I apologize for the length of this fanpost. I wanted to do a thorough job of explaining my results. Also, I have some time now that I am retired (for now) and I thought that I would give this subject some data-driven scrutiny. I was going to wait on posting this until a little after Signing Day, but since there isn't a lot to talk about with Signing Day, I thought that this would provide something for discussion.

This time of year there are usually articles about how recruiting is so important and then endless debates about the value of recruiting. Few would argue that recruiting is not important at all. But we all know that it is not absolutely required for winning games (and even division titles) as teams like WSU and Utah have proven over the last few years. The question then becomes: How important is recruiting? This fanpost will look at just the value of recruiting with respect to wins and losses. I am not going to look at how recruiting relates to other outcomes like division titles or national championships.

Several years ago I read an article talking about recruiting which divided up the schools into recruiting tiers. Teams like Alabama, Clemson, Ohio State, etc. were in the top tier. Teams like UW were in the second tier. As I remember, WSU was a 3rd tier team, and I think that there was a tier 4 for a lot of what is now the Group-of-Five teams. The article looked at the win percentages for teams competing against teams in the same tier or against teams in other tiers. If I remember, for example, there was a 50-50 chance of winning against a team in the same tier, but that went up if you were competing against a team in a lower tier and went down if you were competing against a team in a higher tier. Unfortunately, I can no longer find that article. Even if I could, the article was several years old and looked at recruiting tiers in historical terms so teams that were recruiting better or worse, especially since then, were not up to date.

What I decided to do was to update that kind of analysis, but use roster talent instead of historical recruiting tiers. To do that, I used the 247Sports' "Composite Team Talent" rating which looks at the talent on each team's roster. (It isn't every team as I'll talk about later.) They have talent ratings based on the players on the roster and how they were rated (Composite Rating) during recruiting. There are, of course, obvious flaws in using these numbers, not the least of which is that they don't take into account which of the players actually plays (doesn't redshirt or isn't injured). But, the data was easily available and I decided to look at the data over multiple years so some of those problems would be minimized.

Methodology

For my analysis I looked at all football games played by FBS schools over the years 2016-2019. (247Sports has data only back to 2015.) For each game I looked at the talent rating for the teams involved. My hypothesis was that if recruiting talent was important, the team with the most talent would be winning that game.

As an example of my analysis, UW played its first game of 2019 against Eastern Washington. 247Sports had a Composite Team Talent value of 801 for UW in 2019. This was based on all of the players on the roster for that year. Eastern Washington's was about 15. UW had far more talent than Eastern Washington (by a lot) and won the game easily (47-14), so that was consistent with the hypothesis.

I repeated that calculation for each of the over 3500 games over the four years I studied. I looked at the results for each year separately, then combined data from each of the years together.

As I mentioned above, some teams were not listed for some years in the 247Sports Team Talent rankings. As far as I can tell, all of the FBS schools are listed except one. A few of the FCS ones aren't listed for some years; they may have a value one year, but not the next. The one FBS school that was missing a value was UAB, which had no value for 2016 and was over 300 in the other years. For them I used an average of their 2015 and 2017 values to get their value for 2016. For the FCS schools, if there was no value in the 247Sports rating, I assumed a value of 0. This may seem harsh, but most of those FCS teams had values of less than 50 when they were listed, while most FBS schools had values of well over 200. Again, there was an exception. Sam Houston State which had no value for 2017, but was over 100 for the other 3 years. For them, I averaged the 2018 and 2016 values to get their value for 2017. (I am not sure why there were some schools with no values; fortunately there were not many of any importance.)

Initial Results

Here are the results for each of the 4 years. I use the term underdog here to indicate the team that had a lower Composite Talent Value and favorite for the team that had the higher Composite Talent value.

And here is the results for all four years combined.

The first thing that stands out to me is how consistent the data is across the years; just 3 percentage points different. That tells me that factors such as injuries and redshirts aren't a big factor when looking across all of college football.

The second thing to note is the actual value for win percentage. In just over 2/3rds of the games, the team with ‘more talent' won the game. Some people will point to that and say "see, more talent wins more games". Others may see it and say the opposite "more talent doesn't always lead to wins". I decided to probe deeper.

A Second Factor

If I asked college football fans (and even NFL fans) to predict the outcome of a game between two teams that they did not know much about, what factor do you think they would use to help decide which team to pick? I think for a lot of fans, that would be "which is the home team". So I decided to look through all of the games to determine the win percentage by the home team.

For this analysis I eliminated all of the bowl games, even though a few of them (like last year's Hawai'i Bowl between Hawai'i and BYU) could be considered a home game by one team. There were also some games during the regular seasons that were not played in the home stadium of either team, mostly some non-conference games. I did not eliminate them (it would have been a lot of work). Fortunately there weren't very many of those games compared to all other games, so I don't feel that it impacts this analysis.

For all of the games that I looked at, the home team won in 65% of the games. This was also very consistent over the 4 years worth of data.

It will be interesting to see if the upward trend continues, but for now I'm assuming that this data is consistent year-over-year.

This may seem like an obvious piece of data: the home team has an advantage that improves their chances of winning. But note that the win % by the home team is not much different than that of the team with ‘more talent' (65% vs 68%). It could be argued, therefore, that a talent advantage does not help you to win by much more than just being the home team.

Combining Factors

The two factors that I've showed so far are not mutually exclusive. A home team, for example, can have more or less talent than the team that it is competing against. So how do these factors combine? That is what I looked at next.

In the table below I've compiled the data from the four years and broken out the percentage of games won by each group. (Again, I've eliminated bowl games but left in a few games played at neither team's home stadium.)

The way to read this data is that the percentage in each box indicates the percentage of games won by teams with both factors (row and column). So, 78% of the time that the home team had ‘more talent', they won. That means that 22% of the time the visiting team with ‘less talent' won; that shows up in the opposite corner of the table. When the visiting team had ‘more talent', they won 56% of the time, which means a home team with ‘less talent' won 44% of the time.

The first thing that this shows is the significant advantage that the home team has when it has more talent. That should make sense given the data above; the factors combine to produce an advantage greater than either factor alone.

The next thing that stood out to me is that the team with more talent still has an advantage even when they are the visiting team, but not by a lot. The talent advantage is reduced by being on the road, but it is still present. And, as we saw above, the talent advantage is more than the home field advantage, so this is consistent.

Talent Tiers

So far in my analysis I just looked at which team had the higher talent value. But, the talent differences between teams can be along a range. For example, the talent difference between UW and Eastern Washington is significantly more than between UW and USC. I would expect that the percentage of wins by the team with more talent would increase as the difference in talent between the teams increases.

The talent values from 247Sports go as high as 997 (Alabama in 2017), and as low as under 10 (when there was a value). The talent levels were split pretty clearly in tiers between the Power-5 teams (at the top), then the Group-of-5 teams, then the FCS teams. There were some exceptions, for example, like Boise State having a having slightly more talent than WSU. Almost all of the Power-5 teams had values over 500. (Kansas had 491 one year.) The Group-of-5 teams were generally between 150 and 600. And the FCS schools were almost all below 150.

There are a lot of ways I could have divided the differences; I chose a simple method using increments of 100 up to 400. Even though I could have kept going, I found that the number of games where the difference was more than 400 was so small that it didn't seem worthwhile to continue. So I ended up with 5 categories:

• A difference of less than 100
• A difference of between 100 and 200
• A difference of between 200 and 300
• A difference of between 300 and 400
• A difference of greater than 400

To give you an idea of how those relate, in 2019 there was a difference of less than 100 between UW (801) and Stanford (788). The difference between UW and Utah (801-620) was between 100 and 200. The difference between UW and Arizona (801-570) was between 200 and 300. The difference between UW and Hawai'i (801-421) was between 300 and 400. And as mentioned above, the difference between UW and Eastern Washington (801-15) was greater than 400.

The table below shows the number of games and the win percentage in each group for each of the four years plus the totals for all four years.

I was encouraged that the data again was consistent across all four years. The exceptions were in the 200-300 and 300-400 ranges which had some of the fewest games. There was not as much difference between the 100-200 and the 200-300 range than I expected, especially in 2016. And the win rate even went down slightly in 2017 in the 200-300 range compared to the 100-200 range. Again, I think that looking over several years helps even out the data because there are not a lot of games in some of those groups. Having a win percentage of 95% for a talent difference of greater than 400 also validates for me that there was no need to add categories past that.

As expected, the win percentage does indeed go up as the difference in talent levels increases, and at a fairly consistent (almost linear) rate. There is one thing that is implied in the data that I wanted to point out. If a team has less talent, then their win percentage will be 100% minus the value in the table. For example, the win percentage by teams with less talent and a talent difference of less than 100 will be 43% (100-57), and the win percentage for teams with less talent and a talent difference of greater than 400 is just 5% (100-95). If someone wanted, they could use that to determine the win probability for any team in any game just by looking at the talent differences.

Breaking it Down Further

It should be obvious to many fans of college football that many of the games in the "Greater than 400" category are the ‘cupcake games', like games between Power-5 and Group-of-5 teams or between FBS schools and FCS schools. Those are almost always played at the home stadium of the FBS team. The question I then wanted to look at was whether those kinds of games (home games against ‘cupcakes') were skewing the data. To look at that I divided the games in each category between games where the home team and the visiting team had more talent to get a more complete picture.

I've included the data from all games as a reference, although I've again taken out the bowl games. This data probably takes a little more explanation than the previous data. Here's an example of how to interpret this for each talent-difference category (going from right to left):

• If the talent difference between the two teams is less than (or equal to) 100 (top row of data), then the team with more talent wins 57% of the time (right-most column).
• If the talent difference between the two teams is less than (or equal to) 100 (top row of data) and the visiting team is the one with the talent advantage, they win 51% of the time (third column from the right).
• If the talent difference between the two teams is less than (or equal to) 100 (top row of data) and the home team is the one with the talent advantage, they win 64% of the time (third column from the left).

The first thing to point out is the winning percentage by the visiting team when it has a talent difference of greater than 400. The table shows that there is a win percentage of 43%. However, that is based on just 7 games over the 4 years, so I would not consider it to be a valid data point-that just isn't enough games to draw a conclusion, especially compared to the rest of the data. (That's why I included the ‘*' next to it.) This just points out that teams with more talent seldom go on the road to play teams with much less talent.

When the home team has more talent, the win percentage goes up as the talent difference increases, just like we saw with all games. And because we know that the home team has an advantage anyway, the win percentage by the home team is higher in each category than for all games (comparing columns 3 and 7). This is as expected.

The surprises to me were when the visiting team had more talent. If the talent difference was less than 100, the winning percentage was just 51%. The win percentage by the visiting team did increase as the talent difference increased, but not by a similar rates compared to being the home team. There was a surprisingly big jump between the ‘less than 100' category and the ‘100 to 200' category (51% to 62%), but then the increases were much smaller-almost leveling out. (Again, I'm not going to consider the category of ‘greater than 400'). I'm not sure about the reason for this other than to note that the number of games in the ‘less than 100' category is more than the other 4 categories combined.

Talent Changes

As part of my analysis I had to gather the talent rating for each team for each year. I was able to tabulate the data to see which teams have been increasing their talent over that time period, and which teams have been losing talent. There are well over 200 teams that I looked at, so I don't want to bore everyone by going through all of that data. There are also some of the new Group-of-5 teams that increased talent level, for example, as they moved from the FCS to the FBS level. I decided to do is to pull out some of the Power-5 teams that improved the most and the least.

Illinois is a bit of a surprise at the top. Lovie Smith must be doing a good job of recruiting. Penn State next makes sense as they build back up under Franklin after the sanctions. It should not be a surprise to see UW up near the top as well. Two Pac12 schools near the bottom is not a good look. And Mississippi being at the bottom makes sense because of some of the controversy regarding recruiting there; we'll see if Kiffin can reverse that.

What Does This All Mean?

I'll let people draw their own conclusions from the data because, as I said above, some people will probably use this data to reinforce their beliefs: like talent is what wins games, or talent alone won't win games.

For me, the data was very clear and consistent up until I broke down the data to both home and visitors for each of the talent difference categories. I expect that a part of the lack of consistency in that table had to do with having fewer games in some of those categories. There could also be other factors (like coaching) that confound the data a bit, especially when looking at fewer games.

Even so, I think that there are a few conclusions.

First is that recruiting talent (as measured by the 247Sports Composite Team Talent ratings) is one of the factors that does help to win football games; and the more the talent advantage is, the more likely the team with the more talent will win. It is not, however, the only factor. Home field advantage is also a factor which increases the chance to win. And these two factors combine, with the talent advantage slightly overcoming the home field advantage.

There are other factors as well, but this analysis did not attempt to quantify or even identify what those other factors are, or whether they are more or less significant than talent and home field advantages. Some of these factors could be coaching, weather, and probably experience-especially at the QB position.

What I also see from the data is that while talent can improve the odds of winning, it doesn't guarantee wins, especially if the talent levels are close. UW has beaten teams with more talent, like USC last year (USC's talent was #4 in the country last year). UW also lost last year to less talented teams like Cal, Utah, and Colorado, and two of those games were at home. Just having a better scheme or game plan can also make a difference as well-like when Air Force beat WSU in their bowl game last year despite having a greater than 400 point difference in talent.

Ideally, I'd like to look at the impact of coaching on win percentage to see if that has more or less of an impact compared to just recruiting. That gets difficult with many coaches only staying at a program for a few years, so I'd be dealing with small samples of data. For now I'm going to believe that coaching also has a significant impact on the outcome of games. Whether that is more or less than talent is something that can and will continue to be debated.

I expect that sometime in September of this year is when 247Sports will update their Composite Team Talent for 2020. (They wait for all of the freshmen to arrive and see if there are any late transfers.) When it does come out, I expect that with the way UW has been improving their recruiting, UW may have a talent disadvantage to just three teams on this year's schedule: Michigan, Oregon, and USC. But, I expect UW will be less than 100 points from each of those teams. And I also expect that UW will be less than 200 points from even the top schools like Alabama and Ohio State. For now I'm going to hope that Lake can at least win those games where they have the talent advantage, and maybe pull off an upset (or two!) in the games where they don't. In 2016 won all of the games where it had a talent advantage; they also had 3 games where they had a talent disadvantage and won two of those three games (Stanford, Arizona State, and USC). If Lake can repeat that kind of performance, it should give them the possibility of double-digit wins this year.