I was reading about college squash. It's not a NCAA sport, not enough schools play it, and it's almost entirely Northeastern.
Anyway, no doubles. It's just 9 singles matches.
This was interesting to me, mostly because I was trying to picture what a 9th singles player must be like. It's not that weird, we have singles ladders anyway, and when we're playing someone not that good, we will often take the kid who is somewhere in that part of the ladder and they play singles anyway.
But from a depth perspective, I thought it was an interesting proposition. It's a tennis sim just for me, realism matters to me from the perspective of match play and results, but besides that, I had to ask "if there were no constraints like time, amount of courts, budget, and NCAA rules, what would the ideal dual match format look like?"
That made me consider several things: 1) changng the formats per divsion. D1 having the "doubles point" makes sense, but at the lower levels, there's no reason not to count the doubles matches across the board, which mirrors real life to some degree anyway. So we're making that switch.
But then things got more radical.
D1: 10 singles + 5 doubles = 11 points
D2-D3: 8 singles + 3 doubles = 11 points
D4: 10 singles + 3 doubles = 13 points
Realistically, this doesn't dramatically change the size of rosters, you'll need a few backups in case of injury, but eams already were carrying a surplus anyway, especially at the D3/D4 level, so now it just means that instead of a bunch of players not playing, I get to see everyone actually playing in matches that have impact on the outcomes of the overall dual.
I don't know that it will overly impact who wins the national title, in the sense that, the rich will continue to get richer probably, but it will be especially fun at the lower levels to see how the dice rolls change how close matches are.
In real life, the thought of sitting through that many matches feels insane, but...it's a personnel and court-space thing, once you get past that, it'd actually be kind of fun to think about how many more people would see the court.
My next iteration of the save will be a refreshed intake with the larger rosters and seeing how it all plays out, and how much it distorts or expands the quality of the various levels (with the idea that it'd almost certainly create a drag on D3 and D4 rosters since now we're essentially creating a few more spots for D1/D2 kids that otherwise were going to sit and maybe transfer down, now they won't leave at all.
Measured effects, same season simulated under both formats with the same seed: the top five D1 teams' win rate moved from 93.3% to 88.9%; overall win-percentage spread and upset rates were unchanged within noise; the consolidated doubles point decides 10% of D1 duals (12.2% under the old format); doubles decides 18.5% of D2 duals (14.6% under the old format).
Outcomes decided by the full point engine
Match outcomes are now decided by the per-point engine at full fidelity by default. Previously the fast game-level model decided outcomes and the point engine only reconstructed statistics afterward. The fast model remains available as a speed option on the world hub.
Engine players now carry all 49 graded attributes instead of nine averaged drivers. Each probability in the point engine reads specific attributes: aces read first-serve power and serve variety against the returner's return quality and depth; first-serve percentage reads first-serve accuracy; double faults read second-serve quality and composure, and increase with the server's first-serve power; rally outcomes read groundstrokes, movement, and consistency; conditions read each player's wind, heat, indoor, and crowd tolerances.
Box scores gained a forced-error category, calibrated against charted real college data. Player generation gained playing-style profiles: attributes shift in coherent groups with overall rating held constant, which produces genuine doubles specialists. About a third of programs field a doubles player outside their singles top six.
Point-ending attribution
How points are labeled (winner, forced error, unforced error) was rebuilt three times in this release. The starting problem: deep-lineup players posted completed matches with zero winners, and an elite player posted 47 winners against 2 unforced errors. Both came from anchoring the label shares to an absolute talent reference; players far from the anchor saturated the clamps in opposite directions.
The final model is symmetric and matchup-anchored. Every rally end is a three-way split — the point-winner's winner, or the loser's forced or unforced error, on either side — driven by the gap between the hitter's attacking attributes and the misser's defensive attributes. There is no absolute-level term. Matched pairings produce the same statistical mix at every level: measured 31.7% / 30.0% / 28.5% winners for matched elite, mid, and weak pairings. This matches the reference data: winner/error mixes are approximately 32/41/27 for pro men and 29/37/34 for pro women, and Challenger box scores are statistically indistinguishable from ATP box scores.
Outcome probabilities were not changed by any of this and remain attribute-difference-driven. Measured win rates by overall-rating gap on rostered players: 10 points, 87–89%; 20 points, 98–100%; 30 or more points, 100% across 150 matches per cell.
Three invariants are now enforced by tests: every point is labeled exactly once (a player's points won equals their winners plus the opponent's errors and double faults); matched mixes stay level-blind; mismatches show the gap. The real-world reference data used for calibration is logged in the repo.
Injuries and retirements
The pro league now uses the same injury system as college: shared dice and store, per-league tables, durability-scaled rates. Previously it had no injuries.
Matches can end in retirement. A retirement follows an injury, occurs at 0.2% per completed singles match, and costs the retiring player the line regardless of the score, with dual points corrected when that flips a line. Injury state is wiped when a new save is created.
Injury volume is calibrated per team rather than per player on court: rolls scale by the number of competitors in the dual, so the expanded cards produce the same expected injuries per team per dual as the six-court game did. Retirement rolls scale the same way.
Pro league
The pro league gained: player development toward a peak (ratings previously did not change after graduation), playing-style archetypes drawn from real tennis eras and weighted for the league's format, club coaches who shape their players' styles, coaches drawn from people who exist in the save, and a movement economy with a draft surplus, roster locks, and alumni tracking. A 500 error on the player page for free agents without a club was fixed.
World structure
Every active division and gender universe advances together on one world clock. The Season Hub's separate advance button was removed; POST /world/advance is the only advance route, enforced by a test. Separate advance paths had allowed universes in one save to reach different weeks.
The offseason is a sequence of visible steps — awards, the international cups, the year rollover, the pro-league offseason, preseason — each advanced by one click. The cups and the pro offseason were previously interior steps of the rollover and could not be observed or run separately. Three regressions from this change were found and fixed in the same release.
Rules and data integrity
Service academies (Army, Navy, Air Force, Coast Guard, Merchant Marine) roster US citizens only, enforced through every player-placement path: base rosters, recruiting, all portals, walk-on fill, coach moves, and pro free agents. The Citadel and VMI are state institutions and are not gated.
D3/D4 regular-season and Fall season duals play every singles match to completion (the D3 format). The clinch still decides the winner; the remaining matches fill in the margin and give players completed matches on record. Postseason duals in all divisions keep clinch play.
Box scores show the order of finish.
Player match logs include doubles matches, with the partner and opposing pair listed. Previously the logs covered singles only.
The fall-portal slate enforces one move per player at the write layer, and a bug producing duplicate portal move rows was fixed.
New saves restart the international cups and pro leagues instead of carrying them over from a previous world.
The top navigation bar renders correctly at laptop widths.