Bayle Shanks's website: ideas-computer-quantcoin

summary:

quantcoin is just an idea, not an project that any work has been done on yet, or for which anyone is planning to work on in the forseeable future.

quantcoin is the idea of a bitcoin-like system solving (at least) 4 problems in bitcoin:

(problem in bitcoin) waste of computing power in hashing-proof-of-work
(problem in bitcoin) lack of any fundamental value backing bitcoin price
(problem in finance in general) people keeping automated trading strategies private, hindering the development of an academic field of research studying this area
(problem in finance in general) lack of a trustworthy way to license components of automated trading strategies, hindering the spread of useful components and increasing the cost of programming automated trading

The solutions to these are:

(a) instead of random hashing as proof-of-work, the miners' computing power is publicly auctioned off to solve computing problems for others. This solves the problem of wasing power hashing, and also creates a fundamental valuation floor for the price of quantcoin. (b) automated traders can register strategies in quantcoin on a first-come-first-serve basis. When they do so, their strategy gets to execute before anything else can happen at each network 'tick'. Strategies are executed in the order they were registered. Strategy owners may have to pay ongoing fees in order to fund the computation needed to execute their strategies. (c) in addition to strategies themselves, automated traders can register strategy components, which process inputs and produce outputs, and which can be used by registered strategies in exchange for a cut of the strategy profits.

---

more details

The first problem: in the status-quo, the market will pay you to predict which way prices will go. You only get paid to the extent that your predictions are more accurate than the consensus prediction (the market price). The market pays individually for each prediction; if you find an algorithm that predicts prices, the only way to get paid for it is to run that algorithm continually and submit the results to the market (via buying and selling). This means that you cannot publish your algorithm and still get paid (because if you publish it, the consensus will shift to incorporate your insight). This means that when people discover regularities in price movements, they tend to keep their discoveries secret. This retards the development of the theory of market dynamics, since each person (or collaborating group of people) must rediscover much of the theory themselves, and in addition cannot benefit from the feedback gained by peer review of their supposed discoveries. It would be nice if people could get paid just as much for submitting an algorithm as for incrementally submitting the results of an algorithm.

The second problem: bitcoin is a useful system that allows, (a) a total ordering, and (b) a constraint that a quantity must be conserved, to be globally, securely imposed by consensus upon a set of communications and transactions sent by various anonymous parties. The system relies upon distributing voting power according to computing power, which is evinced by doing useless computation. It would be nice if the computation done could be useful.

The solution: Quantcoin. (todo describe: the basic ideas are (a) use a bitcoin-esque priority system to determine who submitted algorithms first, and then give the algorithms money if they work; the blockchain nicely provides a consensus total ordering and discretization of 'ground truth' prices and (b) use the competitive nature of market dynamics (e.g. the market is trying to be unpredictable) to make the verification of these algorithms viable as a proof-of-work)

design goals:

should be able to integrate predictions of external prices and events
should have a way to express the semantics of bid/ask limit orders, but also be extensible to allow new order types in the future (note: quantcoin is not a broker or exchange; it is an economic protocol that makes brokers and exchanges unnecessary)
should be able to allow algorithms to be written in any language via the notion of a compiler

In traditional markets, if you publish your algorithm, others with better hardware or connectivity can move faster than you, using the same or an improved version to reap the profits before you. In Quantcoin, priority of publication determines which algorithms get to act first.

The money you make is related to the lifetime log return of your algorithm. I suppose it also has to be related to how much you can invest in it because there has to be a way for you to realize losses, e.g. if the algorithm makes some money and then loses all of it, we don't want you to be able to invest nothing and then still walk away with a profit.

You need a way to 'reserve' a strategy before you post it lest it get 'stolen' by the miner. So, sign the strategy and send the signature first. After you insert the actual plaintext strategy, your priority over other strategies will be backdated to your reservation. How to protect against the miner simply modifying the strategy in a way to change its hash but not its function (e.g. adding no-ops), and all miners rejecting all algorithm discovery transaction requests? i guess you assume that eventually there will be an 'honest' miner who will insert your plaintext algorithm discovery transaction.

The miners' reward for simulation depends on the lifetime profitability of the algorithm being simulated, and upon adding new information, e.g. you dont get paid to simulate something which was already simulated in the blockchain. e.g. it's profitable to simulate the most profitable strategies first. Now we have a choice; lifetime profitability compared only to other algorithms' performance during the same amount of time (e.g. profitability per unit time), or total lifetime? In the latter case, the senior algs get more simulation. The former case (per unit time) seems attractive to ensure that new algorithms get tried out, but the latter (sum profits over all time) seems attractive to ensure that algorithms with priority get simulated.

In fact, could define priority that way: an algorithm has priority over latter ones not just if it was introduced earlier, but only iff its sum profitability (scale-free, e.g. total return rather than how much the author actually earned) is greater than theirs. Being earlier is good only in that if you are profitable, you get to start accumulating profits sooner than clones.

In that system, a slightly improved clone will overtake you. So what if person A invents this totally novel algorithm, then person B quickly makes a tweak that improves it? Person A would get little, as the original alg would eventually lose priority to person B. One rejected fix: person A can transfer their earned priority from algorithm A to a new algorithm. But then person A could just clone algorithm B, and person B has no incentive to contribute anything.

So, perhaps we're stuck with priority. In this case, any long-term profitable algorithm must be computed forever before all later algorithms are computed. Hmm..

OK, first off, we dont need to compute algorithms that nobody has any money invested in. And the only person who can invest in them (with priority) is the original author.

proof of work could be the submission of a new algorithm which would have been net profitable for the life of the network. Clearly this difficulty increases as the network gets older; it also increases as the easy algorithms get found. Clearly this is easier to check than to create (although it's still quite difficult to check). But wont this get TOO difficult much too fast? Or will it always be possible to find marginally profitable algorithms?

discoverers of new algorithms can encode transactions in their blocks, too, and get paid for this.

(btw not sure about this, but why doesnt bitcoin have the target difficulty of proof of work be set via auction?)

would be nice to add in zerocash-style zero knowledge proofs for anonymity. Anywhere else we can use zero knowledge proofs? probably

also, a weakness in the current zerocash proposal is the need for initial parameters. apparently if whoever generates the initial parameters saved their intermediate steps, they can create an infinite amount of zerocash! this is a problem b/c surely the NSA will be bugging the machines of whoever is generating the initial parameters. a solution (for zerocoin, but not zerocash?) is apparently 'RSA UFOs' which allow the generation of a sort of initial parameter without the need to trust someone, but apparently its weakness is that it leads to 'ridiculously large' RSA moduli, which i guess make later computations inefficient? perhaps some algorithm can be devised that is like RSA UFOs but which allows an NP search for smaller RSA modulii. In this Quantcoin miners could search for such a thing! So the RSA modulii could become smaller over time. In addition, the network could provide services to others to generate this sort of initial parameter in a good way, for a fee, again giving Quantcoin an intrinsic value.