Anyone who has played fantasy football for any length of time has run across team owners who profess that "fantasy football is mostly about luck", or that "fantasy football is half luck, half skill". Usually - and not surprisingly - the owners with these beliefs have been largely unsuccessful. But still, these disgruntled and probably misguided owners raise an interesting question - how much of success in fantasy football is likely attributed to luck?

Before I get into the nitty gritty of numbers and analysis in an attempt to answer the question above, I need to provide some background and one critical assumption.

BACKGROUND

I have been commissioner of a fantasy football league - Desert Dogs Chapter, FFL - since 1995. In 1997, our concept grew from a single league to a multi-league association (known as BOHICA), in which multiple leagues are run under the same set of rules, with some owners competing in just one league and others in multiple leagues (often in a combination of solo and different partnerships from league to league). With anywhere from 2-4 leagues each year since 1997, BOHICA has had 29 leagues to date (including DDC '95 and DDC '96).

The 29 BOHICA leagues provide a wealth of data for analysis of the impact of luck on fantasy football outcomes. While scoring rules have been tweaked from year to year, these leagues have all used basic scoring with some performance bonuses, with a 7-man lineup (1 each at QB, RB, WR, TE, K, with two flex for additional RB, WR, or TE; no defense). Four of the leagues had incomplete historical game-by-game data and were excluded from that analysis. One of those four also had unreliable historical end-of-season point totals and this league was excluded from the points vs. season winning percentage analysis. Still, given that DDC '95 was a 10-team league while all the rest were 12-team leagues, and that all leagues played a 14-week regular season schedule, the analysis is based on over 2000 game results (and over 4000 individual team game scores vs. W/L result) and over 300 team seasons of play.

CRITICAL ASSUMPTION

The critical assumption is that in the long run, the ability to produce points in fantasy football is based solely on skill. Let me say that again: In the long run, the ability to produce points in fantasy football is based solely on skill. I realize I may lose some of you here, but stay with me for the defense of that statement. The key part of the statement is "the long run". Clearly week-to-week results contain a great amount of volatility. Even season-to-season will have some volatility due to major injuries to key players, aberrational performances by players (both good and bad) and the like. But in the long run, the skilled player will use the draft, trading, player acquisitions, and week-to-week lineup decisions to consistently put points on the board at a higher rate than the lesser-skilled player.

Consider the analogy of another activity which has taken the world by storm the way fantasy football did a few years earlier - Tournament No-Limit Texas Hold 'Em. On a hand-to-hand basis, there's an incredible amount of luck involved; in fact at that micro level it's virtually all luck once a hand is set to play out and the respective odds are established. Over the course of a tournament, a lot of that luck will even out, but there's still enough luck involved that a few seats at the final table are likely to be surprises. But at the end of a year, or over several years, the top pros and best players are almost always the most successful, while the amateurs may have caught lightning in a bottle once at best. So it is in fantasy football: Anything can happen with each player's performance in a given week, and any head-to-head result is not unusual. Over the course of a full season, the stronger owners will succeed a fairly high percentage of the time (and score points on a more consistent basis), whereas the weaker player's results will be more hit and miss. And over many seasons of play, the strong, average and weak players will separate rather definitively. Looking at the results from nearly 30 seasons of play I will attempt to put some more definitive numbers on the table in support of that assertion.

THE TRUE SOURCE OF LUCK IN FANTASY FOOTBALL

So, if we've taken things like injuries, the draft order draw and week-to-week anomalies in player performance out of the equation that we call "luck", what does constitute luck in fantasy football? Luck as I'm defining it in fantasy football is quite simply the impact of head-to-head scheduling on wins and losses. Anyone who's ever played in a head-to-head format league has had the experience of posting the week's second-highest score, only to lose because the one team that scored more points that week was the opponent. Likewise, we have all lucked out with the next-to-worst score of the week, only to record a win. In a 12-team league, each of these events has a 9% chance of occurring, and clearly represents bad luck and good luck, respectively, in terms of the schedule draw for the week.

WEEK-BY-WEEK STATISTIC AND CORRELATION

The first step in digging into the numbers is to define a statistic for use in comparing game results in terms of points to the win/loss result. The statistic I've chosen is the difference between the league points average for the week and each team's points for the week ('Df S-Wk' in the Excel file, 'Individual Game' tab). After developing this statistic for all 4172 individual team game results from the 25 seasons used in this analysis, we can examine the correlation between this statistic and wins (denoted by a 1) and losses (denoted by a 0), in the column labeled 'W/L'. Here is a summary of each league.

What we find over this large data set is that there is a .5900 correlation between Df S-Wk and W/L. This is a slightly stronger than moderate correlation, with 1 representing a perfect relationship between high points and winning, with 0 representing no relationship whatsoever. This is about what we would expect, as clearly scoring lots of points in a given week (relative to the overall average per team for the week) gives a team a great chance to win, but unless a team is the high scorer for the week, there's no guarantee of a win. On a season-by-season basis the results were fairly consistent ranging from .5065 (DDC '03) on the low end to .6499 (DDC '05) on the high end.

SEASON-BY-SEASON STATISTIC AND CORRELATION

Perhaps a more meaningful correlation is one derived from a comparison of season point totals by team against a won-loss percentage for the season. This is the first level at which our long run definition of luck truly can be seen to play out, and we should see the stronger owners succeed at a higher rate. With 28 seasons of data (one 10-team season and the rest with 12 teams), we have 334 season point totals and corresponding won-lost percentages. The statistic we will use here is similar to the one we used for individual game results, as for each team we will calculate the difference between its season point total and the average season point total for the league in that season (Df Pts-SA in the Excel file, 'Season' tab). We can correlate this statistic with the W/L percentage, from the 'PCT' column in the attachment. Here is a summary of the correlation in each league.

What we find this time is a stronger correlation between season point production and won/lost percentage than we found on the game-by-game analysis - again, what we would expect. This piece of the analysis yielded a correlation of .7594 - a fairly strong correlation between season point production vs. the mean and wins and losses. The season-by-season correlations were more variable, as one would expect with only 10-12 data points per season vs. 140/168. The high water mark was a .9380 (DDC '95), while the low water mark was .4405 (DDC '02). It should be noted that 75% of the seasons had a correlation of .69 or higher, with only 7 of the 28 results below that mark.

MULTIPLE SEASON STATISTIC AND CORRELATION

The final step in this exploration is to look at a group of owners over multiple seasons and see if similar correlations are even stronger when looking at data over multiple seasons. But what statistic will best capture owner performance in terms of scoring points over multiple seasons? The statistic I've chosen, labeled 'Df PG-ASA', reflects each owner's differential of points per game versus the average points per game for the aggregate of the specific leagues in which each owner has played. The 'PG' portion is simple to calculate for each owner - it is merely each owner's average points per game over all games/seasons played. The 'ASA' (or Aggregate Scoring Average for all leagues played) is a bit trickier, as it is different for each owner.

Note that this analysis only looks at the 13 owners who have competed in 10 or more seasons in BOHICA and that league average per team/per game scoring has ranged from a low of 31.01 to a high of 40.27 (based on both scoring rule changes and year-to-year differences), with an overall average of 35.50. Note also that since these are multi-league, multi-year players, they are generally more successful than those who have played less frequently, and thus it's not surprising that 10 of the 13 have a positive 'Df PG-ASA'.

I've also chosen to examine correlations of the 'Df PG-ASA' statistic with two different measures of long-term performance and success: Overall Win-Loss Percentage and percentage of seasons in which each owner qualified for the playoffs. In both cases the correlations are the strongest yet, with a .8947 correlation between 'Df PG-ASA' and W-L Percentage, and a .9151 correlation between 'Df PG-ASA' and playoff qualification percentage. These are significant correlations that demonstrate that consistent scoring of points does in fact translate into more wins and making the playoffs more regularly in the long run. It should be noted that BOHICA rules put six teams into the playoffs with the sixth and final playoff spot going to the yet-unqualified team with the most season points, regardless of W-L record. This somewhat inflates the second correlation, but is a league rule that I highly recommend commissioners to consider, as it rewards performance over a season and minimizes the impact on a bad run of luck with head-to-head scheduling, in an environment where there truly is no defense against an opponent's big week.

CONCLUSIONS AND THOUGHTS

Not even the most hard core (and successful) fantasy football player would argue that luck never comes into play. By its very nature, fantasy football takes a set of events that would otherwise be random and puts order to them; this frequently leads to remarkable, dramatic, and/or lucky results. But the data strongly suggests that there is in fact a strong component of skill involved in fantasy football, and that component becomes more and more of a factor the longer play goes on.

While the numbers generated out of this study are not the definitive answers to the question of how much luck is involved in fantasy football, they will do until something better comes along. So for now, put this statement in your hip pocket for the next time that one of your fantasy football buddies spouts off about how the game is mostly or entirely about luck:

"Luck attributed to head-to-head scheduling accounts for roughly 40 percent in any given week. But the impact of luck from all sources reduces to about 25 percent over the course of an entire season and over many seasons the impact of luck reduces to about 10 percent in determining success or failure in terms of winning and losing, making the playoffs or not."