⚾ Case Studies
MLB Case Studies — May 2026
Three signals from our live dataset, verbatim. Two settled at $1.00 and one settled at $0.00.
These three signals come from the public
/stats dataset of
1,381 entries. They were chosen to show
three different game-state shapes the detector fires on, and the third one lost. Nothing here is selected for outcome, and no case on this page tells you what a signal is worth: the entry price does that, and the arithmetic lives on the
break-even table. Full dataset available on request.
Case 1 — Mid-game lead, seventh inning
SEA @ KC
May 24, 2026 · 8:02 PM ET · signal fired during the game
Game state at signal
Top 7th · KC leads 5–2 · run diff +3
Model estimate (runs optimistic)
P(KC wins) = 95.4%
(empirical run-diff table)
Kalshi market at signal
KC contract $0.77 · implied 77.0%
Filters met
model minus market between 0.10 and 0.25 (the gate is capped at the top as well as the bottom: a wider difference is not a better call) · model ≥0.85 · not tied · KC ≤$0.90 · inning ≥3
Game result
SEA 6, KC 8 — KC won
Settlement
Contract resolved at
$1.00. Graded a win in the
public log.
WINgraded
Case 2 — One-run lead, bottom of the ninth
WSH @ ATL
May 25, 2026 · 12:35 AM ET · signal fired during the game
Game state at signal
Bottom 9th · WSH leads 2–1 · run diff +1
Model estimate (runs optimistic)
P(WSH wins) = 94.0%
(empirical run-diff table)
Kalshi market at signal
WSH contract $0.73 · implied 73.0%
Entry
Buy WSH YES @ $0.73
Game result
WSH 2, ATL 1 — WSH won
Settlement
Contract resolved at
$1.00. Graded a win in the
public log.
WINgraded
Case 3 — The loss. Extra innings on the road
DET @ CWS
May 30, 2026 · signal fired during the game
Game state at signal
Bottom 10th · DET leads by 1 · run diff +1
Model estimate (runs optimistic)
P(DET wins) = 100%
(the strongest output the table can produce)
Kalshi market at signal
DET contract $0.79 · implied 79.0%
Entry
Buy DET YES @ $0.79
Game result
CWS walked it off, 4–3 — Detroit lost
Settlement
Contract resolved at
$0.00. The entire stake was gone. Graded a loss in the
public log the same day, like every other loss.
−100%of the stake
This is the cleanest illustration of settlement risk we have: even a 100% model read lost. A "100%" estimate means the historical game states that looked like this one were essentially always converted, not that this one had to be. The model computes from run differential, innings, and outs. It does not see a tiring reliever, who is due up, or a hanging slider. Nothing is decided until the contract settles.
What these three show
- Shape variety, not outcome variety. A mid-game lead, a one-run ninth, and an extra-inning road game: three distinct game-state shapes the detector fires on. Two settled at $1.00 and one settled at $0.00.
- The same kind of read produced opposite outcomes. Case 3 carried the model's strongest possible estimate and still lost outright. The model was not "right" twice and "wrong" once. It reported a disagreement with the price three times, and baseball did the rest.
- Nothing on this page tells you what a signal is worth. That is arithmetic on the entry price you actually get, and it is laid out on the break-even table. About one in 5 of these games ends like Chicago, not like Kansas City, and the underlying record over the full 1,381-entry dataset is on /stats.
Honest caveat about the data above
When this case study was written, our logging captured signal_at, model_p, market_p, gap, and outcome but not the Kalshi price over time after a signal fired, so the numbers above describe the entry and the resolution, not the seconds in between. We have since added that instrumentation, and it did not show what we expected: the median move on the alerted side was 0.0000¢ at every horizon out to five minutes, with a flat spread and rising depth. The correction is here. Read these three cases as entries and settlements, not as evidence of a price that corrected.
Risk reality. Three signals do not constitute a win rate. About one in five MLB signals loses 100%, sometimes in clusters. A high win rate describes the past — it does not by itself make anyone money: your entry price, bet sizing, and discipline through losing streaks decide your actual outcome.
See the full risk disclosure on /method/. This is not investment advice. Sports betting and prediction-market trading involve substantial risk — bet only with capital you can afford to lose.
Want the live alerts in your pocket?
The full methodology is on /method/. Pricing and plans below.