Product Update – APT

Posted on by William Griswold

Greetings,

Ansergy released a new tool we call APT which is an acronym for Algorithmic Power Trading. The name is somewhat of a misnomer as APT serves more as a Hedging/Trading screening tool. The service has been available for clients most of the summer, and now we have made the following changes:

APT Changes

  • Heat Rate – in the past we used Power Price for setting rankings, now the model uses Heat Rates. APT is a fundamental-based system which identifies anomalous markets and screening off of heat rate, versus gas-biased power prices, delivers better results.
  • Historical Rankings – APT now shows where a product was ranked yesterday, two days, last week, two weeks, three weeks, and four weeks ago. Historical rankings are important as you can use these fields to identify fast-moving markets and flag those opportunities the day they arise.
  • Backcasts – any system is only useful if it is right more than it is wrong. APT calls the right side (buy or sell) about 65% of the time, clearly better than a hunch or a flip of the coin.

Backcast Results

You may view a complete set of backcasts measured in win|loss % or margin (btu/kwh) for several different parameters.

By Hour Type

In this example, we returned the model’s performance as measured by Hour Type. Off Peak saw 1748 “trades” and was right 67% of the time. A trade is defined as an APT market call of either Buy or Sell. The On Peak was correct 52% of the time by an average margin of 10 btu/kwh. Variable type “OO” means these were On|Off spreads and the model made the right call 59% of the time.

By Derivative

Rolls led all four derivative types with a 64% win loss rating while the spreads trailed with a 54% call. As you study these results bear in mind, there is no reversal mechanism implemented, and margins would be much higher if they were actually traded.

By RunID

Note the poor performance for Run IDs 3-6 as these are the number of traded days following the market call. Not to be surprised, poor initial returns are to be expected since APT is a “fade” system and always goes against the market move, and it takes the market a few days, or weeks, to recognize its blunders and correct. Peak results are realized, typically between two and three weeks out (Run ID 10-15). An astute trader would typically liquidate those positions during that time which is why we suggest these results are understated if the positions were actually managed.

By Hub

This grouping is total by “hub” and note that some of those are spreads. MidC and NP-MidC had the best performance while the NP-SP did the poorest. The latter is a result of the strong SP market and may also reflect a fundamental issue with the NP forecast.

You can filter on several other parameters; we suggest you spend some time with this report as it can lend insight into what works and what doesn’t with the APT system.

APT – New Fields

All in “red” are new fields.

  • WOW – week on week change in rank. We think this is perhaps the most important field in the entire table and can be used for filtering out the biggest moves. Take for example the #3 ranked trade, an NP15 Oct-Sep roll which moved up 211 spots from a week ago. Check out the Trade Rank chart on this product:
    • Hello! Where it was range-bound for months, suddenly the roll collapses into unchartered waters. Why? Most likely the Prompt month has surged and left the Oct fundamentally too cheap. Point being, the change in ranks does a good job of filtering out a potentially lucrative opportunity – either from the hedging or spec side.
  • D1 through W4 are the historical rankings for that product. D1 = yesterday, D2 = two days ago, W1 = one week ago, etc
  • MP – market price expressed in btu/kwh since all of the rankings are now based upon heat rates.

APT Historical

APT Historical is a new report and is useful for researching how the model called a particular trade over the last four weeks. Using the NP roll as an example:

The model has flagged the roll as a “Strong Buy” for the last three traded days. Before that it was either an “Ignore or Hold” or a “Moderate” buy. We find the history useful for gaining confidence in the market calls. Also note the Buy calls with a “Moderate” emphasis were all followed by small market moves up and the recommendation dropped to a “Hold/Ignore” which should have triggered a liquidation for a profit.

We are working on creating an automated email which will arrive in your inbox every morning that will include the above tables, which we are calling the “Biggest Moves.” The first table shows all 388 products sorted descending on “WOW.” The biggest moves are at the top and would be the trades to investigate further. The second table is just the opposite, these are the most significant reversals and imply a major correction.

Let’s dive deeper on that biggest reversal, the Dec SP-PV LL:

The rank changes were driven by the product coming off its high, though it is still quite rich.

Drop us a note if you would like to discuss these new features either from a spec or hedge perspective; APT is equally applicable to both. Or, if you would like to start getting an APT email each morning let us know.