How do we produce player projections?

How do we produce player projections? - SportsLine.com

2. Daily Fantasy Players doing their own research

We are especially excited about how we successfully use player simulation data to provide valuable Daily Fantasy Optimal Lineup advice. What makes our service unique is we predict what players are the best options based on predictive, not historical data. Most Daily Fantasy players have services and proprietary systems that mine historical data to identify their top players. The industry didn’t need yet another one.

Our Daily Fantasy table shows the following cutting-edge stats:

PLAYER: Includes the player’s game start time, key season stat (ex. ERA for pitchers, batting AVG for batters) compared to his simulation ERA or AVG. More importantly, if you place your cursor over the player’s name in the Optimal Lineup view, you will see COMPARABLE PLAYERS. These are other options at the same position with a similar salary that may be a better fit for you. Maybe you do not like our high expectations for Yasiel Puig and want to pick up Kole Calhoun instead. Or, perhaps you are playing an early slate and need an outfielder starting before 10:10 pm.

MATCHUP: We provide our projected runs for each team and for MLB, the opposing pitcher and his season W-L and ERA. DFS players like using game Over/Unders and lines to determine if it's a favorable matchup according to Vegas. This is another Matchup rating using our Sportsline game forecast.

FANTASY POINT FORECAST: We provide projected Fantasy points based on the league (FanDuel, DraftKings) scoring. It is based on that player's average fantasy points per simulation.

VEGAS LINE EXPECTATION (VLE): This is a unique stat where Fantasy Points are adjusted to meet the expectation of the betting line. Say Bryce Harper is projected for 7 FP and in Sportsline simulations Washington scored 5 runs. But the Vegas line has WAS -110 with an O/U of 11 runs. The Vegas Line Expectation for Washington in this case is 5.5 runs. The VLE projection adjusts the Sportsline FP to meet the scenario where WAS scored +0.5 more runs than in our simulation. Players with high VLE are likely going to be the ones that people not using Sportsline will highly rate. A player low on VLE but high on Sportsline will more frequently be the players with lower ownership percentage AND high upside.

SUBPAR+ / SOLID+ / GOOD+ / GREAT+: These are the probabilities that a player will have a subpar or better game, a solid or better game, a good or better game and a great game. The probability is the percent of simulations that a player met the Fantasy point threshold corresponding to these 4 levels of performance.

These thresholds depend on position and league scoring. The precise values are reported on the table itself (hover over the column header). These thresholds are useful in comparing two equally-rated players, based on their salary and projected FPs. If picking between two minimum-salary players who aren’t expected to put up big stats, we recommend going with the one with the higher GOOD+ percentage. Or maybe you are playing in a Tournament and want to stock up on players who won’t have high ownership percentages. You may want to identify players who have low SOLID+ percentages, but relatively high GREAT+ percentages. These are your HR hitters in MLB with low averages, and your Touchdown scorers in football with low yardage output, and your streaky shooters in basketball.

SALARY: This is the player’s salary for that league. There are certain rare cases where our data feed (provided by Rotowire) contains different salaries for the same player. In these cases we err on the side of caution and rely on the higher salary.

$/FP VALUE: This is a standard DFS metric where we calculate the number of dollars it costs per Fantasy point.

PVAL: This is another unique stat for Position Value. First we divide the player Fantasy Point projection by the position average FPs (the numerator). We then divide the player salary by his position average salary (the denominator). The average player will be 1.00. If Joc Pederson has an FP projection of 12 and the average starting Outfielder FP is 6, his numerator is 12/6 = 2.0. If Pederson has a salary of $4000 and the average OF salary is $3000 then his denominator is $4000 / $3000 = 1.33. His Position Value is 2.0 / 1.33 = 1.50.

SAFETY: This is another unique stat that is similar to the SUBPAR+ to GREAT+ metrics. In this case we calculate how many Fantasy points a player needs to score to be worth what you paid for him. We calculate the percentage of simulations where he gets at least that many fantasy points.

SFTY(HIS): This is our one historical stat. We calculate the percentage of games the player started (min. of 3 starts) over the past 365 days, where he scored at least as many Fantasy points to justify his current day’s salary. In general, better value players and those in our Optimal Lineup will have higher SAFETY than SFTY(HIS) percentages because the player is expected to have a better than average game in most cases. The SFTY(HIS) percentage is also a measure of a player's historical consistency.

OPTIMAL LINEUP

We use cutting-edge stats to calculate our Optimal Lineup for FanDuel and DraftKings. To generate the Optimal Lineup we follow these steps:

1. Rank Players by League: Using our FP projection and proprietary stats described above, we rank every player top to bottom. Before starters are announced, we expect players who are #1 on the depth chart and healthy to play significantly. But this is updated every 10 minutes based on the latest injury reports, reports of players resting, and of course Official Lineups (based on Rotowire’s research). When a player is a Confirmed Starter you will see the icon appear next to their name. We will lower a player’s rank if he is not confirmed but most of the day's lineups are confirmed. We simply would rather recommend someone we know is starting over someone who may end up being a late scratch if their values are comparable.

2. Draft Thousands of Optimal Lineups: We then generate thousands of optimal lineups following the league rules (limiting the # of players from one team, # of games / teams that must be represented, position requirements, etc.). In each of the thousands of lineups we tweak various parameters. For example, in one draft we might only pick players with a PVAL > 1.15 in the first 5 rounds. In the next one we might draft players with PVAL > 1.1 in the first 6 rounds. In another we might only draft PVAL > 1.15 in the first 5 rounds AND a salary that is no more than 50% above the remaining average salary available. We have over 6,000 parameter-driven combinations that identify the lineup projected for the most fantasy points.

3. Maximize our Salary Cap Usage: We take our Optimal Lineup and replace players IF/WHEN there is a player who fits under our cap but has even 0.1 more Fantasy points than our drafted player. For example, say we drafted Cueto for $11,900 and he was projected for 23.7 points on DK. His PVAL was 1.00 and in our Optimal Lineup Draft we only picked players with at least a 1.00 PVAL. But in this part we see Scherzer is available for $12,500 and we have room under the cap. Scherzer is projected for 23.9 fantasy points (+0.2 over Cueto) but his PVAL was just 0.95 so he was not drafted originally. We would replace Cueto for Scherzer in this case.

Lineups are fluid and change throughout the day. As lineups are set, our forecasts change and as games start our Optimal Lineups change. By the end of the day you will only see players who were starting late games. For the most part, you will not see more than 1 or 2 players change within the same hour before games start because it is rare that we have a different Optimal Lineup once our forecasts are run with confirmed starters in games soon to start. That said, there may be more changes than you are used to seeing. Of the thousands of lineups we evaluate, there is often < 0.5 total Fantasy point difference between lineup 1,235 and 1,555 and 2,343 and 5,801, etc. The slightest update to a forecast or change in lineup can adjust projections enough to make one $2500 player better than a $2400 player.

We are continuing to track and refine our optimal lineup model, and welcome your feedback on what you'd like to see.

How do we produce player projections?

Unlike most sites, our player projections are based on predictive, not historical data. Use them to win Daily Fantasy contests or player prop bets.