A typical scenario: a 5-point vertical spread trading at 0.25; If you short the spread, you receive 0.25 credit; the odds of payout are 1:20 (if you win, you are paid 0.25; if you lose, you are down 5); If you long the spread, the odds are reversed. Now if you have 0.9 probability of winning if you short, your Kelly fraction is a negative number; (0.05 * 0.9 - 0.1)/0.05 = -1.1; If you long spread, the Kelly fraction is 0.055. A more accurate computation is to take into consideration "continuous" outcome rather than binary outcome (ie, either gain 0.25 or lose 5). I haven't done this yet (need some time to program). But it seems that Kelly fraction favors long spread trades. Any thoughts?