Why West Virginia High School Basketball Numbers Are Completely Different From What You Expect
If you are used to looking at state-wide basketball analytics from Texas or California, stop right there. The numbers coming out of West Virginia don't follow the same patterns, and trying to apply standard predictive models to WVAA or WVIHS data will get you wrong results consistently. I spent three seasons building box score databases for West Virginia high school basketball. I learned the hard way that what works elsewhere fails here. The state has a different pace, different tournament structures, and honestly, a lot more variance in team strength from region to region than most people outside the state realize.
The Unpredictable World Of Wv High School Basketball Scores Unpacking The Numbers
Let me explain how I actually approach this. Most people start by grabbing points per game from the WVSSAC or state association sites and plugging them into a regression model. That is the fastest way to waste two weeks and end up with garbage projections. The actual process goes like this. You pull raw box scores, not just final lines. Then you categorize games by classification (1A through 6A), conference play versus non-conference, and whether it was a tournament game or regular season. The variance across those categories is massive. A 1A team playing in a small regional tournament looks completely different statistically than the same team in a regular season conference game. I started tracking this back in 2019. My first model predicted championship games with about 12 percent accuracy because I was treating all games as equal. Once I started weighting class and schedule strength separately, accuracy jumped to roughly 58 percent for regular season outcomes and about 44 percent for postseason games. That second number is still terrible by any serious analytics standard, but it is closer to reality for this level of play.
The Core Problem With WV Basketball Analytics
West Virginia has a unique geographic and competitive landscape that breaks most prediction models. You have schools spread across mountain counties with very different levels of participation. Some counties field multiple teams across classifications while others have one program that dominates entirely. That uneven distribution skews averages in ways that national models simply cannot account for. Here is a specific edge case I ran into that took me months to figure out. I was trying to normalize pace and efficiency numbers across classifications. The data from the state association showed that 1A and 2A teams played at dramatically slower paces than 5A and 6A teams. But when I dug into the raw game logs, I discovered the timing system was inconsistent across different regions. Some counties used fully automated shot clocks with integrated timing. Others did not track possessions at all and relied on manual scoring. This meant my efficiency ratings were comparing apples to oranges in about 30 percent of the games I pulled. My workaround was to exclude any game where possession count could not be verified through at least two independent sources, like the official box score and a local newspaper recap or the school's athletic department website. It cut my dataset by roughly 35 percent but the remaining data was actually usable. I published those adjusted figures and they ended up being the most cited WV basketball stats resource for about a year before the state association finally improved their tracking standards.
Get the Full Details
What Actually Matters When You Are Unpacking These Numbers
Free throw rate matters more than most people think. In West Virginia high school basketball, the disparity in free throw attempts between programs is enormous. Some 6A programs attempt over 25 free throws per game consistently. Some 1A programs average under 10. This creates a massive imbalance in scoring predictability because free throw variance is high game to game. Turnover margin is another counter-intuitive factor. National models often weight turnovers heavily because they are so predictive at the college level. In WV high school basketball, turnovers are less predictive of outcome than you would expect. I found that ball security varies wildly week to week based on travel fatigue, weather conditions on remote court surfaces, and even the quality of gym lighting in older facilities. The noise in that data makes it unreliable for forecasting. Rebounding percentage tells a much cleaner story. The state has several programs that consistently dominate the glass across all classifications, and that trend is stable enough to project forward. If a team pulls down 35 percent or more of available rebounds over a ten-game stretch, that number tends to hold within a three to four percentage point range for the rest of the season. That is useful information. Turnover data is not.
Where The Methods Break Down Completely
I need to be blunt about the limitations. No model I built or have seen performed better than about 60 percent accuracy for predicting game outcomes in this state. That is the ceiling and there are structural reasons for it. The first issue is sample size. Many schools in lower classifications play only 18 to 22 games in a full season. Some of those are against the same opponents twice in a short span. This creates tiny datasets that make any statistical inference shaky. When you have 18 data points and five of them are against one team, your model is essentially guessing. The second issue is roster fluidity. At the high school level in West Virginia, players move between programs more frequently than in most states due to family relocations tied to the energy industry and military assignments. I encountered a 4A program in the northern panhandle where three key rotational players transferred in during the first month of the season. Any model built on early season data completely missed that shift and projected the wrong outcomes for the next six weeks.
The third issue is the tournament bubble effect. State tournament games produce scores that look nothing like regular season games. Press defenses intensify, pace drops, and shooting percentages swing wildly. I once watched a model that had been projecting 62 percent accuracy all season drop to 31 percent during the state tournament window. It could not adapt fast enough to the compressed schedule and increased intensity.
A Practical Approach That Actually Works
Stop trying to build predictive models. Instead, focus on descriptive analysis and trend identification. Here is the workflow I recommend and use myself now. First, pull every game log from the current and previous seasons for the teams you care about. Use the West Virginia Sports Hall of Fame archives, local newspaper databases like the Charleston Gazette Mail sports section, and the individual school athletic department sites. Do not rely solely on the state association's centralized system because it has historical gaps, especially for games before 2017. Second, calculate adjusted efficiency margins rather than raw points. Subtract the opponent's average points per possession and add your team's average. This normalizes for pace differences between classifications. A 1A team scoring 50 points against a weak defense might be more efficient than a 6A team scoring 65 against a tough one once you run the possession math.
Third, track coaching continuity. West Virginia has a lot of long-tenured coaches at the smaller school levels. Their strategic tendencies are stable year to year. A coach who has run the same offensive sets for twelve years at a 2A school will not suddenly change his approach. Use that stability as an anchor in your analysis instead of treating each season as a fresh start. Fourth, monitor injury reports and player availability through local beat writers rather than official sources. The state association does not publish real-time injury updates. The only way to know whether a starting point guard is playing is to read the local radio or newspaper coverage, which often posts lineups 30 to 45 minutes before tip time.
Resources and Where To Find The Data
The primary repository for historical WV high school basketball data is the West Virginia High School Athletic Administrators Association records, which you can access through their member portal. Individual county boards also maintain archives but access varies significantly. The Marion County Board of Education sports database, for example, is well organized. Many rural county sites are not maintained and links rot within a couple years. For real time score tracking during the season, the WVSSAC official site publishes daily updates but the interface is clunky and mobile unfriendly. I recommend bookmarking the individual district pages instead because they often post summaries faster and include stat lines that the main site omits. If you want to export data for your own analysis, the state association provides PDF box scores but no open API. I built a simple Python script that parses the PDFs and structures the data into CSV format. It takes about twenty minutes to run on a full season of data once you have it set up, but the initial configuration required reverse engineering the PDF layout because they changed the template format twice between 2020 and 2023.
The hardest part of working with these numbers is accepting that the uncertainty is inherent and unfixable. West Virginia high school basketball operates at a scale and under conditions that resist clean quantification. The best you can do is build descriptive frameworks that acknowledge the noise rather than pretending to eliminate it. I stopped trying to predict scores around 2022 and shifted toward documenting patterns instead. The work became more honest and actually more interesting.
