Free lesson · Alternative data
From forecast accuracy to a costed decision
Open interactive lessonPractice calculationsExplore labs
Start with the idea
A useful prediction must change an executable action enough to cover its marginal costs.
Symbols, units & horizon
- A: account equity in currency at decision
- w≥0: unsigned capital exposure fraction for the selected direction
- a: forecast gross return fraction in that direction over one holding period
- c≥0: expected total cost fraction per exposed currency unit over same period
- expected profit: currency, not guaranteed realized P&L
When and why to use this
Translate an incremental data feature into a transparent decision hurdle.
A useful prediction must change an executable action enough to cover its marginal costs.
Start with a unit exposure and subtract costs that arise from entering, exiting, financing and hedging during the exact holding horizon. Multiply by capital exposure only after all rates share the same denominator. Do not count a dataset’s forecast precision as realized trading revenue.
A strategy may need a hurdle for estimation uncertainty, drawdowns and scarce risk capacity in addition to measured costs. Compare matched baseline decisions and run a prospective paper trace before extrapolating scale.
From forecast accuracy to a costed decision
- Subtract complete horizon-matched costs from gross edge per exposed unit.
- Multiply account equity by exposure weight to obtain capital at work.
- Multiply by net expected return; compare with the actual baseline decision and uncertainty bounds.
Equity $100,000, exposure .2, expected gross return .003 and total costs .002 imply 100000×.2×.001=$20 expected profit for the period.
Apply it in a strategy
- Freeze inputs at the stated decision time and record their units.
- Translate an incremental data feature into a transparent decision hurdle.
- Recompute the example, then change the material assumption and explain the difference.
Research deliverable
From forecast accuracy to a costed decision: produce the worked calculation, a timestamped input record and a written decision addressing this limitation: Forecast error, impact and changing execution conditions can erase a small modeled net edge.
These are synthetic mechanics examples, not historical performance or paper replications. Module evidence and research boundaries record the 12 September 2026 review.
Python implementation
Self-contained teaching example. Python 3.10+; dependencies and input conventions are shown in the code and notation. Run in your own Python environment.
# Python 3.10+; standard library and NumPy only.
# Synthetic teaching inputs; conventions and units are defined in the notation above.
def expected_trade_value(equity,exposure,gross_edge,total_cost):
if equity<=0 or exposure<0 or total_cost<0: raise ValueError('Invalid equity, exposure or costs')
return equity*exposure*(gross_edge-total_cost)
assert abs(expected_trade_value(100000,.2,.003,.002)-20)<1e-12
print(expected_trade_value(100000,.2,.003,.002))Continue learning
Alternative Data: Measurement, Text & Incremental Value — all lessons- From a sampled panel to a population estimate
- A reproducible dictionary score for text
- An event return needs a predeclared benchmark
- Availability delays and signal decay
- Noisy proxies and attenuation
- Measure improvement against a frozen baseline
- From forecast accuracy to a costed decision
- A data investment includes coverage, access and ongoing costs
Quantitative finance and development glossary · Python resources and libraries · Research sources and limitations