EXPLANATION OF DATA AND
Kentucky Takes LSU by 1 Point in OT To Advance To Finals Against Florida
Today's SEC Semi-finals set up a rematch of last Sunday's game with Florida. Kentucky and Florida split their regular season meetings, each winning on their home floor. Tomorrow, these teams will determine whether Kentucky will secure a #1 seed in the NCAA Tournament that begins next week or whether Florida will finally take its first ever SEC Tournament title to Gainesville. How could any fan ask for anything more?
After moving out to a 6 point lead, 20 -14 midway through the first half, Kentucky stuttered down the stretch of the half to go to the locker room down one, 31 - 32. As the teams came out of the gates for the second half, LSU scored the first 4 points, up 5. Time out Kentucky, and a new five players enter from stage right. What do these guys do? Nothing much. They outscore LSU 9 - 2 to convert a 5 point deficit to a 4 point lead, and all 9 Kentucky points are attributed to freshman Crawford.
LSU outrebounded Kentucky tonight by 7, as Kentucky's rebounding woes continued. Kentucky grabbed 13 offensive rebounds and LSU pulled down 17 offensive rebounds. Kentucky had a total of 89 possessions, including 11 second chance possessions, while LSU had 93 possessions, 17 from offensive rebounds. Kentucky's offensive efficiency for the game was an impressive 0.908 ppp, slightly below their season average and almost identical to the 0.905 ppp I predicted against the LSU defense. Kentucky's defense held LSU to 0.847 ppp, slightly below the 0.862 ppp I predicted.
Based on Season performance of both teams, as well as their schedule strength for this year, I predicted a 3 point Kentucky win, 72 - 69.
Click Here To View Graphs
There are four sets of graphs, each set showing data for the current season for all games played and data for last season.
Graph Set 1: Average Offensive and Defensive Efficiency, both in terms of points per possession. The distance between the offensive and defensive data is the average Net Game Efficiency [See Graph Set 2]. A second set of data is also included, the five game running average for offensive and defensive efficiency. By comparing the slope of the average data and the position of the 5 game averages relative to the season average, one can assess the current state of play.
Graph Set 2: Net Game Efficiency after each game of the current and past season. This graph provides important information about the season. Net Game Efficiency values above 0.25 ppp have correlated to NCAA final four quality seasons. Values between 0.10 and 0.25 ppp indicate successful seasons, and probable post season play. Values between -0.10 and 0.10 ppp are indicate of average teams. This graph shows instantly how Kentucky is doing at any time during the season.
Graph Set 3: The Power Rating is the ratio of a team's offensive and defensive efficiencies. Each game has a unique power rating as well. This bar graph shows the distribution of game power ratings. Each bar represents a Power Rating Increment of 0.1, e.g. .8 to .89 is a single bar as is 1.10 to 1.19. Notice for 2003-04 how the data forms a bell shaped distribution around its mean value. The data for each season can be expressed in its statistical terms, a mean, standard deviation, etc. Certainly, a season with an average power rating of 1.25 with a standard deviation of .20 is better than seasons having a mean of 1.10 with a standard deviation of 0.20 or a season having a mean of 125 and a standard deviation of 0.35. Even after three games, a trend is emerging on this graph. It appears that this Kentucky team will be more powerful than its predecessor.
Graph Set 4: Another bar graph of game data. This time it is concurrent plots for offensive [foreground] and defensive [background] efficiencies. Just as the power forms a classic bell shape, so do each of these data sets. The separation of the two sets is important to winning percentage. The greater the separation, the higher the winning percentage. The separation is the Net Game Efficiency. Looking at the difference in this manner provides some insight about why even the very good teams sometimes lose to weaker teams. As this data shows, every team will experience variable performance, offensively and defensively, game to game. These variations can be represented statistically just like the power data can, with means, etc. However, unless the separation between the offensive and defensive data is so large [over the combined standard deviations] that there is no overlap of the two data set, there is always some probability that the better team will lose to the weaker team. The smaller the overlap, the less likely it is that the better team will lose. The greater the overlap, the higher the percentage of games the team will lose. When the data substantially overlap, the team will win and lose about the same number of games.