Methodology
Two separate questions: can this trade go through under the rules we model, and who comes out ahead?
Eligibility and valuation are separate
Eligibility runs rule checks from the league's agreement for the transaction date. Valuation estimates who benefits. A trade can be great value and still blocked — in that case valuation is labeled hypothetical.
Results are not an official league ruling.
Four statuses
- Pass The check ran and the trade satisfies it.
- Fail The check ran and the trade breaks it.
- Conditional Satisfied only if something happens (consent, physicals, roster moves).
- Cannot verify We can't check it from public data. Unknown never counts as a pass.
Verdict precedence
- Any Fail → Blocked
- Otherwise, any required Cannot verify → Cannot verify
- Otherwise, any Conditional → Conditional
- Otherwise → Passes implemented checks
Value units and horizons
Each league uses its own unit: NFL EPA, NBA wins, NHL GAR, MLB WAR, converted to dollars at demo market rates.
Immediate = rest of this season. Next 3 seasons = the following three. Long-term = later seasons, discounted.
Every figure is a range: p10–p90 around a median (p50).
Six components and strategy weights
| Component | Contend | Balanced | Rebuild |
|---|---|---|---|
| Immediate roster impact | 35 | 20 | 5 |
| Long-term asset impact | 5 | 20 | 35 |
| Contract value | 20 | 25 | 25 |
| Financial flexibility | 10 | 10 | 15 |
| Strategic fit | 20 | 15 | 10 |
| Risk | 10 | 10 | 10 |
Money is counted once, in contract value (production value minus cost, including cash). Financial flexibility only measures threshold effects — crossing a tax line, apron or cap — so dollars aren't double counted.
Real-life test: would this trade happen?
A summary on top of the three separate answers. It never changes them. A trade is Realistic only when all of these hold:
- League rules: the eligibility verdict is not Blocked (conditions such as consent or a roster move are listed).
- Each front office says yes: the trade scores at least 50/100 for that team's own strategy (contend, balanced or rebuild). 45–49 means it would ask for a sweetener; below 45 it would say no.
- The players would play: each incoming player's place on his new depth chart. A star traded to a team with a better player at his spot becomes a backup, and real teams rarely pay for that.
- Calendar: days left before the deadline (MLB: the Dec. 1, 2026 labor-agreement expiry).
Rules covered per league, with the article each check encodes, are on the Rules page: NFL cap accounting with bonus acceleration, roster limits, picks and the Nov. 10, 2026 deadline; NBA salary matching, aprons, hard caps, aggregation and cash limits, the Stepien rule and the Feb. 11, 2027 deadline; NHL cap accrual, retention limits, the 50-contract limit, the 23-man active roster and the Mar. 1, 2027 deadline; MLB 40-man roster, 10-and-5 and no-trade rights, CBT, competitive-balance picks, PTBNL rules and the Basic Agreement expiry.
Season simulator: with the trade vs without it
What it does. It plays the rest of the season 10,000 times with today's rosters and 10,000 times with the trade, then compares projected record, playoff, division, top-seed, finals and title odds for every team. Both versions use the same random seasons (same bounces, same upsets), so the difference comes from the trade, not from luck.
Team strength. Each player has a projected value for a full-time role (NFL EPA, NBA wins, NHL goals above replacement, MLB WAR), converted to wins: about 37 points per NFL win, 6 goals per NHL win, and NBA wins and MLB WAR are already wins. Within each position group, players are ranked and get a usage weight (starters 100%, backups less). When a trade moves a player, everyone on both depth charts is re-ranked, so adding a player can only help, losing one can only hurt, and the size depends on who he displaces or who replaces him.
Games. Each game's win chance comes from the two teams' strength gap plus home advantage (a logistic curve). Every simulated season also draws a random error in each team's true strength and in each traded player's value, so the odds stay humble. NHL games can go to overtime (the loser gets a point, about 23% of games).
Playoffs. Current formats: NFL 7 teams per conference with one bye; NBA top 6 plus the 7–10 play-in; NHL top 3 in each division plus 2 wild cards; MLB 3 division winners plus 3 wild cards with two byes. Tiebreakers are simplified (random among identical records).
Data. Every league uses real schedules and results, refreshed automatically once a day, and team strength is an Elo rating built only from game scores. NFL: the 2026 season from nflverse (CC-BY 4.0). NBA and NHL: the 2026-27 seasons (NHL: 84 games) from API-Sports. The NBA lists 80 games per team until the NBA Cup group stage sets the last two, so those two are placeholders against same-conference opponents until they are scheduled. MLB: the 2026 season is over, so the simulator plays all of 2027, with team strength from real 2022–2026 results (through the World Series) pulled toward the average for the new season; the 2027 schedule follows MLB's scheduling formula until API-Sports lists all 2,430 games. If a data feed stops updating, the simulator keeps the last good data, shows its date, and simulates the games after it. Home advantage, the NHL overtime/shootout rate and how far apart teams really are come from real 2022–2024 results. All players, contracts, projections, picks and prospects are demo.
How well it works (NFL backtest). We stopped each season from 2005 to 2025 after Week 8, simulated the rest, and scored the playoff odds against what happened. Fitted on 2005–2016 and tested on 2017–2025: playoff-odds Brier score 0.130 vs 0.131 for treating every remaining game as a coin flip (lower is better), final-wins error 1.80 vs 2.05 wins, and title log loss 2.38 vs 2.77. Honest read: the current standings already carry most of the playoff signal, so the model's edge there is small; it is clearly better at final records and title odds. The trade part (player values) cannot be backtested with demo players.
Backtest on real NBA, NHL and MLB seasons. Using real 2022–2024 results, we rebuilt team strength from scores, stopped each season at two dates, simulated the rest and compared with what happened. Fitted on 2023, checked on 2024. Final-wins error, model vs coin-flip games: NBA from Nov 1, 10.3 vs 12.6 wins; NHL from Nov 1, 5.4 vs 6.8; MLB from opening day, 9.3 vs 12.3. From mid-season the model still predicts final records better (NBA 4.8 vs 7.5, NHL 3.0 vs 4.0, MLB 5.7 vs 6.6), but its playoff odds are about the same as coin-flip games (NBA 0.098 vs 0.098, NHL 0.095 vs 0.093), because the standings already tell most of the story by then. Two seasons per league is a small sample.
Confidence
Combines data completeness (missing contracts or projections lower it), rules coverage (share of applicable checks we could run), projection uncertainty and a sensitivity note about what would change the result.
Make it work
Suggestions are generated, then re-run through the full engine. Each shows its own eligibility verdict. Passing the checks doesn't mean the other team would accept.
Snapshots
Share links store an immutable snapshot: trade, results, data version, rule set and engine versions, and a hash. Reanalyzing creates a new child snapshot instead of changing the original.
AI explanations
Optional and capped daily. AI only rewrites the computed results; its text is validated against the numbers. If it's unavailable or fails validation you get the template explanation instead.