Does the model's predicted probability match actual outcomes? This is the only question that matters long-term.
CONFIRMED edges (books + model agree) should outperform MODEL ONLY edges. If they don't, the book consensus isn't adding real value.
| Signal | Win Rate | % | W/L | Volume |
|---|
When the DataGolf model says 30%, do YES contracts actually resolve ~30% of the time? A well-calibrated model hugs the diagonal.
| Model Prob Bucket | Bar (grey=expected, colored=actual) | Actual | Expected | Diff | W/L |
|---|
Larger edges should win more often — or they shouldn't be trusted. Grouped by market type and edge size.
| Market Type | Edge Bucket | Win Rate Bar | Win % | W/L |
|---|
| MAKECUT | 5–10% |
|
69% | 25/36 |
| MAKECUT | 10–15% |
|
48% | 14/29 |
| MAKECUT | 15–20% |
|
65% | 15/23 |
| MAKECUT | 20%+ |
|
40% | 17/42 |
| R1LEAD | 0–5% |
|
0% | 0/8 |
| R1LEAD | 10–15% |
|
100% | 1/1 |
| R2LEAD | 0–5% |
|
8% | 1/13 |
| R3LEAD | 0–5% |
|
5% | 1/20 |
| TOP10 | 5–10% |
|
89% | 124/140 |
| TOP10 | 10–15% |
|
81% | 48/59 |
| TOP10 | 15–20% |
|
77% | 24/31 |
| TOP10 | 20%+ |
|
80% | 28/35 |
| TOP20 | 5–10% |
|
82% | 108/132 |
| TOP20 | 10–15% |
|
84% | 81/97 |
| TOP20 | 15–20% |
|
93% | 26/28 |
| TOP20 | 20%+ |
|
89% | 33/37 |
| TOP5 | 5–10% |
|
95% | 86/91 |
| TOP5 | 10–15% |
|
69% | 27/39 |
| TOP5 | 15–20% |
|
59% | 10/17 |
| TOP5 | 20%+ |
|
83% | 20/24 |
| TOUR | 0–5% |
|
21% | 31/145 |
| TOUR | 5–10% |
|
87% | 40/46 |
| TOUR | 10–15% |
|
89% | 8/9 |
| TOUR | 15–20% |
|
50% | 1/2 |
| TOUR | 20%+ |
|
33% | 1/3 |
| UNKNOWN | 0–5% |
|
38% | 3/8 |
Which market types are generating real edge vs noise? P&L is from the paper agent's simulated $250 bankroll.
| Market | Bets | Settled | Win Rate | Avg Edge | P&L |
|---|
| TOP5 | 171 | 171 | 84% | 11.8% | $+1135.27 |
| TOP10 | 265 | 265 | 85% | 12.8% | $+2102.98 |
| TOUR | 205 | 205 | 40% | 3.6% | $+1116.30 |
| TOP20 | 294 | 294 | 84% | 12.4% | $+3503.45 |
| UNKNOWN | 8 | 8 | 38% | 0.0% | $-1.32 |
| MAKECUT | 130 | 130 | 55% | 17.7% | $+216.85 |
| R2LEAD | 13 | 13 | 8% | 0.0% | $-7.58 |
| R3LEAD | 20 | 20 | 5% | 0.0% | $-19.24 |
| R1LEAD | 9 | 9 | 11% | 1.8% | $-4.00 |
Edge must be consistent across tournaments — not just one hot week. Low-n tournaments (<10 settled) are noise. Look for positive ROI across 3+ events.
| Tournament | Settled | W/L | Win Rate | Avg Edge | ROI | P&L |
|---|---|---|---|---|---|---|
| BAC26 | 100 | 87/100 | 87% | 12.2% | +99.8% | $+1292.44 |
| BAOUC26 | 116 | 106/116 | 91% | 12.1% | +92.1% | $+987.94 |
| BICA26 | 209 | 168/209 | 80% | 11.6% | +85.9% | $+1478.10 |
| BMC26 | 125 | 99/125 | 79% | 8.3% | +103.4% | $+2258.34 |
| BUBC26 | 70 | 61/70 | 87% | 10.2% | +90.9% | $+210.61 |
| FESJC26 | 49 | 38/49 | 78% | 16.0% | +89.4% | $+534.26 |
| JODC26 | 57 | 16/57 | 28% | 6.6% | +61.7% | $+115.71 |
| ROC26 | 73 | 24/73 | 33% | 9.1% | +60.3% | $+170.30 |
| THOC26 | 122 | 66/122 | 54% | 10.7% | +9.7% | $+162.68 |
| TOC26 | 45 | 38/45 | 84% | 7.2% | +74.3% | $+264.63 |
| TRAV26 | 12 | 2/12 | 17% | 13.7% | -71.6% | $-37.53 |
| WYC26 | 97 | 53/97 | 55% | 12.4% | +74.8% | $+434.85 |
CLV measures whether you consistently get better prices than where markets close. Positive avg CLV is the purest proof of edge — independent of outcome variance.
CLV data accumulates as markets settle. Requires closing prices recorded at market resolution.
Calibration tells you if the model's confidence is justified — not just whether it's profitable.
GOOD SIGNAL
Actual win rate tracks close to model probability bucket. The 30% bucket wins ~30% of the time. Edges are real and the model is honest.
WATCH FOR
High edge buckets with low win rates suggest model overconfidence. Small sample sizes (n < 20) make any bucket unreliable — wait for more data.
Avg predicted edge: 10.9% · Total wagered: $37332.61 · Overall win rate: 69.3% (773/1115 settled)
Monitors ESPN, Kalshi, and DataGolf feeds. Alerts fire via webhook + email when a source goes down during a live event.
⚠ 7 degradation events recorded this session
| Time | Sources | Issues |
|---|---|---|
| Oct 27, 2026 2:27 PM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 27, 2026 11:18 AM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 22, 2026 2:18 AM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 18, 2026 7:57 PM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 17, 2026 4:45 PM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 16, 2026 2:36 AM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |
| Oct 16, 2026 2:36 AM ET | ✓ ESPN: LIVE✓ KALSHI: LIVE✗ DATAGOLF: DOWN | DataGolf predictions DOWN |