Weird AI decision on river

K our villain's running 19/6 over 20 hands. I do not have any reads on him yet apart from the fact his kings were cracked in the previous hand.



PokerStars No-Limit Hold'em, $0.10 BB (6 handed) - Poker-Stars Converter Tool from FlopTurnRiver.com

CO ($12.25)
Button ($17.85)
SB ($9.60)
Hero (BB) ($10.25)
UTG ($9.85)
MP ($15.85)

Preflop: Hero is BB with Q:spade:, 7:spade:
4 folds, SB calls $0.05, Hero checks

Flop: ($0.20) 10:spade:, 2:spade:, K:spade: (2 players)
SB checks, Hero bets $0.15, SB calls $0.15

Turn: ($0.50) K:diamond: (2 players)
SB checks, Hero bets $0.40, SB calls $0.40

River: ($1.30) 3:diamond: (2 players)
SB bets $8.95 (All-In), Hero tanks

i can tell you that you need 46.5% equity and that there are 35 possible combos which beat your hand, which means he needs something like 38 hands that you beat in his range to be neutral EV. as far as his range goes here, im pretty clueless. due to this, i just typically play results oriented here and fold because the % of times i've called overbet shoves on the river and been shown the nuts is >50 at micro.

just remembered that his KK got cracked last hand. makes it even harder.

Probable loose-passive open limps pre-flop, check/calls two bets, then donk-bet shoves river for about 7x the pot. Seems pretty foldy to me. It might help to realize that 22, K2s, K3s and KTo are all in his range the vast majority of the time.

Given history I lean towards a call.
It's one of those spots that look like they matter more than they do imo. Whether it's + or -EV I think it's marginal.