Jump to content


  • Content Count

  • Joined

  • Last visited

Recent Profile Visitors

105 profile views
  1. Suppose we assume that each weight class is an independent measure of ranking accuracy. How well do the different rankers predict the outcome of a bracket? If we assume that scores are continuous and that errors are randomly distributed, when we can test this using a simple AOV, in R (given that stacked.ching is the long version of the table Ching posted): ching.lm <- lm(Score ~ Source + Weight, data=ching.stacked) anova(ching.lm) ## Analysis of Variance Table ## ## Response: Score ## Df Sum Sq Mean Sq F value Pr(>F) ## Source 8 187.0 23.38 0.7211 0.6722 ## Weight 9 9227.1 1025.23 31.6256 <2e-16 *** ## Residuals 72 2334.1 32.42 ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Given that the data sums of scores, we might use a Poisson model, ching.glm <- glm(Score ~ Source + Weight, family=poisson, data=ching.stacked) anova(ching.glm,test="LRT") ## Analysis of Deviance Table ## ## Model: poisson, link: log ## ## Response: Score ## ## Terms added sequentially (first to last) ## ## ## Df Deviance Resid. Df Resid. Dev Pr(>Chi) ## NULL 89 218.387 ## Source 8 3.283 81 215.104 0.9153 ## Weight 9 173.040 72 42.064 <2e-16 *** ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 We might continue with a categorical analysis of the table, perhaps Fisher's Exact test (here, the data are in original table form) fisher.test(ching.table,simulate.p.value = TRUE) ## ## Fisher's Exact Test for Count Data with simulated p-value (based ## on 2000 replicates) ## ## data: ching.table ## p-value = 0.9975 ## alternative hypothesis: two.sided Finally, if we don't assume any distribution, a rank-based test is appropriate friedman.test(Score ~ Source | Weight, data=ching.stacked) ## ## Friedman rank sum test ## ## data: Score and Source and Weight ## Friedman chi-squared = 3.4866, df = 8, p-value = 0.9002 friedman.test(Score ~ Weight | Source, data=ching.stacked) ## ## Friedman rank sum test ## ## data: Score and Weight and Source ## Friedman chi-squared = 56.718, df = 9, p-value = 5.722e-09 Long and short - there is more variation among scores between weight classes than there is variation among the rankers, and the range of differences among rankers is small relative to the error within weight classes. The difference between 544 and 588 is about 4 points per weight class, while high and low totals in any weight class, say 125 (45-65) varies more than that. There may be ways to further decompose the error in the comparisons to distinguish (at least in a statistical sense) among the rankers, but there's not enough information in Ching's table.
  2. The question that springs to mind - is this a function of (a) the university (i.e. the culture of the AD, how is the wrestling program funded and supported relative to other sports?), (b) the coaches specific to the university over that period (does the coach focus more on dual-meet wins or tournament wins?) (c) the conference (does the conference dual meet schedule provide enough information for seeding; is the conference allocation biased?) (d) the athlete base (is the university drawing from a regional pool that underperforms on a national stage?) (e) just dumb luck? Consistent under- or over-performance, relative to seed or other ranking systems, suggests a knowledge gap, and a knowledge gap may be exploited, if the answer is not (e). So, is your data base sufficiently detailed to test (a)-(d) against (e)? That's the part I'm still working on.
  3. Oh, yes, there are a lot of refinements to be made. I just knocked out a crude approximation based on rough placing. There's probably more information hiding in the brackets.
  4. This is a zero-sum measure - for every match where one wrestling performs better than his seed, the other wrestling performs worse than his seed. So, yes, if every person wrestles exactly to their seed, then all measures would be 0. Whether it's harder to gain a positive or a negative - that is more a matter of context. Over all of the wrestlers in a tournament, it balances out - for every wrestler that has a bad tournament, there will be a different wrestler that has a good tournament. That will always happen, and that adds a certain amount of background noise to the analysis. The trickier part is separating, from the noise, whether a coming from a particular school changes the probability of having a good or bad tournament.
  5. My day job includes a lot of statistics and data science, and I've been toying with predictive analytics in wrestling. Specifically, how well do different ranking systems predict performance? So I just happen to have the 2018 tournament data on hand. I've written code to convert seed to expected placement, and then calculate expected team scores from placement. Then, we calculate a difference between expected score and placement score. I don't have the entire set of match results in a useful format, so the expected and actual scores are estimates - they do *not* include bonus points, and average over different routes to placement. That said, the top and bottom 5 teams, by difference between expected score based on seed and expected score based on actual place: HOF +4.17 LH +2.83 NEB +2.63 KENT +1.69 PRIN +1.60 OKST -1.30 PENN -1.44 LEH -2.55 RID -2.75 MIZZ -3.33 This should be read as the average points scored per wrestler. It is a crude measure - I knocked in out over the course of an hour - but it does tend to support the original post.
  6. Seth was a (mostly) no-go for ASU, up until the Sunday morning weigh-in. He felt good Sunday, he wrestled. He tweaked his back mid-week leading up to OSU, the decision to hold him wasn't made until Thursday or Friday - don't hold me to the exact date. My memory of the timeline is that Hahn had Gross wrestling as of the Wed. coaches lunch; announced he wouldn't be wrestling at a Friday evening alumni gathering.
  • Create New...