Braintree Table Tennis League: 92% in Division Three and the Limits of a Win Percentage

In my notebook there is a line written down from last season: Nolan-Bradford, Braintree division three, one defeat all campaign, and that defeat came against Ben Southgate at 16-14 in the fifth game. I keep the line because it shows how one win percentage can carry two completely different meanings. A game finishing 16-14 is a game where both players had their hands on the edge of equilibrium; a two-point margin sits comfortably inside the noise band of almost every model I have ever built. Yet for the whole of the coming season it will be repeated as destiny, as a rivalry, as a weakness nailed into place. The chart says none of that. The chart says that across an entire division three season, exactly once, a player left a two-point dent. Braintree Table Tennis League is a local competition in Essex, operating under the competition system of Table Tennis England, the national governing body for table tennis in England. The new-season preview published by Table Tennis England on its official media channel focuses on two divisions: division two and division three. There are no ITTF ranking points, no prize money, no international qualification places. The only unit of measurement is win percentage across a fixed fixture list, plus the final league position that decides promotion or relegation. That data structure produces a very specific way of reading. The sample is small. Opponents repeat. A division three player meets the same group of people three or four times in a season. Win percentage here measures consistency against a closed pool of opponents, not absolute standard. Anyone reading 92% as a national-level certificate is reading the table wrong. Line-ups are not fixed either: the preview uses phrases such as on occasions and around half of matches to describe the availability of key players. There is also a mid-season movement window, when a young player can be moved up a division around Christmas. Understand those three features and the rest of the table becomes readable. In division two, Black Notley B are the side placed at the top. Their squad stands on three legs. The first is Neil Freeman, who scored 60 per cent in division one last season and now drops into division two. The second is Rev Matthews with 86 per cent in division two itself. The third is Steve Kerns, a former men's singles champion, available for roughly half of the matches. For me, Freeman is the most interesting variable across both divisions, and the reason lies in the translation problem between divisions. Sixty per cent in division one is not a modest figure. In a higher division the density of strong opponents is greater, the wins are harder, the margin for error smaller. When a player like that drops down a level, the reasonable expectation sits in an upward translation, not in holding 60 per cent flat. How far it translates has to be read from the first five matches. That is why I am not betting on the label team to beat, but on measuring Freeman's translation coefficient across the first six weeks. Matthews runs the other way. Eighty-six per cent in division two is a high number, but it was produced inside the very division he still plays in. There is no translation at all. Its value lies in stability: close to nine wins in ten across a season signals a game with few errors and little dependence on inspiration. The weakness of the three-legged structure is the third leg. Kerns adds depth in the big matches, but a line-up depending on a part-time player loses points in the rounds he misses, and those points tend to fall exactly in the head-to-heads against the second-placed side. Sudbury Strollers are the direct challengers. They finished second last season. Dave Fiddeman scored 92 per cent and John Colvin 75 per cent. Those two numbers are enough to create a genuine threat. But the preview is explicit: their fate depends on who backs them up and how often. I read that sentence as a variable, not as commentary. For a team whose points come from two people, the variance sits in the third and fourth slots. If the third player appears in only half the fixtures, Sudbury are not competing on Fiddeman and Colvin; they are competing on the scheduling sheet. Then there is Fiddeman's 92 per cent. I want to pull it out of its divisional context before assigning it any meaning. A 92 per cent win rate in division three speaks about the gap between Fiddeman and the division three opponent pool, not about his position on any ability spectrum beyond this league. Two-source verification here means checking the win rate against the specific opponent list and against the actual number of matches played. Without those two facts, 92 per cent is just a handsome fraction. Division three tells a different story. Finchingfield B finished second last season, lost Lucien Nolan-Bradford, but compensated with Dave Punt, moving down from division two. This is a familiar local-league transaction: an experienced player goes down, a young player comes up, and the two movements rarely cancel each other out exactly. Punt does not replace Nolan-Bradford. Punt changes the team's points structure: he carries a division two win rate into a division whose average win rate is lower. At Black Notley, the club fielding an entirely new F team says a great deal about its membership base. The new players in that side impressed on debut. I do not have enough data to quantify the impression: no win percentage, no named opponents, no game scores. So I file it in the pending column and say plainly that I do not yet have enough evidence. The most interesting part of the preview lies with the junior cohort. Ethan Collins is twelve years old and already has three cadets titles and one junior boys title. At Rayne D, Sai Suresh, fourteen, and Aryaman Singh, thirteen, are about to make their first serious adult-league appearances under the eye of league coach Keith Martin. JJ Calisin, eighteen, is scheduled to move up to division one at Christmas. I remember how I learned to read numbers like these. In 2026 I sat down with ten Hanoi FC matches in the V-League and counted by hand. No API, no data provider. I recorded passes, recoveries in the opposition third, and pass completion under pressure. The result was a defensive midfielder wearing number 8, Nguyen Van Dung, with a PPDA of 9.2, higher than the rest of the squad. The media never mentioned him, because he did not score. My two-thousand-word analysis was shared widely and reached fifteen thousand views. The lesson was not the 9.2. The lesson was that I had built the dataset myself, so I knew exactly where it was weak. The 2026 World Cup taught me the reverse. I built a knockout-stage prediction model using expected goals. I picked Uruguay to beat France in the quarter-final, on the strength of defensive form. France won 2-0 with an expected-goals gap of 2.8 against 0.4, and shots inside the box of nine against four. I was wrong because I read the result instead of the process. Expected goals does not judge a shot; it only illuminates the football people refuse to look at. I spent three weeks re-watching twelve matches, and from then on I set one rule for myself: every judgement must stand on at least two independent sources. Applying that rule to Braintree, I see three signals worth tracking. Freeman's division-translation coefficient. The actual appearance count of Sudbury's third player. And the results of Collins and Calisin against adult opponents, because that is where a junior cohort turns from potential into data. There is a temptation I want to put on the table: treating team to beat as a conclusion. It is only a label assigned by the preview writer, based on a side that has just been relegated and owns three strong players. The label may be right, but it has not been tested. And there is a subtler error. Win percentages in local leagues do not move with the player; they move with the division. The same person, scoring 92 per cent in division three, can fall to 60 per cent in division two without playing even slightly worse. The denominator changes, the numerator stays, the ratio changes. Anyone comparing Fiddeman with Matthews through 92 per cent and 86 per cent is comparing two fractions with different denominators. The Nolan-Bradford and Southgate story works the same way. A 16-14 defeat in the fifth game produces one data point; it does not produce a rivalry. To reach a conclusion I need at least five meetings, and I need to know whether the two will still meet in the same division. A divisional move does not buy a promotion; it buys the probability of a trembling future, and that probability is only settled after the ball has bounced. It is also worth stating the limits of this piece plainly. I have no point-by-point data, no serve statistics, no information on blades or rubbers. Everything above stands on four kinds of facts: win percentage by division, match availability, player age, and last season's finishing position. Those four are enough to rank what to watch. They are not enough to predict the final table. The first six weeks will answer most of the question. I will read Freeman's division two results as a measurement of translation, count how often Sudbury's third player appears, and log every game Collins plays against opponents over twenty. One defeat is a riddle already solved, but hundreds of riddles still sit quietly beneath the win-percentage table. An empty hall does not create ghosts; it creates the cleanest data a monk could ever dream of. And Braintree, with its half-fixed line-ups and its mid-season movement windows, is exactly that kind of empty hall.