"And the population ratios should be - I've been unwisely trying to work this out inside my head, because it's more impressive that way - one INT 18 to six or seven INT 17s to forty INT 16s to two hundred INT 15s to six hundred fifty INT 14s.  So if you've got equal INT 14 and INT 15 representation, then any individual INT 15 is about three times as likely to get admitted to Ostenso wizard academy than an INT 14.  Does all that sound right?"

He'll write it down on the whitewall in case that helps:

INT 18:       1
INT 17:    6-7
INT 16:     40
INT 15:   200
INT 14:   650

"And the population ratios should be - I've been unwisely trying to work this out inside my head, because it's more impressive that way, even if I have a higher chance of screwing up instead* - one INT 18 to six or seven INT 17s to forty INT 16s to two hundred INT 15s to six hundred fifty INT 14s.  So if you've got equal INT 14 and INT 15 representation, then any individual INT 15 is about three times as likely to get admitted to Ostenso wizard academy than an INT 14.  Does all that sound right?"

He'll write it down on the whitewall in case that helps:

INT 18:       1
INT 17:    6-7
INT 16:     40
INT 15:   200
INT 14:   650

(*)  In particular, Keltham is unthinkingly treating their Intelligence detector's integer output as a floor() function instead of a round() function.

"And the population ratios should be - I've been unwisely trying to work this out inside my head, because it's more impressive that way, even if I have a higher chance of screwing up instead* - one INT 18 to six or seven INT 17s to forty INT 16s to two hundred INT 15s to six hundred fifty INT 14s.  So if you've got equal INT 14 and INT 15 representation, then any individual INT 15 is about three times as likely to get admitted to Ostenso wizard academy than an INT 14.  Does all that sound right?"

He'll write it down on the whitewall in case that helps:

INT 18:       1
INT 17:    6-7
INT 16:     40
INT 15:   200
INT 14:   650

(*)  In particular, Keltham is unthinkingly treating their Intelligence detector's integer output as a perfect floor() function instead of a noisy round() function, though it's not like he'd know that was wrong, and it gets him pretty close.