As we have understood, the specification explains how customers will evaluate variable rate tariffs according to their risk. However, we've been unable to find any evidence of this in the code. As far as we can see, customers evaluate the utility of tariffs based purely on expected cost.
Have we misunderstood the current implementation or is this in fact the case? If so, will tariff risk be implemented for the pilot competition?
Because of our tight schedule for the completion of more important subsystems of the server, we thought that the implementation of the variable tariff risk isn't yet developed in a satisfactory level and won't be included in this version of the server.
So the current implementation of the server, based solely in cost evaluation and some degree of random selection of the best suited tariff is the one that will be used in the pilot version.
We understand its impossible to implement everything for the pilot, and the evaluation of fixed tariffs will be fair even without customers evaluating tariff risk. However, how will the evaluation of variable tariffs work? Surely customers will always evaluate a variable tariff's cost according to its expectedMean outside the tariff's notice interval. Is there a penalty to brokers who publish variable tariffs for which the actual mean deviates from the expectedMean?
you hit a key issue we were discussing many times during our game design discussions. The answer is two-fold with only one element being described in the spec so far.
1. A customer evaluating the tariff risk of a published dynamic tariff uses the realized prices to calculate historical payments he would have incurred under this tariff. The mean of these is the cost evaluation while the variance is a proxy for tariff risk (this follows a finance notion of risk). Alternatively we discussed directly evaluating hourly rate variance of the published tariff which would reduce computational load on the customer models.
Note: The biggest problem is therefore evaluating a newly published variable tariff with no history of realized prices.
2. To solve this we aim to supplement tariff utility with a component for broker reputation - i.e. a broker regularly charging maximum variable rates builds up a negative reputation over the course of the game. This has at the moment only been conceptualized.
Sounds great, that was the sort of solution we were expecting. However, since this will not be implemented in the pilot, are you still planning on allowing variable rate tariffs in the pilot? Surely with no notion of risk, a broker has no incentive to meet their expectedMean and can instead charge the maximum price for the life of the tariff with no penalty.
Since balancing is one key tasked of brokers dynamic tariffs may even under a no-reputation scheme still be used in a truthful way as one will not induce shifting by charging the maximum tariff in all periods. But brokers could use the misreport as a systematic way to shift all rates up.
I can't remember right now - are dynamic tariffs defined by a rate band? If so, then the width of the band could serve as an alternative proxy for tariff risk.
I agree, setting the maximum price of a variable rate tariff in all time periods will not induce any load shifting by the consumers. However, it's the misrepresentation of tariff utility which I'm worried about.
Variable tariff rates have a minimum and a maximum rate price. Since a low price is good for the customer, maybe the difference between the expectedMean and the maxPrice should serve as method of evaluating risk. This would penalise tariffs which define a low expectedMean but also a high maxPrice.
Without any notion of tariff risk or broker reputation, I honestly don't see a fair way variable rate tariffs can be included in the pilot competition.