There are plenty of studies that show that odds (closing lines in Pinnacle for example) in general are very accurate BUT they are also biased reflecting how people react to different kinds of bets. Basically it’s just the bookies trying to maximize their winnings.
Over/Under bias: back in the days it might have been possible to make a small win by blindly betting under because most people prefer betting over (2.5 goals, 9.5 corners etc.) As far as I know the bias is still there (and why wouldnt’t it be…), but not so clearly as before.
Favourite/Underdog bias: People looove getting their monies back multiple times whereas @1.32 doesn’t exactly get people ecstatic. Therefore too much money flows to higher odds and bookies also usually want to keep their risk levels down (1€ on 4:1 needs 4€ on 1:4 to keep the books balanced) which is one more incentive to price the underdogs lower than the actual odds might be.
My question is: If you would look at the data what is the actual yield on bets on under vs over? Also, what does the yield look like when comparing small odds vs big odds (ie. 1.3 vs 5.3)?
-My GUESS is that if the margin of an O/U bet would be 2% ( 50%/50% O/U aka 1.96 odds on both O and U) then the bookie might actually have odds like 1.94 for Over and 1.98 for Under.
–> My action (roughly): Bet about 1.5x on the suggested unders, bet less on middling over bets such as 2.5% value O2.5, skip 2.5% or less on O3.5, and skip ~2.0% valued bets on all overs.
I might also bet a bit more on lower odds even though Kelly formula takes that into account very effectively. Usually I just decide on some bets outside rebelbetting whether they are worth it or not based on this.