Fantasy sites offer several different types of fantasy games, including head-to-head, 50/50, and tournaments. Head-to-head pits the user against one other user, and the user whose team scores the most points is paid. 50/50 games enter the user into a large pool of other users, and the top half of high-scoring users is paid. Tournaments also enter the user into a large field, but only the top 10-20% of high scorers are paid.
We plan to enter most of the money in head-to-heads, since this will be the best way to test the accuracy of our automatic player while reducing risk. Tournaments are not optimal because they involve a huge amount of variance and can be disproportionately influenced by fluke events. We plan to use only a few different lineups each week, so 50/50s are riskier than head-to-heads. If we enter a badly predicted lineup into many 50/50s, it is likely that we will lose all the 50/50s for that week, whereas head-to-heads are more independent events.
The “juice”, or percentage of the entry fee that sites keep for themselves, is 10% both on DraftKings and FanDuel, the two largest daily fantasy sites. This means that we would have to have a long-term winning percentage of 55.6% to break even on head-to-heads. Our automatic player aims for a winning percentage of 60%: not too optimistic, but still in the money.
fantasylearning 