What our first 55 graded NBA in-game signals actually showed — a thin, single-period sample from one stretch of last season, the Finals included — and how KXNBAGAME markets move in real time. Numbers first, losses shown, ahead of the 2026-27 season.
DegenHedge measures one thing: the delay between a game changing and a Kalshi market re-pricing to match it. On baseball we've graded hundreds of games. On basketball, the record so far is small: 55 graded in-game games, 43 wins and 12 losses — a 78.2% first-signal win rate, which is the same as saying about 1 in 4.6 of those games lost. That is a preliminary read, not an established rate, and this whole page treats it that way.
This is a data guide, not a betting guide. We are publishing it in the offseason because the market for "kalshi nba" information is nearly empty and we'd rather stake out the honest version of it than let a hype version fill the gap. Nothing here is a pick, a prediction, or advice. When the 2026-27 season resumes in late October, the alerts start again; until then, there is nothing to act on and everything to read.
A yes/no event contract on an NBA game that settles at $1 if the named team wins and $0 if it doesn't. While the game is live, the price behaves like a market-implied win probability — a side trading at 60¢ means the order book is collectively pricing roughly a 60% chance.
Here's the key point: the win probability isn't set by the order book — it's set by the game. A 12-0 run in the third quarter changes the true probability the moment it happens. The price only changes when someone actually crosses the spread with an order. Between those two moments, the scoreboard and the screen disagree.
That window of disagreement is the Kalshi lag, and it exists for the same structural reason in basketball as in baseball: in-game books are thin, no market maker is obligated to keep quotes current through every possession, and human traders react at human speed — with broadcast delay stacked on top. None of that is a knock on Kalshi. It's the default behavior of a retail-driven market attached to a fast game. Our signal anatomy page breaks down exactly what we log each time the gap opens.
For NBA, the model is a Goldman-Stern style in-game win-probability calculation: it estimates P(home wins) from score margin and time remaining, calibrated from thousands of historical games. It does not know who is hot, who is in foul trouble, or which starter just tweaked an ankle — it reads margin and clock, nothing more. That model probability is compared against the live Kalshi KXNBAGAME price for the same team, refreshed continuously, and a signal fires only when the two diverge by enough. Full construction is on the method page.
Two things stood out in the NBA sample, and both cut in the same direction. First, the gaps ran wider: the median model–market gap at fire was 17.5 percentage points, wider than MLB. Second, the entry prices ran cheaper: the median Kalshi price of the picked side was 50¢, cheaper than MLB. Wider gaps and cheaper entries sound appealing — but they came on a fraction of the games. Our MLB record spans 432 graded games at 81.2% (with roughly 1 in 5.4 still losing); the NBA record is 55. That's the sample talking, not a stronger edge — and the only edge here is speed, knowing the game state before the price reflects it.
Read the sample plainly. The 78.2% figure (43W–12L, about 1 in 4.6 games lost) comes from 55 games logged across a single stretch — April 10 to June 14, 2026 — including the Finals. That is one slice of one season, not a full campaign, and one hot or cold run inside it moves the number more than it should. It is nowhere near the statistical footing of the 429-game MLB record. Treat it as a first look you'll want to see confirmed over a much larger sample, not as a number to bank on.
The wider NBA gaps add a second reason for caution. On our MLB data, the buckets where the market disagrees with the model most grade worse, not better — because part of that disagreement isn't lag, it's information the model can't see. NBA's median gap of 17.5pp sits in exactly that zone. Here are three real signals from the log — one win, two losses:
| Date | Game | Model | Kalshi at fire | Gap | Final | Result |
|---|---|---|---|---|---|---|
| 2026-04-13 | MEM @ HOU | 100% HOU | 50¢ | 50 pp | HOU 132–101 | WIN |
| 2026-04-16 | GSW @ LAC | 79% LAC | 55¢ | 24 pp | GSW 126–121 | LOSS |
| 2026-04-10 | PHI @ IND | 66% IND | 13¢ | 53 pp | PHI won | LOSS |
The win, up close. April 13 in Houston: the model read HOU at 100% from margin and clock while Kalshi still quoted the Rockets at 50¢ — a 50-point spread. Nothing clever happened next; Houston simply closed out a 132–101 win the game state already implied, and the price caught up at settlement. That's the lag in its purest form: a timestamp on a price that hadn't reacted yet.
GSW @ LAC (loss). The model read the Clippers at 79% with Kalshi at 55¢ — a favorite-model call, 24 points of gap. Golden State won anyway, 126–121. Settlement risk is always live, even on the side the model likes.
PHI @ IND (loss). The model read Indiana at 66% — but the market had the Pacers at just 13¢. That's a 53-point gap, and the market's number, not the model's, was right: Philadelphia won. When the book disagrees that hard, part of the gap isn't lag at all — a wide gap can mean the market knows something the model does not (an injury, a rotation, a matchup the margin-and-clock read can't see). A bigger gap is not a safer signal. It is sometimes a warning.
The right move in the offseason is measurement, not action. There are no NBA games to trade, and the honest thing to do with a 55-game read is watch it build live as the new season generates a real sample — for free.
The proof channel at t.me/DegenHedgeProof posts every signal as a locked entry shortly after it fires — side hidden so it can't tip anyone's hand, then revealed at settlement, losses included. Pair it with the public results log and the raw CSV. When NBA resumes in late October, you'll be able to watch the sample grow past 55 in real time and judge the 78.2% preliminary read for yourself.
The same mechanism runs year-round on Kalshi's 15-minute BTC markets, where the published filter has graded 92.8% — which still means roughly 1 in 14 published windows loses. Different market, same idea: measure the lag, show every result.
The paid alerts exist for one reason: to deliver these same signals to your Telegram the moment they fire, instead of locked-and-delayed. That's a convenience for when games are live — not something to buy for a season that hasn't started.
The free path above is the primary one, and it's genuinely enough to evaluate everything on this page. If you'd rather line up now, here are the two relevant subscriptions — with the offseason reality stated plainly.
Every signal can lose your entire stake, and the NBA sample proving any of this is small: 55 games from one stretch of last season. The 78.2% figure means 43W–12L — about 1 in 4.6 games lost, and losses do not arrive politely spaced; they cluster. Hit rate is not profit — your entry price, your sizing, and your discipline decide your outcome, and at cheap 50¢ entries a single loss still costs a full contract.
These signals are a fast Kalshi-lag detector, not an independent forecaster of who wins a basketball game — a wider gap can be the market seeing something the model can't. This is informational and entertainment content, not financial or betting advice. Must be 18+ (some platforms require 21+). Past performance does not guarantee future results. If gambling is a problem for you, call 1-800-GAMBLER.
Read the full model details →