Foretec

Every desk wants a proven track record. There is nowhere to build one.

Foretec runs live competitions on real European electricity market data. Participants forecast, bid across day-ahead and intraday, and carry the imbalance. Every entry is scored on the money it would actually have made.

96 settlement periods · one day of imbalance prices in DK1

You have a 100 MW wind farm in DK1. How much do you sell day-ahead?

DK1 DK2
100 MW onshore wind the output you forecast and sell
20 MW · 40 MWh battery optional, an extra column in the submission

where you end up, day-ahead only where you could end up if you also trade intraday

Day-ahead only · – –
Recoverable in intraday –
Drag across the chart.
How this is scored

Sell more than you generate and you buy the shortfall back at the imbalance price; sell less and the surplus settles cheap. The optimal commitment therefore sits below your median forecast. The shaded area is what a second trade in intraday could still recover once the forecast has sharpened. It widens as your day-ahead commitment gets worse, because there is more error left to correct.

100% = selling actual generation at the day-ahead price. Day-ahead €80/MWh, intraday €72 sell / €90 buy, imbalance €108 short / €58 long with occasional spikes. Forecast median 55.5 MW on 100 MW of capacity. Illustrative DK1 conditions, not a specific trading day.

The demo assumes a well-calibrated forecast and a known price asymmetry, so it isolates the commitment decision from the forecasting problem. A real edition is harder: participants supply the distribution as well as the position, and nobody knows in advance which way the system will be short.

The same shortlist circulates, and the evidence arrives after the hire.

€20–37k

Typical agency fee for one quantitative hire in European power trading, at 15–25 % of first-year salary, for a candidate whose models nobody has measured.

€7k / month

What a European transmission system operator already pays forecasters, per market, ranked on accuracy alone. Crowdsourced forecasting is operational, not hypothetical.

0

Neutral, recurring, multi-sponsor competitions for energy trading talent in Europe. Firms run their own, occasionally, behind closed doors.

Every job advert in this market asks for a proven track record, and there is no venue in which anyone can build one. Foretec is that venue: neutral, recurring, and open to the whole market rather than to one employer.

How an edition works

  1. 01

    Qualify

    Participants bring their own data. The platform publishes the settled series every score is computed from, a baseline model and tutorials for the free sources. A qualifying round on historical periods filters the field and opens the leaderboard.

  2. 02

    Commit, daily

    Before each gate closure, every participant submits a probabilistic forecast and a position, then adjusts it in intraday as the forecast sharpens. Submissions are timestamped server-side and locked. Nobody sees the outcome before committing.

  3. 03

    Settle on realised prices

    Positions settle against published day-ahead, intraday and imbalance prices. The score is the share of perfect-foresight revenue captured, which is the benchmark a battery desk already uses internally.

  4. 04

    Meet the top of the board

    Sponsors receive the ranked pool with full performance histories and speak to whoever they want. Participants compete under a pseudonym and choose when to reveal themselves.

What a submission looks like
submission · edition DK1·01 · gate DAY-AHEAD required optional
period position_mw battery_mw p10 p50 p90 00:0041.5−2.028.343.161.000:1543.0−2.029.644.762.800:3044.20.030.946.064.100:4545.1+3.531.447.265.701:0046.8+3.532.748.967.2 ⋮⋮⋮ ⋮⋮⋮ 23:4538.6+1.0 25.240.357.9

96 periods, one row each. A single column of positions is a valid entry; the optional columns are where the edition is won.

Read the full mechanics: gates, settlement formulas, the price-impact model, what the platform publishes and the edition calendar.

The aggregate

The crowd competes too

Every submission is aggregated into a single meta-forecast that enters the leaderboard under its own name, scored on both tracks like any other participant.