BUILD IN PUBLIC · EDGE SCOREBOARD

Does the bot actually have edge?

We're building this in public. Here's whether the bot actually has edge — measured, not claimed. Every lane starts in Shadow: the bot logs what it would bet and we score it against the closing price and the final result, without risking money. A lane only earns real stakes when the numbers below say so.

Shadow intents appear only after their event starts. Live bets appear only after they fill. Nothing here is published before then. Updated 2026-10-08 19:14 UTC.

Where we started (2026-10-05)

MeasureThenTarget
Realized edge, 90d, clean+0.2pp (MLB, 99 bets)> +2pp with 95% CI-low > 0
Claimed edge vs realized12.3pp claimed, 0.0pp realizedCalibration slope 0.8-1.2
CLV beat rate35.7%> 55%, avg CLV > 0
Model edge to result slope0.06>= 0.7

A 2026-10-07 backtest found that, priced correctly, the model has about no edge over Kalshi on NHL puck lines/totals and MLB run lines, and MLB totals run market-only. Expect bets to be rare. That is the point of measuring first.

No measured intents yet

The shadow scan logs every candidate on the slate. Rows appear here once their game starts, and each one is scored at the close and again at the final.

Lanes

A lane is one sport and market type. Live means real filled bets. Shadow means what the bot would have bet (cleared its edge gate), with no money behind it.

MLB · Moneyline

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

MLB · Moneyline reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

MLB · Spread

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

MLB · Spread reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

MLB · Total

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

MLB · Total reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

NHL · Spread

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

NHL · Spread reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

Serie A · Moneyline

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

Serie A · Moneyline reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

Serie A · Spread

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

Serie A · Spread reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

Serie A · Total

Shadow Logged only. No money.

Live · real fills

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Shadow · would have bet

Settled / with close0 / 0
Beat the closen/a
Average CLVn/a
Realized edgen/a
Win rate vs avg pricen/a vs n/a
Flat-unit ROI (after fees)n/a

Last 30 days: beat the close n/a (n=0), realized edge n/a (n=0).

Calibration · all candidates

Serie A · Total reliability diagram Predicted probability against how often it happened, for the model and for the market price. Points on the dashed diagonal are perfectly calibrated. Model Brier n/a, market Brier n/a; lower is better. 0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted No settled candidates yet
Model (circle) Market price (square)

Brier: model n/a vs market n/a · slope n/a · n=0

Chart as a table
SourceBinPredictedHappenedn
What it takes to reach Probation
  • Not yet: Shadow intents with a measured close >= 50 (now: 0)
  • Not yet: Shadow CLV beat rate > 50% (now: n/a)

Recent intents

Every row, wins and losses alike. CLV is the closing price minus our entry price; positive means we got a better price than the market closed at. * marks a close read just after the start or from the last trade.

Nothing to show yet. Shadow rows appear when their game starts; live bets appear after they fill.

How to read this

  • Beat the close (CLV) is the fastest honest signal. If we buy at 45¢ and the market closes at 50¢, we got a better price than the market's final view. A lane that can't beat the close over many bets doesn't have edge, whatever its record says.
  • Calibration: when the model says 60%, it should happen about 60% of the time. Points on the dashed line are calibrated. Brier score is the average squared miss; if the model's Brier isn't lower than the market's, the model isn't adding information.
  • Realized edge is win rate minus the average price paid. A 95% confidence interval that crosses zero means we can't tell it apart from luck yet.
  • Stages: Shadow (no money) → Probation (minimum bets, after shadow beats the close on 50+ intents) → Certified (150+ clean bets, beats the close with confidence, calibrated, edge above zero). Small samples move nothing.
  • Results are past results. They do not promise anything about future ones.

See every published pick and result →   Raw data (JSON) →

21+ only. For entertainment and information; not financial advice. Gambling problem? Call 1-800-GAMBLER.

SIGNAL · LIVE