Building a Fantasy Model

To make the best model, I want to make sure I am posing the right question to be solved.

So what are a few of the ways to describe the daily fantasy contest problem?

Should I simply predict points with more accuracy, predict the best lineup, something else?

I have NFL data from 2009 and already beat DK's projections 67% of the time but this doesn't guarantee me success.

Ideally I'd like to have all past contests with winning scores to know if a team I set is good enough to make money.

Since I don't have that data, I'm wondering what others have done.

What have you guys done or have heard done?

1. Please update your username in your Dashboard > My Profile so your email isn't displaying.

2. If you can predict points scored (projections) better than others, you can create winning lineups. You could use Colossus and upload your projections.

3. Past contests scores won't predict future winning scores. It varies week to week. Good example of this is DK's projections are just average past point scored.

4. Here is what we've done - https://www.dailyfantasycafe.com/tools/lineupoptimizer/nfl

Try to be more specific in what you're asking. What information are you asking for? How to create better projections?

Your point 3 is exactly what i'm talking about. Each week there is a new contest, new head to head matchups so since the competitors change, how can I be confident my model will work?

I might need to make a lot of assumptions.

What if I take your lineup optimizer tool and find the best team(s) given the salary and a site's projected points. Then I use my projected points and come up with optimal team 2. Maybe if team 2 is projected to outscore by 20%, I could be comfortable. Does that make sense?

Any other ways I could approach this?

By model, I'm assuming you mean Projections. You will need to track your projections over time against actual points scored to feel confident. Lineups are only as good as your projections, so yes, I would upload your projections to Colossus and find the optimal lineups.

You still can't be confident you'll win every matchup. You need to find the right matchup and I would recommend starting in 50/50 game times.

Anyone can incorrectly inflate projections, so having your projections be 20% higher on the wrong players, doesn't help you. You need to test your projections over time.

my projections are better 67% of the time. Model vs historic data.

Given only that information, how many head to head games would I win? How many large payout contents would I make money in?

If more information is needed to answer those two questions, what would that be?

my projections are better 67% of the time. Model vs historic data. - This doesn't mean much when it comes to winning in DFS. You only have 8-9 players on your team. You need to score more than other players, not historical projections.

Given only that information, how many head to head games would I win? Impossible to answer. You are playing against another human, not historical projections.

How many large payout contents would I make money in? Combination of first two answers.