…Please stick around UNG. I like typing your municipality. ]]>

Poor wording and I can’t edit.. whoops

]]>Aldridge #29, Kent State

]]>A new team’s rating should be considered provisional for about 6 games, playing against different teams so it can adjust from the initial 40. Still a couple more to go for both GA teams. VCU only gained 1.336 on the exchange, 75th of 220 technical upsets and about 66th percentile. Within the one standard deviation of upset exchanges.

]]>Experience is definitely needed. ]]>

1. Head-to-Head

2. Highest cumulative point differential (3 point differential max/game)

4. Least points against

4. Most points for (max of 5/game unless diferential is <4)

5. Coin Flip ]]>

2012-04-14 Nationals 2012: Saturday Kent def WKU 3 2

2012-10-14 Round Robin Kent def SVSU 2 1

2012-10-21 UMD Tournament Kent def JMU 3 2

2014-01-25 Kentucky Invite Kent def UK 2 1

2014-02-15 BEAST III Kent def JMU 3 2

2014-11-09 Buckeye Invite Kent def OSU 3 2

2015-01-31 UK Invitational UK def Kent 3 2

2015-04-11 Nationals 2015 MSU def Kent 3 2

2015-10-17 Pumpkinfest II Kent def Towson 3 2 ]]>

Excellent!

]]>We could definitely determine how likely it is for one team to win against another team, but I do think the issue is the Gonzalez system is designed to allow a certain amount of technical upsets. These allow teams to advance or fall and their respective ratings move to a better representation of their strength.

About 80% of games are predicted correctly in the system, but that isn’t 95% confidence. But it might be possible to tie this 80% confidence into a deeper calculation of the technical upsets that have already occurred. The Mean upset exchange is 1.273 and the 95% confidence would include exchanges to 1.707. There’s only 10 games that are greater than 1.707. Ten games of 198 upsets equals 5%, but that’s out of 1002 ranked matches overall (0.1%). Haven’t figured it out yet.

The good news is we have regular games coming in and the data is always growing. Larger population allows us to be more confident with our confidences.

]]>I think the crazy part of all of this is that I actually understand what you’re saying and can translate it into everyday language. You’re using a lot of jargon that utilize probability, statistics, and operation research principles aren’t you? I was wondering if it would be possible for you to determine which teams would win within each given tournament if you have to do it with a 95% confidence with one standard deviation from the norm.

]]>Let’s look at the Upsets of the day. There are just 3 technical upsets out of the 11 played in Akron.

a. VCU def Akron 2-1 / 1.250, #90 of 205 upsets of 1016 ranked matches

b. VCU def Kent 2-1 OT / 0.809 actual / 1.618 normalized, #16 of 205 upsets of 1016 ranked matches

c. WKU def Akron 4-2 / 1.123, #141 of 205 upsets of 1016 ranked matches

The mean upset exchange hovers around 1.273, so matches [a] and [c] are within the standard deviation and the great majority of technical upsets will fall around this exchange (technically ~68%). However, match [b] ranks in the 92 percentile in upset games, so we might consider this a big upset for VCU. It does rank 4th in greatest OT upsets.

Taking a normalized exchange for an overtime match and comparing it to a normal match, or even a double worth Nationals match, is a bit subjective. These might not always be the most influential upsets in terms of the overall effect of the team’s rating. But the normalized exchange might help compare these different classes of result.

OT has always been a great indicator of equitable team strength for me, personally. I was super surprised to enter four overtimes into the records, but it seems we get more OTs whenever we put together a schedule that focuses on good even matchups. I had the pleasure to help out Akron in scheduling competitors here, where I focused on getting even matches first, then unique matches we might not see often. 4 in 11 games is 36%, impressive compared to other occurrence percentages. We do a similar thing for Nationals scheduling, where OT occurrence is a small boost above average at 7.53% of total games played. The Records (not including this weekend’s games) have OT appearing in 6.89% of total games played.

]]>