How accurate are the forecasts?

Every forecast was replayed from 2024 onwards using only the data available on its date, then checked against what IRCC actually did. Each model sits next to a simple baseline so the gain is plain to see.

When the next round comes

Accuracy of the next-round dates
Type	Rounds tested	Typical miss	Baseline miss	80% window held
CEC	41	10 days	20 days	93%
PNP	56	5 days	10 days	96%
French	29	16 days	14 days	86%
Healthcare	12	41 days	42 days	83%
Other types	19	95 days	104 days	95%
All types	157	22 days	27 days	92%

Model: next-round timing model. Baseline: each type's average rate of rounds over the past year. A miss is the gap between the most likely date and the actual round.

The cutoff

Accuracy of the cutoff forecasts
Type	Rounds tested	Typical miss	80% range held	50% range held
CEC	42	6.5 points	100%	74%
French	30	24.2 points	93%	63%
Healthcare	13	13.8 points	92%	85%
PNP	57	37.0 points	88%	54%
Other types	23	18.6 points	83%	57%
All types	165	22.5 points	92%	64%

Model: the type's last cutoff with a calibrated range. In 2024-2026 an 80% range held the cutoff 89-93% of the time (conservative).

Chances of an invitation

Accuracy of the invitation chances
Within	Profiles tested	Error (Brier)	Baseline error
1 month	24,140	0.097	0.106
3 months	21,964	0.119	0.151
6 months	20,196	0.125	0.173

Model: simulated draws over six months. Baseline: above the last cutoff, times the chance of a round. Lower is better (Brier). Bands ran low for frequently drawn types, high for rare ones.

Rolling-origin backtests from 1 January 2024 to 13 October 2026, each forecast made only from data available on its date. Rounds data: IRCC.