AEMC releases July 2024 working paper ‘Addressing the risk of algorithmic collusion’

Frequent readers here will recall how we’ve been aware of, and increasingly interested in, ‘the rise of the machines’ for a number of years.

… indeed, our investment in self-forecasting company ETSI in early 2024 was a reflection of our interest, and the benefits that can be delivered to the industry as a whole if such ‘machines’ are designed and deployed well:

(a)  and avoid engaging in the ‘Games that Self-Forecasters can play’, which

(b)  whilst they might appear initially appealing to the parties involved, (in our view) deliver long-term costs to the energy sector as a whole, and so might not be so wise in the longer term.

One of the other ‘machines’ that has arisen has been with respect to auto-bidders, and we have been increasingly looking at the volume and pattern of auto-bidder submissions in recent years.

1)  Back on 26th October 2021 (when in the process of collating Appendix N of GenInsights21) we posted ‘The rise of the auto-bidder’ to illustrate how the volume of bids being submitted had already exploded upwards

… and it’s only continued to grow (including with some very ‘noisy’ rebids).

2)  We appreciated the opportunity to share some thoughts (and concerns) of Declan Kelly in February 2025  via the ‘Hot chips, algorithms and electricity markets’ article.

3)  One of the (many!) articles that we’ve contemplated writing but have not yet had the time to start is discussed internally under a working title ‘Should participants need to pay $10 to make each rebid?’ (with the justification probably obvious to some NEM stakeholders).

So it was with interest recently that two sequential things occurred:

1)  In reading through the Draft Report from the ‘Nelson Review’ Panel after its release on 6th August 2025, we were pleased to see (at the very least) some recognition, and consideration, of these concerns

(a)  more comments to follow in a subsequent article (time permitting) about specific aspects of this

(b)  I recall reading in the Draft Report that the Panel had received input from

2)  But today (Thursday 4th September 2025) we were alerted that the AEMC had belatedly published this 15-page Staff Working Paper titled ‘Addressing the risk of algorithmic collusion’:

… so we’re drawing this to your attention today.

Definitely another for our reading list!