RealWins tells you how many wins a team really has; that is, its wins are weighted for significance and the total gives a better idea of how much a team has accomplished this season.
It also has the effect of telling you when a team is a lock for the NCAA tournament. No team with 30 wins is going to miss the tournament; thus, when a team has a score of 30 in "RealWins" then they are tournament bound, even if they lose the rest of their games.
Using this idea, we can tell you how close your team is to being a lock for the tournament, or how far. Note that this is not the percentage chance of making the tournament, but the chance of being a lock. If a team has an 80% rating come mid-March, it's very likely that they miss the tournament. 90%? They're a bubble team. 70% or lower? Basically no hope.
At mid-season, 70% is a great ranking. The season's only half way done, and the team is already 70% of the way to being an NCAA lock. That's pretty good.
A team's RealWins increase throughout the season as they win more games. It also increases as their previous victims win more games. That's because RealWins goes deeper than a team's own wins. It looks at a team's opponents' wins, and by cycling through again, captures an opponents' opponents' wins. In this way it's similar to the RPI, but with a few key differences:
- Unlike the RPI, it does matter who you beat, not just who you played
- The RPI is an average of winning percentages and fluctuates up and down; RealWins is an aggregate measure and never goes down
- RealWins looks only at wins; it doesn't care about losses at all.
The last point is interesting. It's necessary in order to make RealWins a measure of progress. After a team makes it to a 60% lock chance, they might remain there but never go backwards. It's interesting that it makes for such a good tournament field predictor without even considering losses.
Another thing it doesn't consider is home or road court. Every win is the same, home or away. This is very contrary to what the selection committee claims is important. But when tested, if we include that information, the predictive value plummets.
Imagine if the NCAA tournament selection committee were to announce that they don't take into account a team's losses, or where games took place!
Losses only hurt a team in one way: they miss out on the points they could have gotten for a win. And including a bonus for road wins—and less credit for home wins—just lets a few more non-Major conference teams in, teams that didn't really make it. It's non-Major conference teams that are playing on the road more often, and they'd benefit the most from road games being worth more.
Adding losses does improve the accuracy a tiny bit, but that wouldn't allow for the ranking to be a season-long progression forward. Maybe we'll add losses to a separate ranking, called "Real Record." Because you can't have too many ranking systems.
Comments