those in-the-know can recognize that a good run would have been a record if not for an unlucky event, at the end of the day all that matters is your time (or score). Accounting for the randomness is part of the strategy aspect of the sport.
[0] People who speedrun a particular game will also know of any differences in those probabilities across all known versions of the game. These people have studied these games so deeply they usually know more about how they work than the people who made them.
Probabilities have been calculated from Billy's footage that shows his alleged arcade games rendering as MAME. His random hammer smashes give points way higher than the mean, suggesting he stitched save states in an emulator.
I don't think there's disagreement - my score from 2016 is the one this article is about. The max is whatever the game's RNG will give you, and is probably somewhere close to 1.3m.
As top-tier play has gotten better, the ceiling has increased. We used to think this score was about as good as it was going to get, but only a few years prior, we thought the max was closer to 1.15m. Wild how things change.
API-QL is essentially a glue layer to make our internal interface, Stone, play nicely with Apollo. We might open source some pieces of it, but the vast majority of it is code generation for very specific Dropbox use cases.
