Theoretical studies have shown that in severe operating conditions, valve train friction losses are significant and have an adverse effect on fuel efficiency. However, recent studies have shown that existing valve train friction models do not reliably predict friction in boundary and mixed lubrication conditions and are not sensitive to lubricant chemistry. In these conditions, the friction losses depend on the tribological performance of tribofilms formed as a result of surface–lubricant additive interactions. In this study, key tribological parameters were extracted from a direct acting tappet type Ford Zetec SE (Sigma) valve train, and controlled experiments were performed in a block-on-ring tribometer under conditions representative of boundary lubrication in a cam and follower contact. Friction was recorded for the tribofilms formed by molybdenum dithiocarbamate (MoDTC), zinc dialkyldithiophosphate (ZDDP), detergent (calcium sulfonate), and dispersant (polyisobutylene succinimide) additives in an ester-containing synthetic polyalphaolefin (PAO) base oil on AISI E52100 steel components. A multiple linear regression technique was used to obtain a friction model in boundary lubrication from the friction data taken from the block-on-ring tribometer tests. The model was developed empirically as a function of the ZDDP, MoDTC, detergent, and dispersant concentration in the oil and the temperature and sliding speed. The resulting friction model is sensitive to lubricant chemistry in boundary lubrication. The tribofilm friction model showed sensitivity to the ZDDP–MoDTC, MoDTC–dispersant, MoDTC–speed, ZDDP–temperature, detergent–temperature, and detergent–speed interactions. Friction decreases with an increase in the temperature for all ZDDP/MoDTC ratios, and oils containing detergent and dispersant showed high friction due to antagonistic interactions between MoDTC–detergent and MoDTC–dispersant additive combinations.